Counting the Dead
Published 4 October 2026
A war leaves two ledgers. One is kept by the people who fight it, in categories chosen for administrative reasons. The other is kept by people trying to count the cost in lives, who are working in the dark, usually from outside. They rarely agree, and when a headline sums them it is almost always wrong. This chapter lays out each count, with its source, its date and its scope, and declines to add them.
The United States: nineteen is not nineteen killed in action
The Pentagon publishes casualty tables by operation. The rendered tables, read on 29 September, show the following. Operation Epic Fury has 7 hostile and 7 non-hostile deaths, 14 in all, and 418 wounded in action. The Overseas Operations category, which the table dates from 7 July, has 4 hostile and 1 non-hostile death, 5 in all, and 443 wounded. Together that is 19 deaths, 11 hostile and 8 non-hostile, and 861 wounded.
Two cautions. The 14 Epic Fury deaths include six killed in action and one who died of wounds on the hostile side, and on the non-hostile side six Air Force accident deaths and one Navy death of undetermined cause. The 861 figure is cumulative wounded, not the number now in hospital. A separate Washington Post count of 22 to 23, from anonymous officials, is disputed by the Pentagon and is not used here.
Iran: three counts that must never be added
Iran's forensic chief, via state media, said more than 3,000 had died. The human rights group HRANA, working from event-by-event documentation, published a 7 April snapshot of 1,701 civilians (at least 254 of them children), 1,221 military and 714 unclassified deaths. Those sum to 3,636, and that sum is arithmetic over one snapshot, not a current figure. HRANA noted that its method excludes official aggregates it cannot match to events, and it paused daily updates on 8 April.
HRANA's July report covers 8 to 17 July only: a minimum of 48 dead and 68 injured, 20 of the dead civilians. Its 376 documented incidents are not 376 strikes, since several incidents can describe one attack. It is a ten-day window and does not add to April.
The newest official-adjacent figure comes from the UN. In an interview published on 29 September, the UN's resident coordinator in Iran, Christine Weigand, said Iran's national authorities put deaths above 3,400, with more than 33,000 injured and 30 of 31 provinces affected. The attribution is the point. This is a UN official relaying a government figure, not a UN verification, and it is not stacked on HRANA's 3,636.
Israel
For the early weeks the ambulance service recorded 23 killed by missiles from Iran and Lebanon, and the military reported 12 soldiers killed in southern Lebanon. A farmer killed by Israeli fire on 22 March belongs in neither column. These come from different theatres and mechanisms, and "killed by Iran" is not a safe summary. A later composite from a Washington think tank, which puts Israel at 67 or more, mixes scopes and dates and is not used here.
Lebanon
Lebanon's health ministry reported 4,386 dead and 12,407 injured from 2 March to 17 September. Its separate total since 8 October 2023, 8,953 dead, is a different war's worth of time and is not this war's toll. The ministry's tally does not distinguish fighters from civilians. The Lebanese front has its own chapter.
At sea
The International Maritime Organization's table, as of 24 September, records 85 confirmed incidents and 24 seafarer deaths, and it is the best audit trail in the file because each row names a vessel, a date, a place and a casualty count. Its 16 September statement had 80 attacks and at least 22 deaths, so the later snapshot replaces the earlier one. The 24 are international seafarers. They are not 24 Indians, and they are not all attributed to Iran.
Two Indian cases are corroborated. Reuters and the BBC reported that three Indian seafarers died after the US strike of 10 June on the tanker Settebello, with 21 rescued. CENTCOM said the ship ignored warnings and carried Iranian oil in breach of the blockade. The operator denied any Iranian link, and both positions belong in the record. The other is the Cape Dao on 23 September, where The National interviewed the family of one Indian sailor killed. Neither the US nor Iran has claimed the attack, and the IMO table does not name a weapon or attacker.
Yemen and the rest
Yemen's figures sit in the Red Sea chapter, where two UN-linked tallies of 831 and 838 deaths are left unresolved on purpose. The same discipline applies to Iraq and the Gulf states, whose numbers are thinner still.
The table this chapter will not draw
A reader will want one number. There is not one. There is a Pentagon table with two categories and a cut-off date, a human rights group's April snapshot, a UN official quoting a government, a ministry's count that mixes combatants and civilians, and an IMO list that counts ships from every flag. Adding them would be a kind of fraud, even if each addend were honest. The useful thing a historian can do here is show where each number stops.
Get the next file
One email when a new file is published. No paid service needed.