Action on Armed Violence presents itself as an independent, non-partisan research organization, and the world treats it accordingly. Its Explosive Violence Monitoring Project has run for over a decade, its annual figures are cited by wire services and human-rights groups, and its data now feeds the United Nations: the Secretary-General's May 2026 report on the protection of civilians draws on AOAV's monitor, and the 2025 edition opens with a Guterres quotation calling on states to act on its findings. That standing is the reason the monitor's flaws matter. A partisan pamphlet that overstated civilian harm would persuade no one; a trusted, UN-cited, self-described neutral dataset that does the same thing launders the overstatement into the official record.
AOAV's stated methodology is structurally flawed in a way that inflates civilian counts. Worse, it suspends its own rules in the cases where applying them would have produced a figure unfavorable to the narrative the reports advance.
Here's the methodology: AOAV records casualties from explosive weapons "as reported in English-language media," logging incidents that caused at least one casualty within a 24-hour period. Two rules govern how casualties are classified. The first, stated in the methodology: "All casualties are assumed to be civilians unless otherwise stated." A casualty becomes an armed actor only when a news report explicitly identifies the dead or wounded as a soldier, militant, or armed security official. The second concerns attribution: responsibility is assigned to whoever the news report names. The project then totals these media-derived entries and publishes them under the word "civilian," flat and unqualified, across fifty-odd pages of findings, country profiles, and recommendations. The qualifier that would make the count honest — that "civilian" means "not reported as armed," not "confirmed noncombatant" — sits in a single methodology paragraph the headline reader never reaches.
One of AOAV's headline 2023 incidents shows both rules failing at once. AOAV's third-worst explosive incident of 2023 is listed as "Israeli air strike on an evacuation convoy fleeing north Gaza, 13 October," 270 civilian casualties were attributed to Israel. The footnote sources it to a Sky News article, whose headline reads: "Women and children among 70 killed in Israeli airstrike on fleeing Gaza convoy, Hamas says." The attribution to Hamas is in the title. The body went further: Sky News reported the blast was "blamed on" Israel, quoted an IDF spokesperson saying he was "not aware of any IDF strikes at this time at that location," and stated that "it was not immediately clear who the target was, or whether insurgents were among the passengers." The Associated Press wire carried the same qualification verbatim. The reporting did what responsible reporting does with an unverifiable battlefield claim — it named the source, recorded the denial, and flagged that fighters might be among the dead.
AOAV erased all three caveats. The Hamas claim became a fact of an Israeli airstrike; the denial vanished; the possibility of combatants among the passengers dissolved into a flat count of 270 civilians. And the Israeli account was not a one-line "no comment." Within two days, IDF spokesman Jonathan Conricus argued publicly that the strike "appears to have been a false flag operation carried out by Hamas," asked who would benefit from images of dead evacuees and answered "only one organization: Hamas," and said Israel "did not try to strike anybody, any civilians in that area." The surrounding facts fit that reading at least as well as the alternative: Israel had ordered the evacuation and designated the route, Hamas had told Gazans to defy the order and was reported blocking the southern roads, and Hamas held the clearer motive to manufacture atrocity footage from a convoy of fleeing civilians. There was no evidence of an actual airstrike beyond Hamas's assertion of one. The cause remains formally unresolved; reference works still list the attack type as airstrike or possible IED. AOAV recorded it as settled Israeli strike of 270 civilian casualties, third-worst in the world that year. The media's attribution — "Hamas says" — is the qualifier that made the claim publishable, and AOAV's method strips precisely that qualifier, converting "a party to the war alleges" into "AOAV records."
The civilian-by-default rule produces an even stranger result across Gaza as a whole. In the 2024 monitor, AOAV records 23,432 civilian casualties in Gaza against 612 armed actors — a civilian share of 97.5%, with combatants making up 2.5% of the total and roughly 3% of the dead. If that isn't insane enough, the military casualties it recorded were apparently all IDF soldiers! In other words, Hamas never admits any of its terrorists were killed until years later, Israel admits its soldiers are killed within hours of the incident, so AOAV thinks virtually the only militants killed in Gaza are Israeli. QEDumb.
The same selection problem corrupts AOAV's most quotable Gaza statistic. The 2024 monitor reports that Israel's recorded aerial attacks in Gaza caused, on average, "8 civilians harmed per recorded Israeli air strike, and 5 killed," and the 2023 edition builds a comparable per-airstrike figure. Read the methodology and the number dissolves. An incident enters the dataset only if "at least one casualty from an explosive weapon" was reported; a strike that hit an empty structure, or killed only fighters the media did not count, or fell where no English-language reporter was watching, never appears. The denominator is therefore not "Israeli air strikes" but "Israeli air strikes that produced reported casualties," and AOAV divides the casualties by that pre-filtered set to announce a casualty rate. The strikes that would pull the average down were excluded before the division.
It excluded a huge number of airstrikes. AOAV's 2023 dataset logs a few hundred air-launched incidents in Gaza; the Israeli military, by contrast, said it had struck over 11,000 targets in Gaza by 1 November 2023 and more than 22,000 by mid-December, with independent reporting putting the four-month total near 29,000 targets, roughly 228 a day. Whatever the precise count of distinct air operations, it dwarfs the few hundred AOAV recorded by orders of magnitude. The "civilians per air strike" figure does not describe Israeli targeting; it describes the handful of strikes that left a reportable civilian trail, which is the only kind the method can see. A monitor that counts only the strikes that killed civilians and then reports how many civilians strikes kill has not measured lethality. It has measured its own selection rule.
The absurdity reaches its peak with Lebanon, where the method collides with a fact pattern it cannot survive. AOAV's single worst explosive incidents of 2024 were the September pager operation and the next day's walkie-talkie operation. The explosives were concealed inside pagers and radios that Hezbollah itself had purchased and distributed to its members; the only people physically carrying them were Hezbollah personnel. Hezbollah
acknowledged that hundreds of its fighters carried the devices, and a Hezbollah official said the operation
took 1,500 fighters out of action through injury. Hezbollah's own leadership has long rejected the premise that there is any line between its political and military sides; as leader Naim Qassem
put it, "we don't have a military wing and a political wing... one Hezbollah." A weapon that by physical design could only injure members of the organization, in an operation the organization says wounded 1,500 of its fighters, produced in AOAV's ledger a 97% civilian casualty rate. The monitor even concedes, in its own text, that "armed actors are likely included among these casualties." Here the method does not merely risk error. Its output is contradicted by the admitted facts of the event it is describing, including the facts AOAV itself records.
If the story ended with a flawed method honestly applied, AOAV could fairly answer that open-source monitoring is imperfect and its limitations are disclosed. The defense collapses on the evidence that the method is not applied when it would point away from blaming Israel. The proof is an absence: the al-Ahli hospital explosion of 17 October 2023.
Al-Ahli was the most heavily reported explosive event of the early war. Within hours, Gaza's Ministry of Health announced 471 killed and blamed an Israeli airstrike, and the figure led news bulletins worldwide. Two findings then emerged from exactly the English-language sources AOAV scans. On the toll: the Anglican diocese that runs the hospital estimated around 200 dead, the director of al-Shifa Hospital put it near 250, and US intelligence assessed 100 to 300, likely at the low end. On the cause: US, British, Canadian, and French intelligence, along with Human Rights Watch, concluded the explosion came from a misfired Palestinian rocket that struck the courtyard rather than an Israeli strike on the building.
Now apply AOAV's own rules and watch the contradiction close around the report. AOAV publishes a table of the ten worst explosive incidents of 2023, ranked by civilian casualties; the smallest entry on it is a Pakistani suicide bombing at 193. Al-Ahli is not on the table. It is not anywhere else in the report either: across the entire document, the single most-reported explosive event of the early Gaza war goes unmentioned. There are only two ways AOAV could have reached that result, and each indicts a different rule.
If AOAV applied its standard practice — take the casualty figure reported in the first 24 hours, assume all civilian, attribute to the named party — then al-Ahli enters as 471 killed and 314 injured, attributed to Israel. That is 785 casualties, the single deadliest incident of 2023 by AOAV's stated criteria.
If instead AOAV declined the day-one figure because it was disputed and the misfire finding undercut the attribution — exercising judgment its 24-hour rule does not provide for — then it accepted a corrected toll of roughly 100 to 300 dead and several hundred injured. Even at the floor of that range, killed-plus-injured clears the 193 threshold several times over. It still belongs in the top ten. There is no version of AOAV's methodology under which al-Ahli is correctly absent. Recorded by the rules, it tops the table. Corrected against the rules, it still ranks. Omitted entirely, it reveals a choice.
That choice is the whole case. To leave al-Ahli out, AOAV had to do the one thing it tells the world it does not do: look past the day-one wire copy, weigh the later corrections, and decide the incident did not belong. An event missing from a top-ten ranking might be an editorial judgment about where a line falls. An event missing from the entire report — when AOAV names, profiles, and tallies every other major Gaza incident of 2023, and builds its headline Gaza total from them — is not a line-drawing problem. It is a removal. The capacity to follow up and verify plainly exists; the report's own caveats about Gaza undercounting show AOAV reading the sources closely. That capacity was exercised on the one 2023 Gaza incident almost everyone agrees a Palestinian faction caused and that pointed away from Israel, and it was switched off for the convoy strike Israel denied, for the thousands of unlabeled Gaza dead booked as civilians, and for the Hezbollah fighters counted as bystanders to their own pagers. The verification machinery runs in exactly one direction.
Set the two 2023 incidents side by side and the pattern is unmistakable. The blast that pointed away from Israel — disputed attribution, a toll its own sourcing showed was inflated — was dropped from the ranking despite belonging at or near the top of it. The blast that pointed at Israel — a Hamas claim the IDF denied, with no evidence of a strike beyond the accusation — was elevated to settled fact and placed third. A monitor genuinely indifferent to which side a headline blamed could not sort two events so cleanly along that line.
This is why AOAV's standing is the heart of the problem rather than a footnote to it. The organization's authority rests entirely on the claim that it neutrally records what the media reports. The convoy entry shows it discarding the media's own caveats when they protect Israel. The Gaza and Lebanon totals show the civilian-by-default rule manufacturing combatant-free wars out of conflicts against armed organizations. And al-Ahli shows that the rule can be suspended at will, exercised precisely when suspension serves the narrative. Each example points the same way, and the cumulative weight is hard to read as accident. An organization with this method, applied this selectively, is not producing a flawed measurement of civilian harm. It is producing an argument, and dressing it as data — then handing it to the United Nations, which cites it as the considered judgment of an independent observer.
In short, the only consistency that AOAV shows in its methodology in the Middle East is whatever makes Israel look as bad as possible.
Monday, June 01, 2026
Elder of Ziyon
Buy 
