When the guns went quiet on the evening of May 10, 2025, the four days of fighting between India and Pakistan had killed soldiers in uniform, but it had also killed thirteen-year-old twins who had moved towns for a better school, a doctor’s patients inside a mosque at dawn prayer, a district officer who stepped outside to check on his staff, and a farmer in a Punjab field who never saw the loitering munition that found him. Civilian casualties on both sides of the 2025 India-Pakistan conflict became the most politically charged set of numbers the crisis produced, and at the same time the least independently confirmed.

The short answer to the question most readers arrive with is that both countries killed each other’s civilians, and neither country’s count can be fully trusted. Pakistan said that India’s missile strikes on the night of May 6 into May 7 killed civilians in Bahawalpur, Muridke, Muzaffarabad and Kotli, and its military first put the figure at thirty-one before raising it days later to forty civilians, among them seven women and fifteen children. India said that Pakistani artillery and mortar fire across the Line of Control, together with drones and a loitering munition, killed civilians in Poonch, Rajouri, Uri, Kupwara and one town in Punjab, and its official tally settled at roughly twenty-one civilians alongside its uniformed dead. Independent investigators who later studied the firing along the Line of Control suggested the true civilian total was higher than either government’s headline number, because much of the killing happened in border villages where nobody was counting in real time.

What follows is an attempt to hold both sides of that ledger in view at once, without flinching from either column and without letting either government’s framing become the article’s framing. New Delhi insists its precision weapons spared noncombatants and that the deaths in Pakistan were either militants or the unavoidable shadow of striking infrastructure embedded in towns. Islamabad insists the strikes were aimed at mosques and homes. The Indian side blames Pakistani shelling for the bodies in Poonch and calls it deliberate targeting of towns; the Pakistani side calls the same shelling counter-battery fire against Indian positions placed near civilians. The numbers are contested, the methodology behind them is thin, and the verification gap is wide enough to swallow the entire debate. The honest account is not a single death toll. It is a description of what was claimed, what was documented, what could be checked, and what could not, followed by the harder question of what a doctrine built on precision is supposed to do with the children it still killed.

Background and Triggers

The civilians who died in May 2025 died because of a sequence that began in a high meadow in Indian-administered Kashmir. On April 22, gunmen walked into the Baisaran valley above Pahalgam, a grassland reachable only on foot or by pony, and killed twenty-six people, almost all of them Hindu tourists, several of them shot in front of their wives. The minute-by-minute reconstruction of the Pahalgam attack shows an assault designed less for military effect than for communal wound, and the full explainer on the Pahalgam attack traces how a group calling itself The Resistance Front claimed the killing and then tried to walk the claim back. India blamed Pakistan for backing the attackers. Islamabad denied involvement and offered to cooperate with a neutral inquiry. Neither position moved the other.

For the next two weeks, the two governments climbed a ladder of pressure rather than firing immediately. India suspended the Indus Waters Treaty, closed the Attari-Wagah land crossing, cut off bilateral trade, and revoked Pakistani visas. The fourteen-day record of India’s response to Pahalgam lays out each step as a separate signal, and each signal went unanswered in the way New Delhi said it wanted. On the night of May 6 into the early hours of May 7, the diplomatic phase ended. Indian aircraft and missiles struck nine sites that New Delhi described as terrorist infrastructure belonging to Jaish-e-Mohammed, Lashkar-e-Taiba and Hizbul Mujahideen. The detailed account of Operation Sindoor’s twenty-three minutes and the complete operational guide to Operation Sindoor both describe a fast, multi-target campaign that the Indian government called focused, measured and non-escalatory.

Two of those fourteen days of pressure had a direct civilian dimension that is worth pausing on, because they show that the human cost of the crisis did not begin with the missiles. The suspension of the Indus Waters Treaty was, on its face, a diplomatic and legal measure, but water is not an abstraction in the lives of people downstream. In the days that followed the suspension, local media in Pakistan-administered Kashmir reported sudden changes in the flow of the rivers the treaty governs, water released without the usual notification in one stretch, a sharp drop in another, the riverbed at one Pakistani town visibly drying. Whether or not those reports were precise, they registered as a civilian alarm: farmers and townspeople reading the rivers and concluding that the water itself had become a lever. The closure of the Attari-Wagah crossing, similarly, stranded families. People who had crossed with valid documents were given a deadline to return, and the closure separated households that happened to be on the wrong side of a border that had, until then, been routinely passable. None of this killed anyone. All of it is part of the civilian record, because it is the part of the crisis in which ordinary people first felt the machinery move.

The strategic geography of the strike targets is where the civilian story becomes, in a sense, predictable in advance. The nine sites were not chosen at random, and the logic of choosing them was not a civilian logic. India’s planners selected locations with deep, documented ties to the organizations New Delhi held responsible for Pahalgam, and the most significant of those locations had been, for years, considered too sensitive to strike precisely because of where they sat. The Jaish-e-Mohammed seminary at Bahawalpur and the Lashkar-e-Taiba headquarters campus at Muridke were not hidden mountain camps. They were institutions inside the populous Pakistani heartland of Punjab, woven into towns, surrounded by the ordinary life of cities. To strike them was, by definition, to strike inside a populated area. New Delhi’s argument was that this was precisely the point, that the value of hitting Bahawalpur and Muridke lay in demonstrating that no safe haven was geographically beyond reach, that the old assumption of sanctuary in the heartland was dead. Whatever the merits of that argument as strategy, it carried an unavoidable civilian corollary. A doctrine that selects targets specifically because they sit in places previously considered too sensitive to hit has, in the same motion, selected targets that sit among civilians. The civilian exposure was not a miscalculation. It was structural to the targeting choice, and it was visible the moment the coordinates were set.

Whatever the strikes were aimed at, they crossed a line. Two of the nine targets, the seminary complex at Bahawalpur and the headquarters compound at Muridke, sit deep inside Pakistani Punjab, far from the Line of Control, in the country’s populous heartland. The last time India had struck that deep was the war of 1971. The strikes were also, crucially, aimed at sites that sat among people. A seminary is not an isolated bunker. A headquarters compound that doubles as a school, a clinic and a mosque is woven into a town. The moment New Delhi chose those coordinates, the question of civilian casualties stopped being hypothetical, regardless of how precise the weapons were.

Pakistan’s retaliation widened the civilian exposure in the other direction. Within hours, Pakistani artillery and mortar units opened fire across the Line of Control into the Jammu region, with the district of Poonch taking the worst of it. Over the following days the fighting expanded into the first drone battle between the two nuclear-armed neighbours, then into Pakistan’s broader retaliatory campaign, Operation Bunyan-un-Marsoos, before three separate diplomatic channels converged on the ceasefire negotiated on May 10. The day-by-day timeline of the 2025 conflict holds the full sequence. This article steps out of that sequence to look at one strand of it. The crisis had a military story, a diplomatic story and a weapons story. It also had a civilian story, and the civilian story does not belong to either capital.

The trigger conditions matter because they shape how each death gets read. A civilian killed in Bahawalpur is, to New Delhi, the regrettable cost of a justified strike on a terror seminary, and to Islamabad, proof that India bombed a mosque. A civilian killed in Poonch is, to New Delhi, evidence that Pakistan deliberately shells towns, and to Islamabad, the unfortunate result of counter-battery fire against Indian guns parked among houses. The same fact carries opposite meanings depending on which government is narrating it. Any honest account of the human cost has to begin by refusing to adopt either narration as its own.

The Strikes That Killed Across the Border

The first wave of civilian deaths came from the Indian missile and air strikes on the night of May 6 into May 7. The Indian government’s position was consistent and specific: the operation hit nine sites tied to militant organizations, no Pakistani military facility was targeted, and the locations were selected to avoid damage to civilian infrastructure. Pakistan’s position was equally consistent and pointed in the opposite direction: the strikes hit civilian areas, including mosques, in at least six locations, and they killed ordinary people. The two accounts cannot both be fully true, and the places where they collide are worth examining one by one.

Bahawalpur, a city in southern Punjab, drew the most attention. The strike there hit the Jamia Masjid Subhan Allah, a mosque attached to a seminary complex long associated with Jaish-e-Mohammed. A doctor at a hospital near the site reported that thirteen people were killed, among them a child. The complication that makes Bahawalpur so contested is that one of the people who confirmed deaths there was Masood Azhar himself, the founder of Jaish-e-Mohammed, who said publicly that ten members of his family and four of his aides had been killed in the strike on the group’s headquarters. That statement does two things at once. It supports the Indian claim that the target was a genuine militant nerve centre, because the head of the organization treated it as his own. It also supports the Pakistani claim that the dead included people who were not fighters, because family members inside a residential compound are not, by any standard, combatants. Bahawalpur is the clearest single illustration of why the casualty argument never resolves: the same strike can be a precise hit on a terror headquarters and a strike that killed a child and a household, and it can be both of those things in the same minute.

Muridke, near Lahore, was the second deep-Punjab target. The strike there hit the Markaz-e-Taiba, a sprawling compound founded by Lashkar-e-Taiba’s co-founder that includes a school, a college and a medical clinic alongside its mosque, and that has functioned for years as the organization’s central campus. A leading Pakistani newspaper reported that at least three civilians were killed at Muridke. Here the embedded-infrastructure problem is at its starkest. A compound that contains a working school and a clinic is, by design, hard to strike without hitting something civilian, and that design is not accidental. Organizations that build their headquarters around schools and clinics are, in part, building a shield. That does not make the dead at Muridke any less dead. It does mean the moral weight of their deaths is shared between the side that fired the missile and the side that chose to headquarter a banned militant group inside a campus full of students.

The Muridke complex deserves a closer look precisely because it is the cleanest example of the problem the entire conflict kept producing. International coverage identified the Markaz-e-Taiba as the single most significant of the sites India struck, the institutional heart of the organization rather than a peripheral facility. The compound’s civilian functions were not a disguise layered over a military base; they were genuinely operational, a real school, a real clinic, used by real people, run by an organization that the Pakistani state had formally banned years earlier and yet had allowed to keep operating its campus. That is the contradiction the casualty figure sits inside. When India struck Muridke, it struck a place that was simultaneously a banned terror group’s headquarters and a functioning civilian campus, and the reason it was both at once is that the Pakistani state had permitted it to be both at once. The civilians killed there are a cost. The question the figure cannot answer on its own is how that cost should be distributed between the missile and the decade of official tolerance that put a school on top of a headquarters.

In Pakistan-administered Kashmir, two more sites were hit. The Masjid Syedna Bilal in Muzaffarabad and a mosque in Kotli that India identified as a militant marker were both struck, and Pakistan said five civilians were killed across those two locations. Muzaffarabad and Kotli sit closer to the Line of Control than the deep-Punjab targets, in the contested territory itself, and the strikes there carried a particular charge because they hit religious buildings in a region where the line between a mosque and a militant gathering point has long been deliberately blurred by the organizations that operate there. India’s framing was that these were markers, recognized nodes in a militant geography. Pakistan’s framing was that they were mosques and the dead were worshippers. The five civilians Pakistan reported across the two sites are, like every other figure in this section, a state-issued number that no neutral party stood at the rubble to confirm, and the religious character of the targets guaranteed that the strikes would be read, inside Pakistan and across the wider Muslim world, as strikes on places of worship regardless of what New Delhi said the buildings really were.

Narowal and Sialkot in Punjab were also named among the struck sites. Across all of them, the pattern repeats: India names a militant association for the location, Pakistan names the civilians who died there, and no neutral party stood at the rubble with a clipboard.

Pakistan’s official accounting of the strikes evolved over the days that followed, and the evolution itself became part of the dispute. The military’s first public figure for the May 7 strikes was thirty-one civilians killed and dozens wounded. Several days later, after the fighting had ended, the military’s media wing revised the toll upward, describing forty civilians killed, among them seven women and fifteen children, alongside eleven members of the armed forces, with scores more wounded. The revision can be read two ways. The generous reading is that early counts in a fast crisis are always provisional, hospitals report at different speeds, and the later number reflects a fuller picture. The skeptical reading is that a government locked in an information war has an incentive to let a casualty figure grow. Both readings are available because the underlying records were never opened to anyone outside the Pakistani state.

There is one more layer to the cross-border deaths that complicates the civilian count from the Indian direction. New Delhi’s stated objective was the infrastructure of the banned organizations, and New Delhi claimed to have killed more than a hundred militants in the strikes. Islamabad rejected that figure and said the dead were civilians. The question of militant versus civilian status is its own dispute, examined elsewhere in this series, but it bleeds directly into the civilian count, because every body that one capital calls a fighter and the other calls a noncombatant is a body that cannot be cleanly assigned. The result is a Pakistani civilian toll that is bracketed by two unverifiable claims: an Indian claim that most of the dead were fighters, and a Pakistani claim that almost none of them were. The truth almost certainly sits between those poles. The point of an honest account is to say plainly that nobody outside the two governments knows where between the poles it sits.

The timing of the strikes deepened the civilian exposure in a way that is easy to miss. The operation ran in the dead of night, in the early hours of May 7, and a strike at that hour on a seminary, a mosque or a residential headquarters compound is a strike at the hour when such places are at their most domestic. A seminary at four in the morning is not a command post staffed by alert fighters; it is a building full of people asleep, some of them students, some of them family. New Delhi would argue that night strikes reduce risk, that darkness and surprise mean fewer people moving in the open and a cleaner hit on the structure. There is something to that. But the same darkness that produces a cleaner hit on the structure also produces a fuller building, because the people who use a residential religious compound are inside it, asleep, at exactly the hour the missiles arrived. The choice of timing was a choice that interacted with the choice of target, and the interaction pushed the civilian risk upward even as it pushed the targeting risk down.

Narowal and Sialkot, the two Punjab locations named among the struck sites but rarely discussed in detail, illustrate how thin the public record becomes once attention moves past the headline targets. Both are towns, not isolated installations, and both were named by Pakistan’s military as among the locations India hit. Beyond that, almost nothing entered the public domain: no detailed casualty figure specific to either, no independent assessment of what was struck, no reporting that either capital felt the need to contest at length. The silence around Narowal and Sialkot is itself a data point. It shows that the casualty record is not uniformly thin; it is thin in proportion to how little attention a given site drew. Bahawalpur and Muridke were argued over because they were dramatic. The quieter sites were simply absorbed into the aggregate figure, their specific dead folded into a national total that nobody could break back down.

What can be said with reasonable confidence is narrow but real. The Indian strikes did kill people who were not combatants. Masood Azhar’s own statement establishes that family members died at Bahawalpur, and family members are noncombatants whatever the building’s function. Mosques were hit, and people who were inside mosques died. The strikes were aimed at sites with genuine militant histories, which means New Delhi’s target rationale was not invented. And the strikes were aimed at sites embedded in towns, which means civilian death was a foreseeable outcome of the targeting decision, not a freak accident. Each of those statements survives scrutiny. The precise number does not.

The Shelling That Emptied the Line of Control

If the strikes inside Pakistan were the conflict’s most photographed source of civilian death, the artillery fire along the Line of Control was, by several independent accounts, the larger one. Within hours of Operation Sindoor, Pakistani artillery and mortar units began firing across the Line of Control into the Jammu region of Indian-administered Kashmir. The district of Poonch, a Hindu-majority pocket of hill country pressed up against the de facto border, took the heaviest fire. Indian officials and local journalists described it as the worst shelling the area had seen in more than fifty years, the heaviest since the war of 1971. The reconstruction of Pakistan’s retaliation and the Poonch shelling maps the impact zones in detail. This section is concerned with who was inside them.

To understand why the shelling killed so many civilians, it helps to understand what the Line of Control actually is as a place to live. It is not a clean border with a buffer zone. It is a serrated military frontier that runs through inhabited hill country, and towns like Poonch sit close enough to it that artillery on the Pakistani side can reach them without any feat of range. For decades, the people of these border districts have lived with the knowledge that a flare-up anywhere along the frontier can put their homes inside the beaten zone of guns they cannot see. There are periods of calm, sometimes long ones, governed by ceasefire understandings between the two militaries. And there are periods, like May 2025, when the understanding collapses and the guns open up, and when they do, the geography does the rest. The civilians of Poonch were not killed because they wandered into a war. They were killed because the war came to a district that had always been within reach of it, in a volume that the district had not seen in living memory.

The dead in Poonch were not abstractions, and the most widely reported case made that impossible to ignore. Among those killed were thirteen-year-old twins, a brother and a sister, Zain Ali and Urwa Fatima. Their family had moved to Poonch only three months earlier, and they had moved for the schools. A relative described the ordinary, hopeful logic of that decision: a family relocates so its children can study better, and the relocation places the children directly under artillery that nobody warned them was coming. The twins are the single image that traveled furthest from the Indian side of the ledger, and they traveled because they collapsed the distance between a strategic crisis and a household decision. A school district became a kill zone.

The shelling hit the institutions of ordinary life with grim consistency. Pakistani fire struck in and near the Christ School compound in Poonch, a Catholic school run by a Carmelite congregation, killing two students and damaging the nearby convent of the Congregation of Mother of Carmel. A gurdwara in Poonch was damaged. Homes, schools and religious buildings absorbed the fire because Poonch is built of homes, schools and religious buildings, set close to a border that the artillery could reach without aiming at anything in particular. Indian sources at one point catalogued the damage as more than thirty schools hit and hundreds of homes destroyed, alongside the Sikh temple and the people inside the blast radius. The detail that the dead in Poonch included Hindu, Muslim, Sikh and Christian victims, and that the damaged buildings included a temple, a gurdwara, a mosque and a convent, is not a rhetorical flourish. It is a description of what an area weapon does to a religiously mixed town: it does not distinguish, and Poonch’s casualty list reflected the town’s actual population rather than any single community.

Pakistani fire did not stay inside Poonch. Indian authorities reported shelling and cross-border firing across a wider arc of the frontier, naming Rajouri, Uri, Kupwara, Baramulla and other sectors. The geography of the bombardment matters because it determined the geography of the fear: it was not one town that emptied but a belt of border districts, each of them sending families toward shelters and toward the interior. Uri, a name already heavy in the India-Pakistan story from an earlier attack, saw civilians wounded. Kupwara and Baramulla, further north, registered the cross-border fire along their own stretches of the line. The shelling of May 2025 was not a single atrocity in a single place. It was a frontier-wide bombardment, and the reason Poonch dominates the account is that Poonch was hit hardest, not that Poonch was hit alone.

The death toll on the Indian side moved over the course of the four days, the way such tolls do. Early in the crisis, Indian police and medics reported seven civilians killed by shelling in Poonch. By the following day, the civilian toll in Indian-administered Kashmir had climbed past fifteen, then toward eighteen. India’s consolidated official figure settled at roughly twenty-one civilians killed alongside eight military and paramilitary personnel, with the great majority of the civilian dead concentrated in Poonch district. Those numbers, like Pakistan’s, were government numbers, but they had a quality the cross-border strike figures lacked: they were drawn from Indian territory, from Indian hospitals, recorded by Indian officials with physical access to the bodies and the families. That does not make them neutral. It does make them, on balance, more checkable than a figure announced from across a closed border.

The pattern of how the toll rose is itself worth noticing, because it is the pattern an honest reader should expect from any fast crisis. The first figure, seven, was not wrong; it was early. It reflected what one set of hospitals and one set of officials could confirm in the first hours. As the fighting continued and as access to the more remote affected villages improved, the figure climbed, not because the early count was a lie but because the early count was a snapshot of an incomplete picture. The same dynamic almost certainly operated on the Pakistani side, where the military’s figure rose from thirty-one to forty. The difference is that the Indian rise could be traced through accessible reporting, hospital by hospital and district by district, while the Pakistani rise was simply announced. A toll that rises is not, by itself, evidence of manipulation. But a toll that rises without any accessible trail behind the rise is a toll that cannot be checked, and that is the position the Pakistani figure occupied and the Indian figure, to a meaningful degree, did not.

Two deaths outside the Poonch concentration are worth naming because they show how far the civilian exposure spread. In Rajouri, the neighbouring district, a senior civil administrator was killed by Pakistani shelling, a district officer struck while doing the work of running a district under fire. And in Firozpur district, in Indian Punjab well away from Kashmir, a civilian was killed by a Pakistani loitering munition, the only Indian civilian death recorded outside Jammu and Kashmir. A loitering munition is a drone that waits, then dives. Its reaching a Punjab village is a measure of how the conflict’s geography refused to stay along the Line of Control. The drone phase carried the danger inland on both sides at once.

The framing dispute over the shelling matters as much as the count. New Delhi described the Poonch bombardment as the deliberate targeting of civilian towns, an atrocity aimed at populations. Islamabad described the same fire as counter-battery work, return fire against Indian artillery positioned close to those towns, with the civilian deaths a consequence of where India had placed its guns. There is a real argument buried inside the Pakistani framing. Militaries on both sides of the Line of Control have, for decades, kept positions near villages, and fire directed at those positions will inevitably catch the villages. There is also a real argument buried inside the Indian framing. Artillery and mortars are area weapons. Firing them in volume into a district full of homes is not precision counter-battery work, whatever the stated aim, and the side that opens that fire owns its spread. The disagreement is not resolvable from the casualty figures alone. It is resolvable, if at all, only by examining exactly what was being aimed at and how close it sat to what was hit, and that examination has never been done by anyone both governments would accept.

What survives scrutiny here is, again, narrow and real. Pakistani fire killed Indian civilians, including children, and the killing was concentrated in Poonch. The fire struck schools, homes and places of worship in volume. The Indian toll is more checkable than the Pakistani one, because it was recorded on accessible territory, but it is still a government toll and independent investigators later suggested the real number along the Line of Control was higher than the headline figure. And the question of whether the shelling was aimed at towns or at guns near towns is exactly the kind of question that decides whether a death is a war crime or a tragedy, and exactly the kind of question that no neutral body was positioned to answer.

The Drone Nights and the Death Far From Kashmir

Between the opening strikes and the ceasefire, the conflict passed through a phase that did not look like any previous India-Pakistan crisis. For three nights, the two countries fought with drones, and the drone fighting carried the civilian risk into places the artillery could never have reached. The account of the weapons systems deployed during Operation Sindoor catalogues the hardware. The civilian consequence of that hardware is the subject here.

India described waves of Pakistani drones, by its count several hundred small Turkish-made loitering systems, sent against dozens of sites across its western states, a mix of military and civilian locations. Pakistan described Indian drones, including Israeli-made loitering munitions, intercepted over its major cities, with figures it put in the dozens. As with everything else in this conflict, the numbers were claimed rather than confirmed, and each side’s interception count was a number the other side could not check. But the drones did two things that the missiles and the artillery had not. They blacked out cities, and they killed inland.

The blackouts were a civil-defence response, and they swept across both countries. Indian cities ran blackout drills and air-raid warnings, sirens sounding in places that had not heard them in living memory. Pakistani cities did the same, with schoolchildren in Karachi photographed sheltering under their desks during mock drills. Those drills are not casualties, but they are part of the human cost, because they are the moment when tens of millions of people who were nowhere near the Line of Control nonetheless lived inside the war for a few nights. The conflict reached them as darkness, as sirens, as a child told to get under a desk.

The inland deaths were the sharper consequence. The Firozpur loitering-munition death, already noted, happened because a drone could travel where a shell could not. On the Pakistani side, Indian drones were reported to have fallen on the outskirts of Karachi and in villages in Sindh near the border, hundreds of kilometres from the contested frontier. A drone war does not have a front line in the way an artillery duel does. It has a sky, and the sky covers everyone underneath it. The civilian story of the drone nights is partly a story of specific deaths far from Kashmir and partly a story of fear distributed across two entire countries.

This civil-defence response deserves its own description, because it was the moment the conflict became a lived experience for populations the weapons never reached. Across northern and western India, authorities ordered blackouts and ran air-raid drills in cities that had not practised them in decades. Sirens that most residents had never heard sounded over major towns. Lights went out on schedule, by official instruction, as a precaution against drones and aircraft using illuminated cities to navigate. On the Pakistani side, the picture mirrored the Indian one: mock drills in the major cities, schoolchildren in Karachi photographed crouching under their desks with their hands over their heads, the rehearsal of catastrophe performed by children too young to remember any previous crisis. These drills did not kill anyone. They belong in a casualty assessment anyway, because they are the broadest layer of the conflict’s reach. The fighting along the Line of Control involved a few border districts. The blackouts and the sirens involved the better part of two countries, and for the nights they lasted, tens of millions of people went to sleep, or failed to, inside the expectation of attack.

The drone nights were also the phase in which the information war and the civilian story became hardest to separate. Each government announced interception figures the other could not check. Each described the other’s drones as aimed at civilian areas and its own as aimed at military ones. Wreckage that came down in a field or on a city’s edge became, instantly, evidence in a propaganda contest rather than a fact to be jointly examined. For the civilians underneath, the effect was a kind of doubled fear: fear of the drones themselves, and fear amplified by a media environment in which both states had every incentive to make the threat sound larger when describing the enemy’s drones and the defence sound total when describing their own. The civilian who sheltered through those nights was sheltering inside both a real danger and a managed narrative, and the two were, by design, difficult to tell apart.

The drone phase also widened the verification problem rather than narrowing it. Artillery leaves craters and fragments that can, in principle, be traced. A missile strike leaves a damaged building that satellites can photograph. A drone shot down over a city, or one that crashed in a field, produces wreckage whose origin and intent are far harder to establish, and both governments had every incentive to describe that wreckage in whatever way served the larger narrative. By the time the ceasefire took hold, the civilian toll had three sources, the cross-border strikes, the Line of Control shelling and the drone nights, and the drone nights were the source about which the least could be said with confidence.

Key Figures

Most accounts of the 2025 conflict’s civilian story run through governments and militaries, but it was lived by particular people, and a few of them, by accident of circumstance, became the figures through whom the larger toll was understood.

What follows is not a complete list, because a complete list does not exist and could not, given the verification conditions described later in this account. It is a small set of people and institutions who, for one reason or another, became visible, and whose visibility makes the abstraction of a casualty count briefly concrete. Some of them are victims. One of them is the opposite of a victim. Several of them are not individuals at all but the institutional bodies that produced the numbers everyone else then argued over. They are grouped together here because each, in a different way, is a point at which the civilian story of the conflict can actually be seen rather than merely totalled, and an account that stayed at the level of the totals would have described the conflict without ever once describing a person inside it.

The Twins of Poonch

Zain Ali and Urwa Fatima were thirteen years old, a brother and sister, and they died in the Pakistani shelling of Poonch. Their family had moved to the town only three months before the conflict, and a cousin explained that the move had been made for the schools, for the children’s education. They became the most widely reported civilian deaths on the Indian side because their story carried a particular weight: a family had made an ordinary, careful decision about their children’s future, and that decision placed the children under artillery. The twins were not a statistic the Indian government deployed; they were a fact that reporters found, and the fact did more to communicate the human cost of the shelling than any official figure. They are named here because an account of civilian casualties that lists only numbers has already failed at the thing it is supposed to do.

Masood Azhar, Counting His Own Dead

Masood Azhar, the founder of Jaish-e-Mohammed, is not a civilian and not a victim in any sense this article uses the word. He appears among the key figures because of what he said. After the strike on Bahawalpur, Azhar stated publicly that ten members of his family and four of his aides had been killed. That statement is one of the strangest pieces of evidence the conflict produced, because it simultaneously confirms the Indian targeting rationale and the Pakistani civilian claim. It confirms that Bahawalpur was a genuine militant headquarters, because the head of the organization treated the strike as a strike on his own base. And it confirms that noncombatants died there, because the relatives of a militant leader are still noncombatants. Azhar counted his dead for his own purposes. The count nonetheless became part of the public record of the strikes, and it cannot be dismissed from either column it touches.

The District Officer of Rajouri

A senior civil administrator in Rajouri district was killed by Pakistani shelling during the conflict. He was not a soldier. He was the kind of official who keeps a district functioning, and he was killed doing that work while the district was under fire. His death is named here because it marks a category that casualty figures usually erase: the civilian who dies not because he is hiding from the war but because he is trying to manage it, the administrator, the police officer, the medic, the teacher who stays. The conflict’s civilian toll included people whose job was to hold the civilian world together while it was being shelled. That category matters for the count as well as for the record, because deaths of this kind tend to be classified inconsistently. A district officer killed by enemy fire can be folded into an official total, or treated as a service death, or simply noted in a local report that never reaches the national figure, and which of those happens depends on administrative accident rather than on any principle. The officer of Rajouri stands in for an unknown number of others whose work placed them in the open during the shelling and whose deaths sit in the uncertain space between the civilian column and the uniformed one.

The Counters Themselves

The last key figures are institutional rather than individual: the spokesmen and media wings who produced the numbers. On the Pakistani side, the military’s media wing issued the casualty figures, first thirty-one, later forty civilians and eleven service members. On the Indian side, state police, hospital officials and government spokesmen produced the toll that settled near twenty-one civilians. These institutions are key figures in the civilian story because, in the absence of any neutral counter, they were the only counters. Every number in this article passed through one of them. Naming them as figures rather than treating their numbers as facts is the single most important analytical move an honest account can make, because it keeps the reader aware that the casualty record of this conflict is not a measurement. It is a set of competing testimonies.

The Household at Bahawalpur

The people who died inside the seminary compound at Bahawalpur are harder to name than the twins of Poonch, because the Pakistani state did not release a granular list and the one public accounting came from Masood Azhar, who had his own reasons for the count he gave. But the doctor’s report of thirteen dead, including a child, and Azhar’s statement that ten of his family members were among the dead, together establish a household. A child died there. Relatives of a militant leader died there, and relatives of a militant leader are not themselves combatants. Whatever else the Bahawalpur compound was, it contained, on the night of May 7, a family, and the family is named here as a figure in the conflict’s civilian story precisely because the structure around them, a seminary tied to a banned organization, has been used by both governments to argue the family out of the civilian column. They belong in it. A doctrine that cannot say so plainly has lost the thread of what a civilian is.

The Journalists Who Stood at the Rubble

A last set of individual figures worth naming are the reporters, local and international, who did the work that the verification system did not. With no neutral investigative body present, the closest thing to independent documentation of the civilian toll came from journalists who reached the affected places and recorded what they found: the reporters in Poonch who found the family of the twins and asked why they had moved to the town, the correspondents who later traced the firing along the Line of Control and concluded the toll was higher than the headline figures, the photographers whose images of the Christ School compound and the struck mosques entered the record. They were not neutral counters in the formal sense, and several of them worked for outlets with national affiliations. But they had physical access, they named specifics, and they were willing to publish findings that complicated their own governments’ tidy numbers. In a conflict where the official casualty record was a set of competing testimonies, the journalists were the figures who came closest to producing evidence, and a casualty assessment that does not acknowledge them has misunderstood where its own more reliable facts came from.

Consequences and Impact

The dead are the sharpest measure of the civilian cost, but they are not the whole of it. The 2025 conflict displaced, frightened and damaged far more people than it killed, and the consequences outlived the four days of fighting.

The largest immediate consequence was displacement. As the shelling intensified along the Line of Control, Indian authorities evacuated tens of thousands of civilians from border villages, and thousands of people spent consecutive nights in shelters and bunkers. Along the Pakistani side of the frontier, families fled the border districts in the same way. The account of the Kashmir evacuation and civil-defence response sits alongside this one in the series, but the displacement belongs in any honest casualty assessment, because a family that survives the shelling by abandoning its home and its standing crop has still suffered a real and lasting harm. Displacement is not death, but it is not nothing, and the conflict produced it on a scale measured in tens of thousands.

Damage to civilian infrastructure was the second consequence, and it was concentrated where the fire was heaviest. In Poonch and the surrounding districts, schools, homes and religious buildings were destroyed or damaged in numbers that Indian sources counted in the dozens for schools and the hundreds for homes. The Christ School compound, the gurdwara, the convent, the ordinary houses of the border villages: each was both a structure and a piece of a community’s continuity, and the rebuilding would take far longer than the ceasefire took to arrive. On the Pakistani side, the struck compounds at Muridke and Bahawalpur contained schools and clinics whose civilian functions were damaged regardless of the functions tied to militant groups that shared the address. Infrastructure damage is the part of the human cost that persists in concrete after the fighting stops, and the part that is still visible, in a half-rebuilt school or a patched roof, long after the casualty figures have settled into the archive.

The third consequence reached deep into ordinary civilian life through disruptions that had nothing to do with the front and everything to do with the war’s gravitational pull on a region’s normal functioning. India closed a large number of airports across its northern and western states, cancelling hundreds of flights and stranding travellers who had no connection to the crisis beyond their geography. Pakistan closed its airspace, which rerouted international air traffic far beyond South Asia. Both countries suspended their flagship cricket competitions, the kind of suspension that sounds trivial against a death toll but that registers, for hundreds of millions of people, as the moment the war touched something they followed every day. Financial markets in the region wobbled on rumour and fear. None of this is a casualty. All of it is part of the impact, because the impact of a four-day war between two large states is not confined to the people in its blast radius; it is the sum of every normal thing that stopped, every plan that was abandoned, every routine that the crisis interrupted across a region of more than a billion and a half people.

The fourth consequence was psychological and national in scale. The civil-defence drills, the blackouts, the sirens and the shelter nights reached populations that the weapons themselves never touched. For a few nights, two countries of well over a billion and a quarter billion people respectively lived inside the expectation of attack. Children in Karachi sheltered under desks; cities across northern and western India went dark on schedule. That experience does not appear in any casualty figure, and it should, because it is among the broadest layers of the conflict’s civilian impact: not the people who were hit, but the far larger number who spent the crisis braced to be.

A fifth consequence was the grief that the ceasefire did not address. When the fighting stopped, both governments moved quickly to narrate the conflict as a success, each claiming military achievement, each claiming the other had been deterred. The analysis of the ceasefire’s fragile aftermath traces how fast the triumphal framing set in. What the framing left out, on both sides, were the families. The relatives of the twins in Poonch, the household at Bahawalpur, the farmer’s family in Firozpur: none of them were part of either victory story, because grief does not fit inside a victory story. The consequence worth naming is that the conflict’s civilian dead were, almost immediately, narrated past. The casualty figures became debating points in an information war within days of the bodies being counted.

Taken together, these consequences describe a civilian cost that the casualty figure, whatever its true value, badly understates. The dead are the part of the harm that can be written as a number. The displaced families, the broken schools, the cancelled flights, the darkened cities, the sheltering children and the grief that no ceasefire reached are the part that cannot, and they are larger than the death toll by orders of magnitude. A border village whose population fled, whose standing crop was lost and whose school was shelled has suffered something real and lasting even if it recorded no deaths at all. The human cost of the 2025 conflict is best understood as a set of widening rings. At the centre are the people who were killed. Around them are the people who were wounded. Around them are the people who were displaced. And around all of it are the tens of millions who simply lived for a week inside the expectation that the next ring might reach them. An honest casualty assessment names the centre as precisely as the evidence allows, because the centre is where the irreversible harm sits, but it refuses to pretend the centre is the whole of the figure, because the conflict’s reach was always wider than its body count.

The Verification Problem

Every number in this article is a claim, and the central analytical fact of the 2025 conflict’s civilian casualties is that almost none of the claims were independently verified. This is not a minor caveat to be cleared in a sentence. It is the structural condition of the entire casualty record, and understanding why verification failed is more useful than pretending a confident total exists.

Start with what verification of a wartime civilian death actually requires. It requires physical access to the site, to confirm what was hit. It requires access to the bodies or to credible medical records, to confirm who died and how many. It requires the ability to distinguish combatants from noncombatants, which in a conflict involving militant infrastructure is genuinely difficult. And it requires a counter with no stake in the result, an investigator whose figure does not serve a government. The 2025 conflict failed all four conditions at once.

Physical access failed because the border was closed and the fighting was fast. The cross-border strike sites inside Pakistan were never examined by anyone India would accept, and the Indian shelling sites were never examined by anyone Pakistan would accept. International news organizations said plainly that they could not independently confirm the casualty figures from either side. A major international monitoring group that tracks the harm done by explosive weapons issued a statement during the crisis calling for exactly what did not exist: independent verification of the sites struck, the weapons used and the scale of the harm, and transparent reporting of civilian casualties by both governments. That call is itself the cleanest evidence of the gap. Monitors do not demand verification of conflicts that have been verified.

The combatant-noncombatant problem failed verification in a particular way. When India strikes a seminary that is genuinely tied to a militant group, and bodies are recovered, the bodies are claimed by India as militants and by Pakistan as civilians, and no neutral process exists to sort them. The Bahawalpur case shows the problem in miniature: the same dead can be a militant leader’s family, which is to say noncombatants, and the population of a functioning militant headquarters, which is to say a legitimate target’s occupants, and the category they fall into depends on who is describing them. A casualty count cannot be more accurate than the combatant-status determinations underneath it, and those determinations were never made by anyone outside the two states.

The independent-counter problem is the deepest. Every figure in the public record of this conflict came from a government or a military media wing. Pakistan’s casualty numbers came from its military’s media arm. India’s came from its state police, its hospitals and its officials. Neither set of counters was neutral, because both were also combatants in the information war that ran alongside the shooting war. This does not mean the numbers were invented. It means the numbers were produced by institutions with an interest in the result, and no institution without such an interest was present to check them.

There is a further asymmetry inside the verification failure that is worth stating precisely, because it is the one place where the two casualty records are not equally opaque. The Indian civilian toll was generated on territory that, while heavily militarized, was not sealed against its own press. Reporters reached Poonch, walked the damaged streets, found the family of the twins and counted the broken schools, and Indian hospitals and district administrations produced records that could at least be cross-checked against that reporting. The Pakistani civilian toll was generated under different conditions. The deep-Punjab strike sites were treated as sensitive locations tied to militant organizations, access to them was tightly controlled, and the fullest single public accounting of the dead at Bahawalpur came from the militant leader whose base had been hit. This does not make the Indian figure true and the Pakistani figure false. It means the Indian figure is checkable in a way the Pakistani figure is not, and an honest assessment has to weight its confidence accordingly rather than treating two numbers of similar size as though they rested on similar evidence. Post-conflict analysis by independent researchers who studied the firing along the Line of Control reinforced the point from the other direction. Their work suggested that the cross-border shelling killed more people than either government’s headline number captured, and that a large share of the conflict’s total civilian deaths occurred in the border villages where the counting in real time was thinnest. The figure that was most checkable, in other words, may also have been the most undercounted, which is the kind of conclusion that only an inquiry without a stake in the outcome can reach and that neither government had any incentive to reach for on its own.

The findable pattern, the thing this article can offer that a wire report cannot, is a tiered map of confidence. At the highest tier of confidence sit a small number of statements: civilians died on both sides; children died on both sides; the killing was concentrated in Poonch on the Indian side and in the struck compounds on the Pakistani side; the Indian toll is somewhat more checkable than the Pakistani one because it was recorded on accessible territory. At the middle tier sit the rough ranges: an Indian civilian toll in the low twenties, a Pakistani civilian toll that the Pakistani military placed at forty after first saying thirty-one. At the lowest tier of confidence sits everything precise: the exact totals, the combatant breakdown of the dead inside Pakistan, the drone-phase casualties, and the true Line of Control total, which independent investigators who studied the firing afterward suggested was higher than either government’s headline. An honest reader should hold the top tier firmly, the middle tier loosely, and the bottom tier as openly contested. That tiered structure is not a failure to reach a conclusion. It is the conclusion. The verification problem is not noise around the casualty story. It is the casualty story.

Analytical Debate

The civilian casualties of the 2025 conflict generated a debate that ran along several fault lines at once, and the debate is worth laying out directly, because the way an observer answers it determines what they think the conflict was.

Start with the simplest fault line: how many civilians died, and on which side more. India’s position is that its precision weapons minimized civilian harm in Pakistan and that the meaningful civilian toll was the one Pakistan inflicted on Indian towns through indiscriminate shelling. Pakistan’s position is the mirror image: that India’s strikes deliberately hit mosques and homes and produced the larger civilian toll, while its own fire was militarily targeted. The available evidence does not cleanly vindicate either. The Pakistani military’s own revised figure of forty civilians exceeds India’s figure of roughly twenty-one, which on the face of it supports the Pakistani claim of a larger toll. But the Indian figure is the more checkable of the two, and independent investigators suggested the Line of Control toll was undercounted, which would narrow or even close the gap. The most defensible position is that the two civilian tolls were closer to each other than either capital wanted to admit, and that both were probably higher than the early headline numbers.

The second fault line is about intent, and it is the one that decides moral and legal weight. Did either side deliberately target civilians? India accused Pakistan of exactly that in Poonch. Pakistan accused India of exactly that at Bahawalpur and Muridke. The honest analytical answer is that intent is the hardest thing to establish from outside and the easiest thing to assert from inside. India’s strikes were aimed at sites with real militant histories, which argues against an intent to kill civilians as the objective, even as it concedes that civilian death was a foreseeable result of striking embedded targets. Pakistan’s shelling was area fire into a populated district, which makes the claim of careful counter-battery targeting hard to sustain at the volume reported, even if specific Indian gun positions near towns were the nominal aim. Neither side’s conduct is cleanly exonerated and neither is cleanly convicted, and anyone offering a confident verdict on intent is supplying certainty the evidence does not contain.

The third fault line is about precision itself, and it is the one most relevant to the larger argument this series makes. India presented Operation Sindoor as a precision operation, and in a narrow technical sense the claim has support: guided weapons, named coordinates, deep targets struck without a broader bombing campaign. The comparison with the 2019 Balakot airstrike shows how far Indian strike capability had developed. But precision is a property of the weapon, not of the outcome. A precise weapon aimed at a target embedded in a town produces civilian casualties with precision. The debate, then, is whether precision is a meaningful moral category when the target selection guarantees civilian exposure regardless of how accurate the strike is. Analysts of conflict between nuclear-armed states, including Caitlin Talmadge of Georgetown, whose work examines how limited wars between such states actually unfold, have pressed the point that limited and precise are not the same as harmless, and that the civilian cost of a restrained operation is still a cost the operation owns. The precision framing is not false. It is incomplete, and the incompleteness is exactly the size of the civilian toll.

Verification forms the fourth fault line, and the dispute is whether the gap should be read as innocent or culpable. One view holds that the absence of independent counts is simply what happens in a fast crisis between two closed adversaries, an unfortunate condition rather than a choice. The opposing view holds that both governments actively benefited from the fog, because an unverifiable toll can be shaped, and that the failure to allow independent investigation was therefore at least partly a decision. Observers of the conflict’s humanitarian dimension, including Michael Kugelman of the Wilson Center, who has written extensively on the human stakes of India-Pakistan crises, have noted that the information environment around the 2025 fighting was contested from the first hour and that the casualty figures were instruments in that contest as much as they were measurements of it. The debate over the verification gap is, in the end, a debate over whether the world should have known more, and whether the not-knowing served someone.

The fifth fault line is about proportionality, and it is the one that connects the casualty count to the question of whether the conflict should have happened in the form it did. One side of the debate holds the two events against each other directly. The Pahalgam attack killed twenty-six civilians, and the conflict launched in response killed, by the combined figures of both governments and the higher independent estimates of the border firing, a larger number of civilians than that. On this reading, the response cost more innocent life than the atrocity that triggered it, and a response that kills more noncombatants than the attack it answers has failed a basic moral test regardless of its strategic logic. The other side of the debate rejects that arithmetic as too narrow. It argues that a response is not measured only against the single attack that prompted it but against the attacks it is meant to deter, and that the relevant comparison is therefore not twenty-six against forty but the casualty curve of a region with a credible deterrent set against the casualty curve of a region without one. On this reading, a response that is painful in the moment can still be defensible if it lowers the long-run civilian toll by making the next Pahalgam less likely. The honest difficulty is that the second argument cannot be proved inside the short window the casualty figures cover. Deterrence is a claim about attacks that did not happen, and attacks that did not happen leave no bodies to count. The proportionality debate therefore does not resolve into a number. It resolves into a prior choice about which civilians belong in the ledger at all: only the ones already dead, who can be counted, or also the ones a deterrent is meant to protect, who never can be.

The Moral Accounting

This series argues a single connected thesis across all of its articles: that the campaign of targeted eliminations and the conventional military operations are two arms of one doctrine, and that a state which shelters terrorism eventually finds the shelter becoming the threat. Most articles in the series engage that thesis by confirming it. This one engages it by pressing on the place where it strains, because the civilian dead of May 2025 are the part of the doctrine that the doctrine’s own logic cannot fully absorb.

The case for the thesis, applied to this conflict, is coherent and should be stated fairly. The argument runs that the militant infrastructure struck at Bahawalpur and Muridke was real, that it was embedded in towns by deliberate choice, and that the choice to headquarter banned organizations inside campuses full of students was itself a decision to use civilians as shields. On that reading, the civilian deaths at the struck compounds are a cost that belongs, at least substantially, to the side that built the shield, not only to the side that fired the missile. The argument continues that the alternative to striking embedded infrastructure is not a clean war but no response at all, and that no response is what produced Pahalgam in the first place. The doctrine’s defenders would say that the civilian toll of Operation Sindoor, however painful, is smaller than the civilian toll of the attacks the operation was answering, and that a doctrine which reduces total civilian death over time is morally defensible even when it produces civilian death in any given instance.

That argument is not empty. But it does not reach far enough, and the honest version of the thesis has to say where it stops.

It stops at the twins in Poonch. The doctrine’s logic can account for the dead at a militant headquarters by pointing to the shield. It cannot account for thirteen-year-old children killed by artillery in a town their family moved to for the schools, because Poonch was not a militant headquarters and the twins were not anyone’s shield. They were killed by Pakistani fire, which means they are not even a cost of India’s doctrine in a direct sense; they are a cost of the escalation the doctrine set in motion. And that is precisely the point at which the thesis has to be honest about its own reach. A doctrine that says shelter becomes the threat is an argument about Pakistan’s choices and India’s response. It is not an argument that can absorb the death of children in an Indian border town, because those children were not killed by the precision the doctrine celebrates. They were killed by the area fire that the doctrine’s escalation predictably summoned. The doctrine can claim that Pakistan chose to shell Poonch. It cannot claim that the shelling of Poonch was no part of the doctrine’s consequences, because the operation that the doctrine produced is what the shelling was answering.

It stops, too, at the household at Bahawalpur. Even granting every part of the embedded-infrastructure argument, the family members who died there were people, and the precision that killed them was precision aimed at a town. A doctrine that wants to be taken seriously as a moral framework, rather than only as a strategic one, has to be able to say the sentence the victory narratives would not say: that the operation killed noncombatants, that some of them were children, and that this is a debit, not an asterisk. The thesis survives as a strategic description. It does not survive as a moral one unless it carries that debit openly.

There is a distinction worth drawing here that the victory narratives on both sides deliberately collapsed, because collapsing it was useful to them. Moral ownership of a civilian death is not the same as legal fault for it, and neither is the same as strategic responsibility for the chain of events that produced it. A doctrine can be strategically sound, legally defensible in the narrow sense that it struck objects with a genuine military character, and still carry moral ownership of the noncombatants its strikes killed. Those three things can all be true at once, and an honest account refuses to let any one of them cancel the others out. The Indian government’s case that Bahawalpur and Muridke were legitimate targets does not erase its moral ownership of the family members who died inside them. The Pakistani government’s case that its fire along the Line of Control was counter-battery work does not erase its moral ownership of the twins in Poonch. Each capital spent the days after the ceasefire trying to use the legal column and the strategic column to zero out the moral one, and that attempt is precisely the move this section is built to resist. The moral column does not zero out against the others. It is added separately, and it stays on the page.

So the inverse engagement resolves like this. The civilian casualties of 2025 do not refute the thesis that the shadow war and the open war are one doctrine. The targeting was real, the embedded infrastructure was genuine, the escalation logic was real, and the campaign continues to look, strategically, like a single connected body of action. But the civilian dead mark the doctrine’s outer limit as a moral account of itself. They are the cost the doctrine must carry rather than the cost the doctrine can explain. Every civilian death above zero, on both sides, is a number the doctrine has to hold, not narrate past. The thesis is strong enough to describe the strategy. It is not strong enough, and should not be asked, to make the children of Poonch or the household at Bahawalpur into anything other than what they were, which is people who were alive in April and were not alive in May, and whose deaths no framing improves.

Why It Still Matters

The civilian casualties of the 2025 conflict matter beyond the four days that produced them, and they matter in ways that the ceasefire did not settle.

They matter, first, because the verification gap was never closed and is therefore a template. If the next crisis between India and Pakistan unfolds the way this one did, the casualty record of that crisis will be produced the same way: by two governments who are also combatants in the information war, with no neutral counter present, on a closed border, in a fog that benefits whoever wants to shape a number. The 2025 conflict did not just produce contested casualty figures. It demonstrated that contested casualty figures are now the expected output of an India-Pakistan war, and that demonstration will shape how the next one is reported, argued and remembered. The analysis of India’s post-Pahalgam defence doctrine describes a posture of standing readiness rather than episodic response, and a posture of standing readiness implies more crises, which implies more casualty records produced under the same broken conditions.

They matter, second, because the embedded-infrastructure problem is not going away. As long as militant organizations headquarter themselves inside towns, inside campuses, inside compounds that contain schools and clinics and mosques, any strike on that infrastructure will kill civilians, and the moral argument over who owns those deaths will recur identically every time. The 2025 conflict did not resolve that argument. It only ran it once, loudly, and the next iteration will start from the same place, with the same two framings, unless the underlying condition changes.

A third reason they matter is that the displacement and the fear reached populations far larger than the dead. Tens of thousands of people were evacuated from border villages, and two entire countries spent nights braced for attack. That broader civilian impact is the part that no casualty figure captures and the part that recurs most reliably, because every escalation, even one that kills relatively few, distributes fear across the whole region. The human cost of the 2025 conflict is not only the people in the ground. It is the much larger number who learned, again, that they live inside the range of the next crisis.

They matter, finally, because of what the victory narratives left out. Both governments moved within days to describe the conflict as a success, and neither success story had room for the families. An account of the 2025 conflict that names the twins of Poonch, the household at Bahawalpur, the officer in Rajouri and the farmer in Firozpur is not taking a side between New Delhi and Islamabad. It is taking a side for the proposition that the people who were killed are part of the record, that they belong in the history of the conflict and not only in the grief of their families, and that a war remembered only through its strategic outcomes has been remembered dishonestly. The civilian casualties matter because they are the part of the 2025 conflict that the conflict’s own narrators were quickest to forget, and the cost of forgetting them is paid the next time the same fog is allowed to settle over the same kind of war.

None of this is an argument that the conflict should be remembered only as a tragedy, or that its strategic questions do not deserve serious attention. They do, and the rest of this series gives them that attention. It is an argument that the strategic record and the human record are not the same record, and that the second one is the one far more likely to be lost. Soldiers killed in uniform are counted by the institutions that sent them into the field. Fighters killed in strikes are counted, and argued over at length, by the states that targeted them. The civilians have no such institution behind them. They were counted, when they were counted at all, by governments for whom the number was an instrument, and by reporters working against a window that was already closing. That is the reason an account like this one needs to exist. Its purpose is not to settle the toll, which cannot be settled honestly from outside, but to refuse the alternative, which is to let the people who were alive in April and were not alive in May slide out of the history altogether while the conflict is remembered through its maps, its weapons and its competing claims of success. The civilian casualties still matter because they are the part of the record with no natural keeper, and the part of any record that has no keeper is the part that quietly disappears.

Frequently Asked Questions

Q: How many civilians were killed in the 2025 India-Pakistan conflict?

There is no single confirmed figure, and that is the most important fact about the count. Pakistan’s military first said thirty-one civilians were killed by Indian strikes, then revised the figure to forty civilians, including seven women and fifteen children, alongside eleven service members. India’s official toll settled at roughly twenty-one civilians killed by Pakistani shelling, drones and a loitering munition, alongside eight military and paramilitary personnel. Independent investigators who later studied the firing along the Line of Control suggested the true civilian total was higher than either government’s headline number, because much of the killing happened in border villages where no neutral party was counting in real time. Both figures came from governments that were also fighting an information war, and no independent body verified either.

Q: Were there civilian casualties on both sides?

Yes, and this is one of the few things that can be stated with high confidence. Indian strikes on the night of May 6 into May 7 killed civilians in Pakistan, in Bahawalpur, Muridke, Muzaffarabad and Kotli, a fact confirmed in part by Masood Azhar’s own statement that members of his family died at the Bahawalpur site. Pakistani artillery and mortar fire across the Line of Control killed civilians in India, concentrated in Poonch district, with further deaths in Rajouri and a loitering-munition death in Firozpur in Punjab. Children died on both sides. The conflict’s civilian cost was genuinely bilateral, and no honest account can present it as one-sided.

Q: Did India’s precision strikes cause civilian deaths?

Yes. India described Operation Sindoor as a focused, measured operation using precision-guided weapons aimed only at militant infrastructure, and in a narrow technical sense the strikes were precise. But precision is a property of the weapon, not of the outcome. The targets at Bahawalpur and Muridke were embedded inside towns, inside compounds that contained schools, clinics and mosques, and a precise weapon aimed at a target inside a town produces civilian casualties precisely. Masood Azhar’s confirmation that family members died at Bahawalpur, and Pakistani reports of civilians killed at the other sites, establish that the strikes killed noncombatants regardless of how accurate the weapons were.

Q: How many Indian civilians were killed by Pakistani shelling?

India’s consolidated figure was roughly twenty-one civilians killed, with the great majority in Poonch district, where the bombardment was described as the worst in more than fifty years. The dead included thirteen-year-old twins, Zain Ali and Urwa Fatima, whose family had moved to Poonch only three months earlier for the schools, and two students killed in or near the Christ School compound. A senior civil administrator was killed by shelling in Rajouri, and a civilian in Firozpur in Indian Punjab was killed by a Pakistani loitering munition, the only Indian civilian death recorded outside Jammu and Kashmir. The reconstruction of the Poonch shelling maps the impact in detail.

Q: How many Pakistani civilians were killed by Indian strikes?

Pakistan’s military initially reported thirty-one civilians killed in the May 7 strikes, then raised the figure several days later to forty civilians, among them seven women and fifteen children, with scores wounded. The deaths were reported across Bahawalpur, where a doctor said thirteen people including a child were killed near a struck mosque, Muridke, where a Pakistani newspaper reported at least three civilian deaths, and Muzaffarabad and Kotli, where Pakistan said five civilians died. The figure is complicated by the separate dispute over how many of the dead were militants rather than civilians, a question examined elsewhere in this series, and by the fact that the count came entirely from Pakistani state sources with no independent verification.

Q: Is there an independent civilian death toll for the conflict?

No, and the absence is structural rather than accidental. International news organizations stated plainly that they could not independently confirm the casualty figures from either side. The border was closed, the fighting was fast, and no neutral investigator had physical access to the strike sites inside Pakistan or the shelling sites inside India. A major international group that monitors the harm done by explosive weapons issued a statement during the crisis calling for exactly what did not exist: independent verification of the sites struck, the weapons used and the scale of the harm. Every number in the public record passed through a government or a military media wing, and no body without a stake in the result produced a count.

Q: Why are the casualty numbers disputed?

They are disputed because four conditions that verification requires all failed at once. There was no physical access to the sites for neutral investigators, because the border was closed. There was no agreed way to distinguish combatants from noncombatants among the dead inside Pakistan, because the strikes hit genuine militant infrastructure embedded in towns. There was no independent counter, because every figure came from a government that was also fighting an information war. And the numbers themselves evolved during the crisis, with Pakistan’s figure rising from thirty-one to forty, which can be read as a fuller picture emerging or as a figure being shaped. The dispute is not noise around the casualty story. It is the casualty story.

Q: Did either side deliberately target civilians?

Each side accused the other of exactly that, and intent is the hardest thing to establish from outside. India accused Pakistan of deliberately shelling civilian towns in Poonch; Pakistan called the same fire counter-battery targeting of Indian guns positioned near towns. Pakistan accused India of deliberately striking mosques and homes; India said it struck only militant infrastructure and that civilian death was an unintended result of targets embedded in populated areas. The defensible analytical position is that India’s targets had real militant histories, which argues against civilian death as the objective, while Pakistan’s area fire into a populated district is hard to reconcile with careful targeting at the volume reported. Neither side is cleanly exonerated and neither is cleanly convicted.

Q: What happened to civilians in Bahawalpur?

Bahawalpur, a city in southern Pakistani Punjab, was struck because of its association with Jaish-e-Mohammed, whose seminary complex and headquarters mosque were located there. A doctor at a nearby hospital reported that thirteen people were killed, including a child. The case is uniquely contested because Masood Azhar, the founder of Jaish-e-Mohammed, publicly stated that ten members of his family and four of his aides had been killed in the strike. That statement confirms both that the site was a genuine militant headquarters and that noncombatants, including family members, died there. Bahawalpur is the clearest single illustration of why the casualty argument never resolves: the same strike was a hit on a terror base and a strike that killed a household.

Q: What happened to civilians in Poonch?

Poonch, a Hindu-majority district in the Jammu region pressed against the Line of Control, took the heaviest Pakistani shelling of the conflict, described by Indian officials and local journalists as the worst in over fifty years. The dead included thirteen-year-old twins whose family had relocated to the town for its schools. Pakistani fire struck in and near the Christ School compound, killing two students and damaging a Carmelite convent, and damaged a gurdwara and large numbers of homes. Indian sources catalogued more than thirty schools hit and hundreds of homes destroyed across the affected districts. The shelling also triggered the evacuation of tens of thousands of people from border villages, who spent consecutive nights in shelters.

Q: How did the drone phase affect civilians?

The three nights of drone fighting carried the civilian risk into places that artillery and missiles could not reach. India described several hundred small Pakistani loitering drones sent against dozens of sites; Pakistan described Indian drones intercepted over its major cities. The drones did two things to civilians. They prompted blackouts, air-raid sirens and civil-defence drills across both countries, reaching populations far from the Line of Control, with schoolchildren in Karachi photographed sheltering under desks. And they killed inland: a Pakistani loitering munition killed a civilian in Firozpur in Indian Punjab, and Indian drones were reported to have fallen near Karachi and in Sindh villages. The weapons systems used in the conflict are catalogued separately in this series.

Q: How many civilians were displaced by the conflict?

Tens of thousands. As the shelling intensified along the Line of Control, Indian authorities evacuated tens of thousands of civilians from border villages, and thousands of people spent consecutive nights in shelters and bunkers. Families on the Pakistani side of the frontier fled the border districts in the same way. Displacement does not appear in the casualty figures, but it is a real and lasting harm: a family that survives by abandoning its home and standing crop has still suffered a significant loss. The displacement was, by scale, one of the largest civilian consequences of the conflict, larger in raw numbers than the death toll by a wide margin.

Q: What humanitarian organizations reported on the casualties?

International monitoring and news organizations played a limited but important role, mainly by documenting that the figures could not be verified rather than by producing figures of their own. Major international news organizations stated that they could not independently confirm the casualty claims from either side. A prominent group that tracks the impact of explosive weapons in populated areas issued a briefing during the crisis urging both governments toward transparent reporting of civilian casualties and calling for independent verification of the strike sites and the harm caused. Their contribution was to mark the verification gap clearly, which is itself a finding: monitors do not call for verification of conflicts that have been verified.

Q: Was the 2025 conflict’s civilian toll larger than the Pahalgam attack that triggered it?

The Pahalgam attack on April 22 killed twenty-six civilians, almost all of them tourists. The combined civilian toll of the four-day conflict that followed, taking both governments’ figures together and the higher independent estimates of the Line of Control firing, was larger than that, though the exact total is unverified. This comparison is sometimes used to argue that the response cost more civilian lives than the attack, and sometimes used to argue that the response was still proportionate because it aimed to prevent future attacks. The full explainer on the Pahalgam attack covers the trigger event; the comparison itself is contested and depends on whether one counts only immediate deaths or also the attacks a response is meant to deter.

Q: How does this conflict’s civilian cost compare to earlier India-Pakistan crises?

The 2025 conflict killed more civilians than the 2019 Balakot crisis, which produced very few civilian deaths, and it involved the deepest strikes into Pakistani territory since the 1971 war. The Poonch shelling was described as the heaviest since 1971. What made the 2025 civilian cost distinctive was not only its scale but its breadth: missile strikes deep inside Pakistan, the heaviest Line of Control shelling in over fifty years, and a drone phase that carried casualties and fear far inland on both sides. The comparison with the 2019 Balakot airstrike and the day-by-day 2025 conflict timeline place the event in the longer sequence of India-Pakistan military crises.

Q: Did the ceasefire address the civilian casualties?

No. The ceasefire that took effect on the evening of May 10 stopped the firing, but it did not include any joint accounting of the civilian dead, any agreed verification process, or any acknowledgment of the toll by either side directed at the other’s victims. Within days, both governments were narrating the conflict as a military success, and neither victory story had room for the families of the dead. The analysis of the ceasefire’s fragile aftermath describes how quickly the triumphal framing set in. The civilian casualties were, almost immediately, narrated past, and the casualty figures became debating points in a continuing information war rather than the subject of any shared reckoning.

Q: Why does the verification gap matter for the future?

It matters because it is now a template. The 2025 conflict demonstrated that contested, unverifiable casualty figures are the expected output of an India-Pakistan war: produced by two governments who are also information-war combatants, on a closed border, with no neutral counter present. If the next crisis unfolds the same way, its casualty record will be produced under the same broken conditions, which means it will be just as contestable and just as easy to shape. A military posture of standing readiness, described in the analysis of India’s post-Pahalgam defence doctrine, implies more crises over time, and each one will generate a casualty record under the same fog unless the underlying conditions for verification change.

Q: What is the most honest thing that can be said about the conflict’s civilian toll?

That civilians died on both sides, that children died on both sides, that the killing was concentrated in Poonch on the Indian side and in the struck compounds on the Pakistani side, and that the Indian toll is somewhat more checkable than the Pakistani one because it was recorded on accessible territory. Beyond that small set of high-confidence statements sit rough ranges, an Indian toll in the low twenties and a Pakistani toll the Pakistani military placed at forty, and beyond the ranges sits everything precise, which remains genuinely contested. The most honest position holds the high-confidence statements firmly, the ranges loosely, and the precise totals as open questions. The verification gap is not a caveat to the casualty story. It is the casualty story.