Every correction made to a published entry is listed here, with its date, what changed, and
the version that was replaced. Corrections are never silent.
A reader can therefore see how often an entry has been wrong, and what changed.
What counts as a correction
A figure, a date, a locator or a quotation was wrong
corrected, with the previous value shown
The source published a correction, an erratum or a new version
the entry is updated and says which version it now rests on
A source was withdrawn or retracted
the entry is marked retracted and stays visible with the reason
We dropped a caveat the source attached to a figure
corrected — this is a falsification even when unintentional
A typographical fix that changes no meaning is not a correction and is not listed here.
Everything else is.
While a check is in progress
When a source publishes something that may affect a published entry, the entry carries a
banner saying it is under review, before anyone has decided what to change. The banner
states that a correction exists on the source; it does not assert anything new, and the
finding itself is not touched until a human has read both.
The entry cites the final version of 14 June 2024, which was read on 4 October 2026. The earlier date was that of the advance unedited version of 27 May 2024.
date of consultation in the citation block (urldate)
The quotations follow the 14 June version. The cause the report gives for the limit on contact in Gaza is now carried with it. The phrase “Israeli-sourced” is dropped: the report’s footnote on the hostages cites the Israeli Ministry of Foreign Affairs at the end of the paragraph without saying which figures it covers. The cover sentence belonged to the advance version.
source_reserves
Paragraph 5: Israeli officials “repeatedly publicly announced Israel’s refusal to cooperate with the Commission’s investigation”; in Gaza its “ability to reach out to victims and witnesses was limited”. Hostages: “Two Israeli-sourced accounts of that day give different hostage figures”, quoting “At least 252 people were abducted to Gaza as hostages”. Cover: “Advance unedited version”, “Distr.: General 27 May 2024”, “an overview of those papers”. → Paragraph 5: Israeli officials “repeatedly announced the refusal of the Government to cooperate with the Commission’s investigation”; in Gaza “the ability of the Commission to make contact with victims and witnesses was limited owing to the continuing fighting on the ground and the major disruption of communications”. Hostages: “Two accounts of that day give different hostage figures”, quoting “At least 252 people were abducted and taken to Gaza as hostages”. Paragraph 1: “Two conference room papers present the detailed findings of the Commission on both situations”, to be read “in conjunction with those papers”.
In the 14 June version the footnote is attached to the end of the sentence, not to “at least 280 women”, and it is worded differently. The reserve and the method note now say what the footnote says and where it sits. The report gives only a web address for the website; the description of that page as one “for victims of terror” was the entry’s reading of a word in the address, and is removed.
source_reserves ; method
Reserve: “The footnote marker sits on “at least 280 women”, so how far it reaches — that figure alone, or the whole civilian breakdown — is not stated.” The footnote read “Israeli government data does not provide disaggregated figures for the number of civilians killed” and “The Commission based this figure on a manual count through the website of the Israeli National Insurance”. Method: “… linking its page for victims of terror.” → Reserve: “The footnote marker sits at the end of the sentence giving the civilian breakdown, and the note speaks of “the above figure” without saying which of the figures in that sentence it means.” The footnote reads “Israeli Government data do not provide disaggregated figures for the number of civilians killed” and “The Commission based the above figure on a manual count obtained through the website of the National Insurance Institute of Israel”. Method: “… and gives a web address on that website, marked “(in Hebrew)”, whose path reads “TerrorVictims”.”
The WHO report’s sentence names no place; “in Israel” was the entry’s addition, and a count “since the hostilities started” may include deaths elsewhere.
source_reserves
“WHO’s situation report of 1 November 2023 (entry whoboth-2023-001) gives a different Israeli-sourced figure for people killed in Israel: “According to Israeli authorities, ~1400 Israelis and foreign nationals have been killed”, counted since the hostilities started.” → “WHO’s situation report of 1 November 2023 (entry whoboth-2023-001) gives a different Israeli-sourced figure: “According to Israeli authorities, ~1400 Israelis and foreign nationals have been killed”, counted since the hostilities started.”
The version of the report dated 14 June 2024 was brought in by hand on 4 October 2026 and read in full. It is not worded like the advance unedited version of 27 May 2024 on which the entry rested, and the address of the report opens the 14 June version. Paragraph 9 is on page 3 in both and gives the same counts of people (more than 1,200 killed, at least 809 civilians, at least 280 women, 68 foreign nationals, 314 Israeli military personnel, at least 252 hostages); the date of the hostage count differs (21 May 2024 in the advance version, 26 May 2024 in this one). The quotations, the section title and the date now follow the 14 June version.
published ; locator ; quote ; supporting_quotes ; sourcing
published 2024-05-27, from the advance unedited version. Quote: “… Of these, at least 809 were civilians, including at least 280 women, 68 foreign nationals and 314 Israeli military personnel.” Section: “Factual findings: acts committed by the Hamas Izz ad-Din al-Qassam Brigades and other Palestinian armed groups on 7 October 2023 in Israel”. → published 2024-06-14, from the version the report’s address opens. Quote: “… Among those killed, at least 809 were civilians, including at least 280 women and 68 foreign nationals, and 314 were Israeli military personnel.” Section: “Factual findings: acts committed by the Izz al-Din al-Qassam Brigades and other Palestinian armed groups on 7 October 2023 in Israel”. The two supporting sentences on children and hostages follow the same version.
The address shown after “re-opened … at the address below” was cut after 90 characters, so the link led nowhere. It now gives the full address of the document.
attestation line shown under the table (address)
The thread called the source a “daily briefing” (it is a press statement of that date), said the Ministry was “counting its own dead”, introduced “a third item” when it had given only one, and derived the Arabic word from شهيد (shahid), a derivation not confirmed by an Arabic speaker. The thread now says only what the statement says. The entry itself is unchanged by this; its corrections are listed in its own journal.
social media thread (Bluesky and Mastodon)
Opening message of the thread sent earlier: “On 20 October 2023 — the fourteenth day — the Palestinian Ministry of Health stated at its daily briefing: 4,137 killed and 13,162 injured in the Gaza Strip. It is a party to the conflict counting its own dead. What is recorded is what it said, on the day it said it.” (and the following messages of the same thread). → Opening message of the replacement thread: “On 20 October 2023, the fourteenth day of the war, the Palestinian Ministry of Health — a party to the conflict — stated that 4,137 people had been killed and 13,162 injured in the Gaza Strip. A statement of that date, not an independent count.” (and the following messages of the same thread).
The full report in English, read end to end on 4 October 2026, states the 70% figure on its first page with a wider base (“all land seized by settlers to date”) than in its findings on page 20 and than in the Hebrew summary page (the area taken over for grazing). The previous reserve gave the Hebrew version and said the English summary contained no occurrence of the figure, which holds for the summary page but not for the full report. The claim, the quotation and the figure of 786,000 are unchanged.
source_reserves
The reserve on the 70% figure read: “The Hebrew summary carries, in the same paragraph, a figure the English summary does not: settlers took over 70% of the total grazing-seized area within about two and a half years (machine translation of the Hebrew, not reviewed by a human). That is a statement about the pace of the seizure, and the English summary contains no occurrence of it.” It gave the Hebrew summary’s version only. → The reserve now adds that the full English report gives a 70% figure in two places and words its base differently: “In less than three years, 70% of all land seized by settlers to date has been taken under the guise of grazing activities” (printed p. 1) and, in its findings, “within just two and a half years, settlers managed to seize 70% of the total area they had taken over since the 1990s” (printed p. 20). “The English summary” is now “the English summary page”.
The thread said the opinion “expressly excludes Israel’s conduct in Gaza after 7 October 2023”, and that reading it as a ruling on the war “attributes to the Court what it declined to address”. The Court writes that the policies and practices contemplated by the request “do not include conduct by Israel in the Gaza Strip in response to the attack” of 7 October 2023; the rest was the register’s reading. The Court’s words are now quoted instead. The entry itself is unchanged by this; its corrections are listed in its own journal.
social media thread (Bluesky and Mastodon)
Opening message of the thread sent earlier: “The International Court of Justice, by 11 votes to 4: Israel’s continued presence in the Occupied Palestinian Territory is unlawful. The opinion expressly excludes Israel’s conduct in Gaza after 7 October 2023. The Court says so itself.” (and the following messages of the same thread). → Opening message of the replacement thread: “The International Court of Justice, by 11 votes to 4: Israel’s continued presence in the Occupied Palestinian Territory is unlawful. The policies and practices put to the Court “do not include conduct by Israel in the Gaza Strip in response to the attack” of 7 October 2023.” (and the following messages of the same thread).
A second address for the same opinion was opened and compared on 4 October 2026. The page references, paragraph numbers and quotations of the entry are unchanged; the sentence only tells a reader who opens the other address where to look and what differs.
sourcing.locator
… Both page pairs were taken from the document itself. The operative paragraph number is given as well. → … Both page pairs were taken from the document itself. The operative paragraph number is given as well. The same opinion is served by the Court’s website at …186-20240719-adv-01-00-en.pdf, an 83-page edition with its own pagination, downloaded by hand on 4 October 2026. In it, paragraph 1 is on printed page 7 (page 10 of the file), paragraph 81 on printed page 27 (page 30), paragraph 284 on printed page 77 (page 80) and operative paragraph 285 on printed page 78 (page 81). The paragraph numbers and the cited sentences are the same in the two editions; the one difference is in the parenthesis that ends paragraph 284: it reads (paragraphs 72-83) in the Reports text and (paragraphs 72 to 83) in that edition.
The address shown after “re-opened … at the address below” was cut after 90 characters, so the link led nowhere. It now gives the full address of the document.
attestation line shown under the table (address)
The address shown after “re-opened … at the address below” was cut after 90 characters, so the link led nowhere. It now gives the full address of the document.
attestation line shown under the table (address)
The thread said the authors “mark that one as the outlier”; they write that the Bayesian model averaging model “estimated fewer deaths” than the other models. It also summarised the authors’ list of excluded deaths as “the collapse of health services, hunger or water”, which is stronger than their words (“health service disruption, food insecurity, and inadequate water and sanitation”). The authors’ words are now quoted instead. The entry itself is unchanged by this; its corrections are listed in its own journal.
social media thread (Bluesky and Mastodon)
Opening message of the thread sent earlier: “The Lancet: a capture–recapture analysis estimates 64,260 deaths from traumatic injury in the Gaza Strip, 7 Oct 2023 to 30 Jun 2024 (95% CI 55,298–78,525). The authors’ own bound: traumatic injury only — not deaths from the collapse of health services, hunger or water.” (and the following messages of the same thread). → Opening message of the replacement thread: “The Lancet: a capture–recapture analysis estimates 64,260 deaths from traumatic injury in the Gaza Strip, 7 Oct 2023 to 30 Jun 2024 (95% CI 55,298–78,525). The authors’ own bound: traumatic injury only. They “do not account for non-trauma-related deaths”.” (and the following messages of the same thread).
The publisher’s PDF of the journal issue was brought in on 4 October 2026 and compared with the PDF read on 20 September 2026. The text is identical, the pages correspond as stated, and four page breaks differ by one line. The figures, quotations and page references of the entry are unchanged.
sourcing.document ; sourcing.locator
sourcing.document: “… was read end to end on 20 September 2026 from a copy held by the authors' institutional repository at the London School of Hygiene & Tropical Medicine, which carries the publisher's own pagination. …” sourcing.locator: “… The journal issue prints the article on pages 469–477; the page-by-page correspondence between the two is not established here.” → sourcing.document: “… was read end to end on 20 September 2026 from the PDF ‘Published online January 9, 2025’, held by the authors’ institutional repository … The PDF of the journal issue (Lancet 2025; 405: 469–77) was downloaded by hand from the publisher’s page, in an ordinary browser, on 4 October 2026 and compared with it word for word: the text is identical, and the quotations of this entry were checked against that file. …” sourcing.locator: “… page 1 of the file is p. 469 and page 7 is p. 475. The two PDFs break pages at slightly different places: at the foot of pages 4, 5, 6 and 7 of the PDF of 9 January 2025, one line stands at the head of the next page in the issue PDF.”
The address shown after “re-opened … at the address below” was cut after 90 characters, so the link led nowhere. It now gives the full address of the document.
attestation line shown under the table (address)
The thread said the Comptroller was “a party to the conflict counting its own dead”. The report is an audit of the administration of compensation, not a count of deaths; the figures are given as context. The thread now says what the report says and that the Comptroller is an organ of a party to the conflict. The entry itself is unchanged by this; its corrections are listed in its own journal.
social media thread (Bluesky and Mastodon)
Opening message of the thread sent earlier: “The State Comptroller of Israel, on 7 October 2023: hundreds of soldiers and over 1,000 Israelis and foreigners murdered, thousands injured, 251 hostages taken into Gaza. It is a party to the conflict counting its own dead. “Over 1,000” is the report’s own approximation.” (and the following messages of the same thread). → Opening message of the replacement thread: “The State Comptroller of Israel, an organ of a party to the conflict, writes that on 7 October 2023 Hamas “murdered hundreds of soldiers, over 1,000 Israelis and foreigners”. “Over 1,000” is the report’s own approximation.” (and the following messages of the same thread).
The two references now follow the version of the Commission’s report dated 14 June 2024, which the report’s address opens; the hostage sentence is quoted in that version’s words, and the figures are unchanged (see the journal of entry coi7oct-2024-001).
source_reserves
“… the UN Commission of Inquiry, in A/HRC/56/26 (advance unedited version of 27 May 2024), says “At least 252 people were abducted to Gaza as hostages”” and “The Commission of Inquiry, in A/HRC/56/26 (advance unedited version of 27 May 2024), paragraph 9, writes …”. “Two Israeli-sourced accounts of that day …”. → “… the UN Commission of Inquiry, in A/HRC/56/26 (14 June 2024), says “At least 252 people were abducted and taken to Gaza as hostages”” and “The Commission of Inquiry, in A/HRC/56/26 (14 June 2024), paragraph 9, writes …”. “Two accounts of that day …”.
The thread stated, in the register’s own voice, that an outpost is “unlawful under Israeli law itself”; predicted which figure “is most likely to be lifted”; and said that outposts and settlements “do not add up”, an inference the report does not make. The report’s own words are now quoted instead. The entry itself is unchanged by this; its corrections are listed in its own journal.
social media thread (Bluesky and Mastodon)
Opening message of the thread sent earlier: “Peace Now: in 2025 the Israeli government established 86 outposts in the West Bank, 58 of them “agricultural farms”. The summary of that same report says 60.” (and the following messages of the same thread). → Opening message of the replacement thread: “Peace Now: in 2025 the Israeli government established 86 outposts in the West Bank, 58 of them “agricultural farms”. The summary in the English edition says 60.” (and the following messages of the same thread).
In paragraph 9 the words “According to Israeli sources” open the first sentence only. The footnote speaks of “this figure” without saying which, so the earlier statement that the whole civilian breakdown is the Commission’s own count went further than the report. Paragraph 21 gives the foreign nationals as “At least 68”; the floor had been dropped from the claim.
claim ; method ; source_reserves
claim: “… states, attributing the figures to Israeli sources, that more than 1,200 people were killed directly in the 7 October 2023 attack, of whom at least 809 were civilians — including at least 280 women and 68 foreign nationals — and 314 were Israeli military personnel.” method: “The total of more than 1,200 is attributed to Israeli sources. The civilian breakdown is not …” Reserve: “The civilian figures are the Commission’s own count, not a figure Israel published.” → claim: “… states that, according to Israeli sources, more than 1,200 people were killed directly in the 7 October 2023 attack, and that of these at least 809 were civilians — including at least 280 women and at least 68 foreign nationals — and 314 were Israeli military personnel.” method: the sentence giving the total opens “According to Israeli sources”; the sentence giving the civilian breakdown carries no attribution of its own. Reserve: a footnote to the paragraph states that the Commission based “this figure” on a manual count through the website of the Israeli National Insurance; the footnote marker sits on “at least 280 women”, so how far it reaches is not stated.
The figure 1,123 is written nowhere in the report; it was a sum made here. Since 809 is a floor and 1,200 is “more than”, a sum below 1,200 shows no gap, so the statement that the figures do not sum is withdrawn. What the report does not say, how its figures partition the total, is kept.
source_reserves
The figures do not sum, and the sentence does not say how they partition. … Read the other way, 809 civilians and 314 military make 1,123, short of “more than 1,200”. Both civilian figures are floors, and the 40 children the paragraph names next are not placed in either group. … → The sentence does not say how the figures partition the total. … The figures for civilians and for women are floors, and so is the figure for foreign nationals, which paragraph 21 gives as “At least 68”. The 40 children the paragraph names next are not placed in either group. The breakdown is not given as a partition of the total, and adding or subtracting these numbers to obtain a third would be an operation the source does not perform.
Two sources of the register attribute to Israeli authorities different totals of people killed on and after 7 October 2023; only the hostage figures had been compared.
source_reserves
(none) → WHO’s situation report of 1 November 2023 (entry whoboth-2023-001) gives a different Israeli-sourced figure for people killed in Israel: “According to Israeli authorities, ~1400 Israelis and foreign nationals have been killed”, counted since the hostilities started. The two figures are attributed to Israeli sources at different dates.
In the report, section II is “Legal framework”; the factual findings are section III. The printed page number, which the provenance had said could not be established, is printed on each page.
locator
A/hrc/56/26, paragraph 9 (PDF page 3), Section II.A, “Factual findings: … on 7 October 2023 in Israel — Introduction” → A/HRC/56/26, paragraph 9 (page 3), Section III “Factual findings: … on 7 October 2023 in Israel”, sub-section A “Introduction”
The text on which every quotation was checked is the advance unedited version dated 27 May 2024; the date of 14 June 2024 came from the cover of another file. The date shown is now that of the text read.
published ; sourcing.document ; source_reserves
published: 2024-06-14. No reserve on the nature of the version read. → published: 2024-05-27. Reserve added: the report’s cover reads “Advance unedited version” and “Distr.: General 27 May 2024”, and the report states that it is “an overview of those papers”, the conference room papers A/HRC/56/CRP.3 and A/HRC/56/CRP.4, and “should be read in conjunction with them”. The provenance now says that a file served by the United Nations documents server carries the date 14 June 2024 and that the two files have not been compared word for word.
Two glosses in the register's voice, that the authors limit the generalisation to students and that the count is a floor, went further than the authors, who in the Discussion extend their findings with caution to other young Palestinians and write only that it is “possible” that the most affected were underrepresented. The population is “nearly 2400”, not “roughly”.
source_reserves ; methodology_note
The authors bound the generalisation to that group: “an appropriate sample size was ensured …”. The survey selects against the very thing it measures, and the authors say so: “… it is possible that those most affected by the conflict … were underrepresented in the sample.” A count of displacements taken from those who still had connectivity is a floor. → The authors write that “an appropriate sample size was ensured to retain the generalizability of study results to its population” — the study’s population being the nearly 2,400 students enrolled at the two schools. The authors write: “… it is possible that those most affected by the conflict … were underrepresented in the sample.” They do not say how this would affect the displacement counts. In the Discussion the authors call the sample “a representative sample of male and female medical students from the 18–24 age group”, and write that the findings “may help infer the psychological impacts of the assault on other Palestinians from similar age groups”.
The questionnaire counts displacements “during the conflict” and the article says that only 27 (7.96%) “said they had not been displaced”. The claim gave no period and said “never”, which reads over a lifetime.
claim ; figure.unit
In an online survey of 339 medical students … 8.55% reported having been displaced ten or more times, while 7.96% reported never having been displaced. (unit: of the 339 respondents displaced ten or more times) → … 8.55% reported having been displaced ten or more times during the conflict, while 7.96% reported not having been displaced. (unit: of the 339 respondents displaced ten or more times during the conflict)
The respondents are 339 medical students, not adults in general. The population had been corrected in the claim before publication but not in this field, which stayed online and appeared above a line saying the survey is of medical students. A note on the page that referred to the drafting of the entry is removed.
method
Cross-sectional survey of adults in Gaza recording, among other variables, the number of times each respondent had been displaced. → Cross-sectional, internet-based survey of undergraduate medical students from the two medical schools in the Gaza Strip, recording, among other variables, the number of times each respondent had been displaced “during the conflict”.
Neither version of the briefing says that the count is partial; it says that “A large number of victims are still under the rubble”, without saying whether they are counted. The word “daily” is not in the document. The qualification partial-data attributed to the source an admission it does not make.
method ; figure.qualifier ; figure.unit
method: “Partial by the source's own account. A running count announced at a daily press briefing. …” qualifier: partial-data. unit: “… as counted by the Palestinian Ministry of Health”. → method: “A running count announced at a press briefing. Neither version of this briefing states how the count is compiled, what it excludes, or whether the victims still under the rubble are in it. … The Ministry does not describe its count as partial.” qualifier: unverified. unit: “… as stated by the Palestinian Ministry of Health”.
The two versions differ in more places than the three that are carried, for instance an Arabic item saying that Christian families were wiped from the civil register, which has no English counterpart. The wording no longer says there are only three differences.
source_reserves ; corrections[0].notice
… the two texts are not the same. Three items. … (banner: the entry now carries the three items where they differ) → … the two texts are not the same. Three examples. … (banner: the entry now carries three examples of where they differ)
Statements about the register's own posture are removed and the Ministry’s characterisations are attributed to it. The document does not say that the Ministry counts its own dead. The Lancet study is now carried with both of its statements, the underestimate and the possible inclusion of non-trauma deaths.
source_reserves ; data_provenance
“The words are the Ministry’s, not ours. … These are the source’s characterisations; the registry reports that the source makes them and takes no position on them.” “The source is a party to the conflict counting its own dead. What is recorded here is what the Ministry stated …” “An organ of a party to the conflict, publishing figures about that party's own losses. … a capture-recapture analysis published in The Lancet estimated that the Ministry's list under-counted traumatic-injury deaths …” → “… These are the Ministry’s own characterisations.” “The source is a party to the conflict. The figure is the Ministry’s statement of that date, not an independent count.” “An organ of a party to the conflict. … a capture-recapture analysis published in The Lancet writes that its results support ‘the view that the MoH figures are more likely to underestimate than overestimate mortality’, and that ‘some non-trauma deaths might have been included’ in the Ministry’s lists …”
The full English report carries the 70% figure, so only the English summary lacks it. Neither edition calls the firing-zone point a contradiction; the word was the register's. Two findings of the report were stated in the register's voice before being quoted and are now attributed to the report.
source_reserves
(1) “The Hebrew version carries, in the same paragraph, a figure the English does not: settlers took over 70% of the total grazing-seized area within about two and a half years.” (2) “The report states it as a contradiction in the state's own rules: …” (3) “Most of the land taken is not ‘state land’, even by Israel’s own definitions. In the source’s words: …” → (1) “… a figure the English summary does not …”, with the English rendering of the Hebrew sentence marked as machine translation, not reviewed by a human. (2) “The report says that around 41% of the seized land is inside areas settlers are themselves officially forbidden to enter. In the source’s words: …” (3) “The report says that most of the land taken is not ‘state land’, even by Israel’s own definitions. In the source’s words: …”
The full report, read on 3 October 2026, writes “approximately” and “assessments” where the summary writes “at least”, so the register's own note, which stated the figure to be a floor, could not stand. The removed date is that of the summary page; neither the page nor the full report dates the figure itself.
registry_note ; source_reserves ; figure.as_of
registry_note: “The figure is a floor, not an estimate: the report says ‘at least’.” The figure carried the date “as of 2025-04-06”. No reserve described how the full report qualifies its own figure. → registry_note: “The report qualifies the figure differently from one text to another: ‘at least’ in the English summary, ‘about’ (כ-) in the Hebrew summary, ‘approximately’ in the key findings of the English full report.” Reserves added: the full report calls the figure an assessment (“Our assessments indicate that settlers living in shepherding outposts currently control approximately 786,000 dunams”); it says that “control” does not necessarily mean that Palestinian entry is entirely prohibited; it is dated December 2024. The date “as of 2025-04-06” is removed from the figure.
The description of the method was not in the cited page. The full report describes it, and its own words are now carried.
method
Joint field and mapping study by Kerem Navot and Peace Now of land taken over by settlers for grazing. → The summary page does not describe how the 786,000 dunams were measured; it refers to “An analysis of the land seized by Israeli settlers”. The full report’s methodology section (printed p. 16) says the lands were “assessed and mapped by observing signs that Palestinian agricultural lands had remained uncultivated for extended periods, as well as evidence of road construction, water infrastructure installation, and the construction of fences and gates”, and that “interviews and discussions with Palestinian shepherds and farmers” provided “crucial insights”.
The statement about the publisher's standing is in neither the page nor the report and is removed. The origin of the data, which the document does state, is carried instead.
data_provenance
Peace Now is an Israeli organisation whose Settlement Watch has been the reference on this subject for decades, and which counts the policy of its own government. It is not a party to the conflict, but it operates under the authority of one of them and documents that authority's acts. Kerem Navot is likewise an Israeli research organisation. → The 786,000 figure is the authors’ own assessment. The table’s ownership categories rest on Israeli state sources, as its footnote and row labels say: “Lands marked as private on Israeli Civil Administration maps”, “Lands with no ownership information on Israeli Civil Administration maps”, “State lands registered in the Israeli land registry”. The 80,000-dunam allocation figure is the State Attorney’s Office’s, as the authors report it. Peace Now and Kerem Navot are Israeli organisations, not parties to the conflict, operating under the authority of one party and documenting that authority’s acts.
The sentence on the allocation is in Part Two of the English edition. The example about a single plot is in neither the summary nor the full report, and is removed.
source_reserves
The 80,000-dunam allocation figure … was located under “Part One: Seizing Control of Hundreds of Thousands of Dunams Through Shepherding Outposts”. The table note read: “A plot can be a firing zone and private Palestinian land at once. Read down the percentage column and the seven rows total 152%, not 100%; …”. → The 80,000-dunam allocation figure is located under “Part Two: Grazing Land Allocation Contracts with Settlers”. The sentence “A plot can be a firing zone and private Palestinian land at once” is removed; the table note now reads that the report’s own “only 40%” for state land matches the sum of two of those rows, 15.5% and 25.3% (40.8%).
The appeal is dated October 2023 on its cover; 24 October is the date of its figures, and the interval to 1 November is eight days. The report writes “WHO has documented 216 attacks” with no qualification, so the qualifier at-least was the register's reading, not the source's.
source_reserves ; figure.qualifier
Reserve: “WHO’s own figures a week earlier … Its funding appeal of 24 October 2023 states: ‘There have been 171 attacks on health care in oPt leading to 493 deaths and 387 injuries.’ … the 491 here covers Gaza alone, eight days later.” Qualifier: at-least. → Reserve: “WHO’s own figures eight days earlier … Its funding appeal of October 2023 gives, as of 24 October 2023: ‘There have been 171 attacks on health care in oPt leading to 493 deaths and 387 injuries’. …”. A reserve is added: the report gives no count of fatalities or injuries for the 123 attacks in the West Bank. Qualifier: exact.
Neither the report nor WHO’s appeal says that a documented attack is one that WHO was able to verify, or who supplies the reports behind the count. Those statements about WHO’s method and about the origin of the count were the register's.
source_reserves ; data_provenance
Reserve: “A documented attack is an attack WHO was able to verify. It is a floor on what occurred, not a total.” data_provenance: “WHO's own surveillance system — one of the few counts in this registry produced by neither party.” → Reserve: “The report gives 216 as the number of attacks WHO has documented (‘WHO has documented 216 attacks’) and does not say that it covers every attack that occurred.” data_provenance: “The report gives this count as WHO’s own: ‘WHO has documented’, footnoted ‘*WHO SSA’. It does not say who supplied the reports behind it.”
The earlier statement that the robot sent no request to the Court was wrong: its requests were refused. The provenance now says what was attempted and from which copies the document was read.
sourcing.document
… The robot sent no request to the Court, and the document is not kept here, so it could not be re-opened on 20 September 2026 when the other published sources were re-read. → The Court’s website refused the robot’s requests for the PDF — three on 17 September 2026, and again at the re-check of 28 September 2026 — … The PDF, the printed I.C.J. Reports text of 107 pages, was opened … by hand, on 19 September 2026, and brought in by hand a second time on 3 October 2026; it was read from those copies. The robot could not re-open the document on 20 September 2026 when the other published sources were re-read.
The opinion excludes conduct by Israel in Gaza “in response to” the attack of 7 October 2023, not all conduct in Gaza after that date; the words “expressly excludes … after 7 October” and “declined to address” were the register's, not the Court's. The Court's own sentence on how its reply is to be read (paragraph 284) is now carried.
source_reserves
The opinion expressly excludes Israel’s conduct in Gaza after 7 October 2023. The Court writes that the policies and practices put to it by the General Assembly “do not include conduct by Israel in the Gaza Strip in response to the attack carried out against it by Hamas and other armed groups on 7 October 2023”. The request was adopted in December 2022. Reading this opinion as a ruling on the war would attribute to the Court something it declined to address. → The Court writes that the policies and practices contemplated by the General Assembly’s request, adopted on 30 December 2022, “do not include conduct by Israel in the Gaza Strip in response to the attack carried out against it by Hamas and other armed groups on 7 October 2023”. It adds that its reply “rests on the totality of the legal grounds set forth by the Court above, each of which is to be read in the light of the others, taking into account the framing by the Court of the material, territorial and temporal scope of the questions (paragraphs 72-83)”.
The earlier sentence said of the document what was true only of the register's robot: the Court's website refuses automated requests, while the PDF opens in an ordinary browser. The same wording is removed for every entry.
attestation line shown under the table
This document is reachable from no public address today. The copy it was read from is named in Provenance. → This document could not be re-opened automatically on 2026-09-28; see Provenance.
The figure of 450 is the report's own, on its summary page. The claim, the quotation and the figure of 61,000 are unchanged.
supporting_quotes
None. The claim gave an annual average of about 450 recognitions in 2009 to 2022, and no quotation of the entry carried that figure. → One sentence of the report added, with its page: “about 61,000 people were recognized as victims of hostile attacks by the audit end date (March 2024), compared to 450 people who were recognized on average each year from 2009 to 2022” (printed p. 78, Summary).
The Abstracts volume, the document most readers open first, gives only the second total.
source_reserves
The report gives two totals a few pages apart: “about 61,000” and “a total of about 61,100”. Both are the source’s, and both are approximations it marks as such. → … The Comptroller’s Abstracts volume of the same series, dated February 2025 on its cover, gives “about 61,100 individuals by the audit completion date (March 2024)”. All three are the source’s, and all are approximations it marks as such.
The report’s own reserve in the other direction was missing: the unified recognition left out part of the residents of the affected areas. The restriction in the report’s description of the 54,700 had also been dropped.
source_reserves ; method
No reserve on the extent of the recognition. method: “… covering residents of areas within 0–7 km of the Gaza Strip and participants in the dance parties; …” → Reserve added: the report states that the recognition “did not encompass all residents situated in areas where hostile attacks or terrorist infiltrations took place, particularly in Ofakim, where unified recognition was limited to residents of select streets”, and that it was later “expanded to include additional streets”; the Ofakim Municipality “has asserted that the entirety of the city … should be classified as a war zone”, and a petition was filed with the High Court of Justice in May 2024. method: “… residents of regions within 0–7 km of the Gaza Strip ‘which were subject to terrorist infiltration or direct combat’, and participants in the dance parties; …”
No sentence of the report says that a death does not necessarily count among the recognised. The report says that everyone recognised through the unified process is classified as having a temporary mental injury, and does not say whether people who were killed are among the 61,000.
source_reserves
“Recognised” is an administrative decision conferring a status and an entitlement. It is not a count of people killed or injured: a person compensated for a temporary stress injury counts here, and a death does not necessarily. → “Recognised” is an administrative decision conferring a status and an entitlement. It is not a count of people physically injured or killed: the report classifies everyone recognised through the unified process as having a temporary mental injury (acute stress disorder). The report speaks of “the rights of citizens and foreign nationals suffering harm due to hostile attacks, as well as the rights of bereaved families of individuals killed in such events”, and does not say whether people who were killed are among the 61,000.
A judgment of the document’s value and a summary that called the report’s findings “failings” were not the report’s words. The sentence quoted ends item 2 of a list and does not begin a paragraph. The title shown for the document was written nowhere in it; the document prints “Regulating Assistance for Victims of the October 7th Attack”.
data_provenance ; registry_note ; locator ; source_reserves[].locator
data_provenance: “… What makes the document valuable is that it documents the failings of its own government — a reason to read it, not a reason to treat it as neutral.” registry_note: “This document is about the war's impact on Israeli civilians and the failings of the Israeli state towards them. It says nothing about Palestinian losses, and its figures cannot be set beside a Gaza toll.” locator: “Printed p. 69, paragraph beginning ‘As of the audit end date’” → data_provenance: “An organ of the State of Israel, a party to the conflict, auditing its own administration.” registry_note: “This document is an audit of how Israel recognises and compensates victims of hostile attacks since 7 October 2023, of the long-term cost to the state treasury, and of the need for a response to thousands of people who experienced psychological trauma from the attack but are not legally recognised as victims under the Compensation Law. It says nothing about Palestinian losses.” locator: “Printed p. 69, Key Findings, under ‘Economic Estimates for the Increase in Compensation for Victims of the October 7 Attack…’, item 2, last sentence, beginning ‘As of the audit end date’”. Printed pp. 66–67 for the sentence that crosses the page break.
The earlier text called the survey “independent” and said that the two studies converge “on under-reporting rather than exaggeration”, which were the register's words. The corroborating source itself states a limit to the corroboration, and a letter and the authors’ reply in the same journal had not been mentioned.
methodology_note
A commentary on this article was published in the same issue (The Lancet 2025; 405(10477): 440–442). Corroboration: the Gaza Mortality Survey, an independent household survey, … “closely matches a separate estimate made using capture-recapture methods” — this study. Two different methods over two different periods, converging on under-reporting rather than exaggeration (entry gmsurvey-2025-001). → Published responses in The Lancet are named: the commentary by Smith, Shellah and Samudzi (its text has not been read); in the issue of 7 June 2025 a Correspondence by Zivot, Paltiel, Sim and Wald, “Gaza deaths and lawful warfare”, and the authors’ reply (“Our estimate is approximately 1·7 times the Gaza Ministry of Health’s estimate of deaths (64 260 vs 37 877) for the same time period, not three times”). Corroboration: the Gaza Mortality Survey (a household survey by a different team of authors) writes that its estimate “closely matches a separate estimate made using capture-recapture methods”; its authors also write that two of this study’s three data sources “are components of the GMoH database itself”, so that the study “can offer only limited and partially circular assessment of GMoH overall numbers”; and a published letter writes that the survey’s estimates “should not be regarded as reliable measures of conflict-related deaths”.
The word “outlier” and the weight given to the authors’ own limit were the register's, not the article's. Four statements the authors make about the limits of the figure were absent, among them one that runs in the opposite direction to the others.
source_reserves
(1) “… The last falls below the lower bound of the published interval, and the authors mark it as the outlier.” (2) “The authors bound their own estimate, and the bound is the most important thing about it: …” (3) “930 people reported missing were excluded from the main analysis. The authors run a separate sensitivity analysis that includes them as assumed decedents; the headline figure does not.” (4) No reserve on non-trauma deaths in the lists, on residual dependence among lists, or on the eight models. → (1) “… the authors write that the BMA model ‘estimated fewer deaths’ than the other models.” (2) “The authors write: ‘Our findings underestimate the full impact of the military operation in Gaza, …’” (3) “… the overall estimate would increase to 68 662 (95% CI 58 735–84 472; see appendix p 8)”. (4) Added: the authors assumed that the Ministry lists “captured only deaths due to traumatic injury”, and “some non-trauma deaths might have been included, potentially resulting in an overestimation of deaths due to traumatic injury”; “some residual dependence among lists might not have been reflected in the estimates”; the headline figure is an average of eight models, whose estimates of deaths outside any list run from 20,862 to 47,825.
The article writes “suggesting” and “implying”, never “concludes”; and it says “more likely to underestimate than overestimate”, not “not over-reported”. The authors also write that some non-trauma deaths might be in the Ministry’s lists, which would overstate traumatic-injury deaths.
claim ; data_provenance
claim: “… and concludes the Palestinian Ministry of Health under-reported mortality by 41%.” data_provenance: “… The study's finding is that this authority under-reported, not over-reported.” → claim: “… suggesting that the Palestinian Ministry of Health under-reported mortality by 41%.” data_provenance: “… The study estimates that this authority under-reported traumatic-injury deaths by 41% over the period, and writes that its results support ‘the view that the MoH figures are more likely to underestimate than overestimate mortality’.”
The closing statement, that the repository does not forbid the registry, could not be kept: the robots file of that host, read on 3 October 2026, names the crawler this registry uses. The article, its quotations and its figures are unchanged.
sourcing.document
The peer-reviewed article, 9 pages of body text followed by references, was read from the authors' institutional repository at the London School of Hygiene & Tropical Medicine, which carries the publisher's own pagination. It was read end to end on 20 September 2026. thelancet.com forbids this registry by name in its robots file, so no request is sent there; the article was read where it is not forbidden. → The peer-reviewed article, 9 pages including its references, was read end to end on 20 September 2026 from a copy held by the authors' institutional repository at the London School of Hygiene & Tropical Medicine, which carries the publisher's own pagination. thelancet.com forbids this registry by name in its robots file, so no request is sent there.
The same one-hostage difference was marked “contested” here and “single-source” in the entry on the Commission’s report. Neither text disputes the other: they do not cite each other, and the earlier one gives a floor. Both entries now carry the same label, and the difference stays stated in both.
corroboration ; contested_by
corroboration: contested. The “Who disputes it” block of the page was empty: its data was stored as organization, document and ground, names the page template does not read. → corroboration: single-source. The block is removed. The Commission of Inquiry’s figure (A/HRC/56/26, advance unedited version of 27 May 2024, paragraph 9: “At least 252 people were abducted to Gaza as hostages”), against this report’s 251, is stated in a reserve, as it is in the entry on the Commission’s report.
The report’s sentence lists “hundreds of soldiers, over 1,000 Israelis and foreigners” without saying whether the soldiers are included in the 1,000. It does not say where its figures come from. Other figures for the same day, in the Commission’s report and in the Comptroller’s own other volumes, had not been stated.
figure.unit ; method ; source_reserves
unit: “Israelis and foreigners murdered, besides hundreds of soldiers”. method: “… The figures are the state's own, given as context for the audit rather than established by it.” → unit: “Israelis and foreigners murdered, in a sentence that also names ‘hundreds of soldiers’”. method: “… The report gives these figures without a source or a method; they appear in its Abstract as background to the audit.” Reserves added: the Commission of Inquiry’s totals for the same day (“more than 1,200 persons were killed”, “at least 809” civilians, 314 military personnel), and the Comptroller’s other texts (“about 1,000 Israeli and foreign citizens” in February 2025; “about 1,200 people” and “over 240” hostages in January 2024).
A judgment of the document’s value and a summary calling the report’s findings “failings” were not the report’s words, and had no bearing on the claim. The report’s note 4 says that individually recognised victims include people who were physically injured.
data_provenance ; registry_note ; source_reserves
data_provenance: “… What makes the document valuable is that it documents the failings of its own government — a reason to read it, not a reason to treat it as neutral.” registry_note: “This document is about the war's impact on Israeli civilians and the failings of the Israeli state towards them. …” Reserve: “… a single blanket decision … of which about 54,700 …”, “not people physically harmed on the day”. → data_provenance: “An organ of the State of Israel, a party to the conflict, auditing its own administration.” registry_note: “This document is an audit of how Israel recognises and compensates victims of hostile attacks since 7 October 2023, … It says nothing about Palestinian losses.” Reserve: the report’s own words on the 54,700 (“in a joint decision … which were subject to terrorist infiltration or direct combat”) and “not people physically injured or killed on the day”.
The estimate in the other entry is not independent of the list: it takes from it the 6% share of double orphans, and it starts from the Ministry’s own death total. The adjective “independent” was the register's. The authors’ conclusion that the method likely under-reported orphans was not carried, and the link between the figures on identities and the first two exclusions was the register's reading.
registry_note ; data_provenance ; source_reserves
registry_note: “An independent estimate built from the same data is higher than this list: entry orphans-2025-002. The two are published together.” data_provenance: “… What the study does with it is independent of that authority.” Reserve: “They give the scale of the first two: …” → registry_note: “The study’s own estimate for the first three months of the war is entry orphans-2025-002.” data_provenance: “… The study’s authors are affiliated with the Max Planck Institute for Demographic Research, UCLouvain and the London School of Hygiene and Tropical Medicine, and declare no competing interests.” Reserve: the authors write that the Ministry’s method “likely resulted in the under-reporting of orphans as it excluded cases where parents’ deaths were not formally recorded or where identification was impossible, and it omitted newborns who had not been registered in the civil registry due to the ongoing conflict”; in the next sentence, the 24,682 of 34,654 casualties with identities.
The date and nature of the Ministry’s release are reported by the study, which cites a post on the Ministry’s Telegram channel; the entry had stated them as its own. The study calls the 911 double orphans an estimate. The age percentages are the list’s, not the study’s own estimates, and only the under-5 share is compared. The sentence quoted is in the Methods section, not the Results.
claim ; method ; locator
claim: “The Palestinian Ministry of Health released on 1 May 2024 a list of 15,173 children aged 0–17 who had lost one or both parents due to the war; 911 of them — 6% — had lost both.” method: “… compares its age breakdown against the study's own demographic estimates: 25% under 5, 32% aged 5–9, 29% aged 10–14, 14% aged 15–17.” locator: “Results, ‘Comparison of patterns of excess mortality’” → claim: “The study reports that on 1 May 2024 the Palestinian Ministry of Health released a list of 15,173 children aged 0–17 who had lost one or both parents due to the war, and that the list ‘provides an estimate’ of those who had lost both: 911, or 6%.” method: “… It gives the list’s age breakdown — 25% under 5, 32% aged 5–9, 29% aged 10–14, 14% aged 15–17 — and notes that its own estimate for children under 5, 28%, is consistent with the list’s 25%.” locator: “Methods, ‘Comparison of patterns of excess mortality and orphanhood from other sources’”
The title given for the publisher’s correction notice was wrong. The correction itself, which fixes inverted author names and stray symbols in an equation, touches no figure.
sourcing.correction ; sourcing.document ; organization
sourcing.correction: “… « Correction to: A demographic assessment of the impact of the war in the Gaza Strip on the mortality of children and their mothers », …” sourcing.document: “… the sentence above taken from the Results section.” → sourcing.correction: “… « Correction to: … on the mortality of children and their parents in 2023 », …” sourcing.document: “… the sentence above is taken from the Methods section …” organization: the department is given for the first author only.
The comparison with the Ministry list is the authors’, and it sets an estimate for October to December 2023 against a list dated 1 May 2024; the claim stated it as the register's and without the dates.
claim
The same study estimates that, in the first three months of the war, 23,597 children (95% CI 22,906–24,310) lost one or both parents — significantly more than the 15,173 on the Ministry of Health list. → The same study estimates that, in the first three months of the war, 23,597 children (95% CI 22,906–24,310) lost one or both parents, which the authors call “significantly higher than the list of 15,173 orphans released in May by the Palestinian MoH in Gaza” (the study dates that list 1 May 2024).
13,101 of 21,822 is 60%, not “about two thirds”, and the identified share is not “a third”. The interval was shown without saying what it contains, and the limits the authors state in the other direction, and the debate on the Ministry’s total that the study reports, were absent.
method ; source_reserves
“The age-and-sex distribution is available for 13,101 of the 21,822 deaths — about two thirds. The remainder is redistributed on the assumption that the identified third is representative.” No reserve on what the confidence interval covers. → “The age-and-sex distribution covers 13,101 of the 21,822 deaths; the authors write ‘two-thirds of war-related deaths in 2023 (13101/21822)’. They assume that ‘this age- and sex-specific death distribution reflects the distribution of all deaths due to the war before December 31st’ …” Reserves added: the model is deterministic and the authors ran a Monte-Carlo simulation, with uncertainty “incorporated through two elements: 1) uncertainty in the background U5MR estimates from the UN IGME and 2) stochasticity from the additional deaths due to the war”; the authors call their estimates “conservative” and write that “our estimates should be considered with caution”; the Ministry’s counts “have been regularly questioned” while the WHO, Israeli intelligence services and the UN “consider that the MoH estimates reflect the actual number of deaths”.
The authors write “potential decline” and “likely”, where the entry stated a decline and a finding. The adjectives “independent” and “plainly”, and the statement that the two are not rival counts, were the register's. In the public data file the grounds of the contestation had been exported empty; that is repaired for every entry.
source_reserves ; data_provenance ; registry_note ; contested_by
Reserve: “the quality of the Ministry’s data declined as the war went on”. data_provenance: “… analysed independently of it. The study's finding is that the authority's own list understates the number.” registry_note: “The Ministry list gives 15,173 and this estimate gives 23,597. The authors say plainly which of the two is a floor. Both are published here, and neither is preferred.” contested_by: “… the authors describe the list as a lower bound rather than a competing total, so the two are a floor and an estimate rather than two rival counts.” → Reserve: “Among the authors’ reasons is ‘the potential decline in the quality of MoH mortality data as the war progressed’ …”. data_provenance: “… The study’s authors are affiliated with … and declare no competing interests. They write that the Ministry’s method for its orphan list ‘likely resulted in the under-reporting of orphans’.” registry_note: “The Ministry list is entry orphans-2025-001.” contested_by: “The authors write that the list ‘gives only a partial view of the orphan situation’ and that it does ‘provide a lower bound for our estimates’.”
The two editions carry different dates, and the English edition is the one quoted. The date shown for the entry is the Hebrew edition's; the other is now stated.
sourcing.document
Both editions were opened on 19 September 2026 … and both were read end to end on 20 September 2026, side by side. → … The English edition is dated “03.2.26” and the Hebrew edition “02.2.26”; this entry carries the Hebrew date.
The source writes “agricultural farms” in quotation marks, in both editions. The quotation marks were dropped from the claim and from the figure's unit and are restored.
claim ; figure.unit
Peace Now states that in 2025 the Israeli government established 86 outposts in the West Bank, 58 of them agricultural farms. → Peace Now states that in 2025 the Israeli government established 86 outposts in the West Bank, 58 of them “agricultural farms”.
Both editions were read again on 3 October 2026. Neither describes how the outposts were counted; the method that had been written is in neither.
method
Annual count by Peace Now's Settlement Watch team, drawing on aerial photographs, government documents and field visits. → The page does not say how the 86 outposts were counted. The displacement figures in the same section are attributed to B’Tselem data.
The legal qualification was written in the register's own voice and is in neither edition. What the report says about outposts is carried instead, with its place in the document.
source_reserves
“Outpost” renders מאחז (ma’ahaz), the word the source uses in Hebrew: a settlement established without Israeli government authorisation, and therefore unlawful under Israeli law itself. “Settlement” does not carry that. (Another reserve read: “An outpost is by definition established without Israeli government authorisation; …”) → “Outpost” renders מאחז (ma’ahaz), the word the source uses in Hebrew, and “settlement” renders התנחלות. The source’s table of new jurisdiction areas labels some entries “Legalization of an outpost”, and its text speaks of “the regularization of existing outposts”. The sentence on authorisation is removed from the second reserve.
A statement about the publisher's standing, which the document does not contain, and a forecast about readers are removed. The facts that remain are stated with the document as subject.
data_provenance ; registry_note
data_provenance: Peace Now is an Israeli organisation whose Settlement Watch has been the reference on this subject for decades, and which counts the policy of its own government. registry_note: The number most likely to be lifted from this report is 60, because it sits in the English summary. → data_provenance: The count of outposts is Peace Now’s own; the page names no source for it. The displacement figures in the same section are attributed to B’Tselem data. registry_note: The English edition’s “Key Data” list gives 60 “agricultural farms”. The body of the English edition and the Hebrew edition give 58, and the Hebrew edition does not print 60.
The two sentences that follow the one quoted state the authors’ own consequences of the verification asymmetry, for Palestinian injuries and for the Israeli count. They, the definition of an injury and the exact scope of the 7 October 2023 exclusion were absent.
source_reserves
“The database does not record the two sides the same way. The authors state it in their limitations: ‘Palestinian casualties are only recorded if there is two-source verification, whereas this is not the case for Israeli casualties.’ In the methods the exception is put on Israeli injuries …” → … The authors add that this “may lead to underestimation of Palestinian injury rates” and that, since two-source verification is only used for Palestinian casualties, “there is a risk of misclassification bias for Israeli casualties, including those that may be Palestinian citizens of Israel”. … Also added: those injured “are recorded only if they were physically hurt in a relevant incident and received medical treatment at a clinic or hospital, or by paramedic personnel”, and “minor injuries are not recorded within the database leading to underreporting”; the scope of the 7 October 2023 exclusion (“data on casualties in the West Bank and Israel in other contexts is updated beyond 7 October 2023”).
The word “criteria” is not in the article, and it restated, in another form, the claim that the study applied one set of criteria to both sides, which the correction of 20 September had already removed. The casualties are recorded in the UN humanitarian office’s database; the study takes them from it after excluding Gaza.
claim
… records 69,312 Palestinian casualties meeting the study criteria — … → … covers 69,312 Palestinian casualties after excluding those that occurred in Gaza — …
“It is not the Gaza war” was ambiguous, since the period contains the 2014 Gaza war, whose Israeli casualties are in the database. The two instructions to the reader, and the sentence saying that a single origin allows the two counts to be set side by side, were the register's. The department named in the author line is the first author’s only.
source_reserves ; data_provenance ; organization
Reserves: “… it is not the Gaza war”; “… they must never be merged into a single death toll” / “… the two are never merged.” data_provenance: “… That single origin is what allows the two to sit side by side; …” organization: “… Traboulsi et al. (Department of Emergency Medicine, Yale University, New Haven, Connecticut)” → Reserves: “Casualties that occurred in Gaza (n = 43,584) were excluded from the analysis.”; “The study reports the two separately.” data_provenance: “… the UN humanitarian office’s Protection of Civilians database, into which, the article says, its field staff regularly enter the data; for incidents resulting in Israeli injuries, the information ‘is based on media reports’. The database does not apply the same verification rule to the two sides.” organization: “(first author: Department of Emergency Medicine, Yale University, New Haven, Connecticut)”
The two sentences that follow the one quoted state the authors’ own consequences of the verification asymmetry, for Palestinian injuries and for the Israeli count. They, the definition of an injury and the exact scope of the 7 October 2023 exclusion were absent.
source_reserves
“The database does not record the two sides the same way. The authors state it in their limitations: ‘Palestinian casualties are only recorded if there is two-source verification, whereas this is not the case for Israeli casualties.’ In the methods the exception is put on Israeli injuries …” → … The authors add that this “may lead to underestimation of Palestinian injury rates” and that, since two-source verification is only used for Palestinian casualties, “there is a risk of misclassification bias for Israeli casualties, including those that may be Palestinian citizens of Israel”. … Also added: those injured “are recorded only if they were physically hurt in a relevant incident and received medical treatment at a clinic or hospital, or by paramedic personnel”, and “minor injuries are not recorded within the database leading to underreporting”; the scope of the 7 October 2023 exclusion (“data on casualties in the West Bank and Israel in other contexts is updated beyond 7 October 2023”).
The word “criteria” is not in the article, and it restated, in another form, the claim that the study applied one set of criteria to both sides, which the correction of 20 September had already removed. The casualties are recorded in the UN humanitarian office’s database; the study takes them from it after excluding Gaza.
claim
… records 4,118 Israeli casualties meeting its criteria — … → … covers 4,118 Israeli casualties after excluding those that occurred in Gaza — …
From 1 May 2014 to 30 June 2024 is ten years and two months, not eleven; eleven calendar years are touched.
method
Same dataset, same eleven-year period and same analysis as entry westbank-2025-001: … 1 May 2014 to 30 June 2024. → Same dataset, same period and same analysis as entry westbank-2025-001: … 1 May 2014 to 30 June 2024.
“It is not the Gaza war” was ambiguous, since the period contains the 2014 Gaza war, whose Israeli casualties are in the database. The two instructions to the reader, and the sentence saying that a single origin allows the two counts to be set side by side, were the register's. The department named in the author line is the first author’s only.
source_reserves ; data_provenance ; organization
Reserves: “… it is not the Gaza war”; “… they must never be merged into a single death toll” / “… the two are never merged.” data_provenance: “… That single origin is what allows the two to sit side by side; …” organization: “… Traboulsi et al. (Department of Emergency Medicine, Yale University, New Haven, Connecticut)” → Reserves: “Casualties that occurred in Gaza (n = 43,584) were excluded from the analysis.”; “The study reports the two separately.” data_provenance: “… the UN humanitarian office’s Protection of Civilians database, into which, the article says, its field staff regularly enter the data; for incidents resulting in Israeli injuries, the information ‘is based on media reports’. The database does not apply the same verification rule to the two sides.” organization: “(first author: Department of Emergency Medicine, Yale University, New Haven, Connecticut)”
The report gives some figures without naming a source (for instance 216 attacks on health care, marked “WHO SSA”), so “every figure is attributed” was wrong, and what WHO does or does not establish is not something the report says. It says who the source of each toll is.
source_reserves ; method
Reserve: “Every figure in this report is attributed, and WHO establishes none of them itself. They are the counts of two parties, relayed by a UN agency, and the agency says so.” method: “… WHO establishes neither count.” → Reserve: “The dashboard note on the first page reads ‘Source for Gaza and West Bank death and injury figures: Ministry of Health’, and the Israeli sentence opens ‘According to Israeli authorities’. The report does not give either set of figures as a WHO count.” method: “… The report does not give either count as its own.”
The report does not date the Israeli figure, so “two days apart at most” could not be supported; the same series, WHO’s appeal of October 2023, gave the same orders of magnitude “as of 26 October”. The locator of the reserve also named the wrong sentence: the Israeli sentence is the fourth of the overview, not the third.
source_reserves
… The two figures are two days apart at most, and the report does not place them in the same window. → … The report gives no date for the Israeli count, so it does not place the two figures in the same window.
Both versions were read end to end on 20 September 2026. Neither states that the count is compiled from hospitals and morgues; that description came from a different document about a different period.
method
Running count compiled by the Ministry of Health from the Gaza Strip's hospitals and morgues, announced at a daily press briefing. It counts the dead recorded by the health system. → A running count announced at a daily press briefing. Neither version of this briefing states how the count is compiled, nor what it excludes.
Both versions were read end to end on 20 September 2026. The word "crimes" is in neither of them.
source_reserves
Its statements of that day count "massacres" and name the Israeli action "aggression" and "crimes". → Its Arabic statement of that day uses massacre and massacres, the aggression, the occupation, the violations and ethnic cleansing. Its English statement of the same day uses "massacres" and "violations", and none of the others.
Both versions were read end to end, side by side, on 20 September 2026. The figures of 4,137 killed and 13,162 injured are identical in both.
source_reserves
The entry presented the Ministry's English statement as its rendering of the briefing, and said nothing about how the Arabic differs. → The Ministry published this briefing in Arabic and in English and the two texts are not the same. Where the Arabic reads 37 massacres against families of the Gaza Strip, the English reads "This includes 37 families". Where the Arabic reads massacres against 521 families, 3,109 shahid among them, the English reads "resulted in the killings of 521 families, including 3,109 citizens". And the Arabic carries an item on ethnic cleansing that the English does not carry at all.
Page 102 and printed page 851 differ by 749, not by the 750 the entry stated. The two page pairs come from the document; the offset was derived from them. The Court refuses this registry's robot, so the document could not be re-opened to settle which number was the slip, and the derived number is removed rather than guessed.
sourcing.locator
the file's page 102 carries the printed page 851 — an offset of 750, established from the running heads across the document and not assumed → The quoted operative paragraph is on printed page 851, which is page 102 of the PDF; the excluded-conduct paragraph is on printed page 786, page 37 of the PDF. Both page pairs were taken from the document itself.
Three statements the entry made about itself did not hold. None of them needed the document to check.
source_reserves, sourcing.document
"and this is the reserve most often dropped"; a reserve reading "This is an advisory opinion, which is what the document calls itself" under the locator "not in this document"; and "No request was made to the Court" beside "downloaded (...) at the address given above". → The frequency claim is removed and the fact kept. The self-contradicting reserve is removed; the method note already states the General Assembly requested an advisory opinion. The provenance now reads that the robot sent no request and a person opened the page by hand.
The correction was read on 19 September 2026, and the banner still said its content had not been read. The banner itself was stating something untrue.
corrections[0].notice
A correction has been published on this article (The Lancet 2025; 405(10477): 468). Its content has not been read here. No figure from this study should be cited without noting that a correction exists. → A correction has been published on this article (The Lancet 2025; 405(10477): 468). It corrects the appendix, for labelling inconsistencies. The sentences cited here are in the Summary, the Results and the Discussion.
The article was read end to end on 20 September 2026. The supporting quotations announced sentences other than the one quoted and carried that same sentence.
supporting_quotes
A single item repeating the main quotation word for word, under the locator "Same document, sentences other than the one quoted above". → Five distinct sentences from the article, each with its own page and section, and the gaps between them stated: the Ministry of Health total of 37,877 (p. 1), the 34,995 deaths estimated outside any list (p. 6), the 45.4% list coverage (p. 6), the BMA model's 50,663 (p. 6), and the authors' minimum-estimate conclusion (p. 8).
The article was read end to end on 20 September 2026. These four reserves are stated in the article and were absent from the entry. The figure of 64,260 (55,298-78,525) is unchanged.
source_reserves
Three reserves: traumatic injury only, 930 missing people excluded, and the method infers unlisted deaths. → Four added. The 41% is measured against the Ministry of Health's own 37,877, a total that already includes 9,692 unidentified decedents. Of the three alternative estimators the authors report, the Bayesian model averaging one gives 50,663, below the lower bound of the published interval. The confidence interval is a Wald interval, and the authors write that better-coverage methods were computationally prohibitive. Hospitals in Gaza City and North Gaza stopped reporting on 2 November 2023, and the authors could not stratify by geography.
The report was read end to end on 20 September 2026. It is written in English and says "thousands of terrorists". The claim is attributed to the State Comptroller from its first words, so the word carried there is the report's own.
claim
attacked Israel with thousands of missiles and the infiltration of thousands of fighters → attacked Israel with thousands of missiles and the infiltration of thousands of terrorists
The report was read end to end on 20 September 2026 and the offset established from the spread numbers, the alternating running heads, and the text running on across each page break.
locator
Abstract, paragraphs beginning "On October 7th, 2023, during the holiday of Simchat Torah..." — with a note that the printed page number could not be established with certainty. → Abstract, printed p. 62. In this file the printed page is the file page plus 60.
Both editions were read end to end, side by side, on 20 September 2026. The earlier reserve was written from the English edition alone.
source_reserves
The two figures are on one page, in one language, in one document. → Only one of the report's two language versions carries the second figure. The English edition opens with a "Key Data for 2025" list reading "86 new outposts, including 60 'agricultural farms'", where its body reads 58. The Hebrew edition has no such summary list, and the figure 60 does not appear anywhere in it.
The article was read end to end on 20 September 2026. It states this exclusion three times, in the methods, the results and the limitations.
source_reserves
The entry gave the casualty counts without recording that the database excludes the hostilities that began on 7 October 2023. → Casualties from the hostilities that began on 7 October 2023 are not in this database. The authors write that they were "awaiting independent verification" at the time of analysis, and that their absence "may lead to underestimation of both Palestinian and Israeli death and injury rates."
The article was read end to end on 20 September 2026. Its methods state: "The analysis covers the period from 1 May 2014 to 30 June 2024."
period_covered
1 January 2014 to 31 December 2024 → 1 May 2014 to 30 June 2024
The article was read end to end on 20 September 2026. The entry asserted a symmetry of protocol that the source states twice it does not have.
method
Applying a single set of inclusion criteria to Palestinian and Israeli casualties alike. The entry also carried the tag same-protocol. → The two sides do not enter the database under the same verification rule. The authors state: "Palestinian casualties are only recorded if there is two-source verification, whereas this is not the case for Israeli casualties."
The article was read end to end on 20 September 2026. It states this exclusion three times, in the methods, the results and the limitations.
source_reserves
The entry gave the casualty counts without recording that the database excludes the hostilities that began on 7 October 2023. → Casualties from the hostilities that began on 7 October 2023 are not in this database. The authors write that they were "awaiting independent verification" at the time of analysis, and that their absence "may lead to underestimation of both Palestinian and Israeli death and injury rates."
The article was read end to end on 20 September 2026. Its methods state: "The analysis covers the period from 1 May 2014 to 30 June 2024."
period_covered
1 January 2014 to 31 December 2024 → 1 May 2014 to 30 June 2024
The article was read end to end on 20 September 2026. The entry asserted a symmetry of protocol that the source states twice it does not have.
method
Applying a single set of inclusion criteria to Palestinian and Israeli casualties alike. The entry also carried the tag same-protocol. → The two sides do not enter the database under the same verification rule. The authors state: "Palestinian casualties are only recorded if there is two-source verification, whereas this is not the case for Israeli casualties."
The report was read end to end on 20 September 2026. Its Gaza toll is bounded at 30 October and its Israeli toll is not, so the two figures in it cover windows that differ by up to two days.
source_reserves
The entry carried the two tolls from this report without recording that the report gives an end date for one and not the other. → The two tolls do not carry the same end date. The Gaza toll is given "From 7 to 30 October"; the Israeli toll is given "since the hostilities started", with no end date, in a report whose stated cutoff is 1 November 2023, 13:00.
The report was read end to end on 20 September 2026. The three percentages were listed side by side, which reads as a partition; the source's own counts show the children fall inside the female and male figures.
method
For Gaza it cites the Ministry of Health (approximately 41.5% children, 45% female, 55% male of those killed). → The Gaza breakdown does not describe three separate groups. The source gives approximately 41.5% children (3,650), 45% female (3,956) and 55% male (4,849) of 8,805 killed. Its own female and male counts add to the 8,805 total; the children are counted within them.