Records · Verification table
Child hunger Do page: verification of what changed after the 5 November fresh read, 6 November 2026
The working record behind the pages, published as written. Source file source/verify/child-hunger-do-2026-11-06.md. All records.
Scope. Every claim in a line added or changed in content/child-hunger-do.html between ee6f7b8 and HEAD (4764ebf for this file), from git diff ee6f7b8 HEAD -- content/child-hunger-do.html, that has no row in source/verify/child-hunger-2026-11-05.md. Every author name, title, year, venue, sample size and effect in the source list written on 5 or 6 November was treated as unverified until read on the paper's own record. The help-now box, "What happens after" and the languages line were checked against the quotes in source/impact/needs-2026-11-06.md (read 08:55 to 09:21 EST today); five of those sources were also re-read. Mid-run, the coordinator passed on five findings from the Learn and Practice re-verification that recur on this page; each was checked from the source (rows H7, H8 to H12, A1, F2, S1a, S16a, S16b, X1). Only content/child-hunger-do.html was edited.
Method. Sources were fetched on 2026-11-06 between 09:35 and 09:58 EST (14:35 to 14:58 UTC; date stamps 09:35, 09:57, 09:59). Default: curl with a browser user agent, HTML stripped to text, PDFs read with pdftotext -layout, JSON and XML APIs (NCBI E-utilities, Crossref, OpenAlex, Europe PMC, ClinicalTrials.gov) read as returned. Pages that render with JavaScript or refuse curl were loaded in headless Chrome ([local working file], which returns document.body.innerText). Where curl got 403, WebFetch was used: for PDFs, only to save the file, which was then read with pdftotext (GAO-09-570, GAO-15-666, the Christian and Barrett working paper); for FRAC's page, WebFetch's quotes are the tool's copy and were not checked byte for byte (marked in the row). No verdict rests on a search snippet or on memory. One web search was run, to look for an IRC statement of its staff numbers; it found none, and the row cites only what was read.
Verdicts. CONFIRMED: the source's words support the page as written. CORRECTED: the page was changed to match the source. NOT IN SOURCE: the cited source does not say it; reworded or removed. NEWER DATA: a newer reading from the same evaluator changes the claim. BLOCKED: 403/405 or a bot wall. UNREACHABLE, UNFETCHABLE: none this run.
Grades (the site's own scale, content/about.html): primary = "the original document"; research = "peer-reviewed work"; evaluator; charity's own figures = "an organisation reporting on itself"; news; our arithmetic; internal = this repo's own files.
Fetch log (all 2026-11-06).
- NCBI E-utilities abstracts, PMIDs 30398672, 22336792, 35294044, 23498849, 35090531, 35303461, 32589867, 32589873, 23746772, 30834852: 200 (one 429 rate limit, retried, 200).
- Europe PMC full text PMC7311356 (Goldman et al.): 200. Europe PMC search for PMID 35303461: no open full text.
- ClinicalTrials.gov API, NCT03751475 (OptiMA-DRC): 200.
- PLOS Medicine, ComPAS article page: 200.
- GiveWell: top charities, cost-effectiveness criteria, Our mistakes, node/4024 (→ IRC January 2024 grant page), node/4359 (→ Taimaka grant page), ALIMA Chad February 2024 grant page, ALIMA Cameroon June 2025 grant page, Top Charities Fund, sitemap.xml, Metrics Report 2025 PDF: all 200.
- Gates Foundation, Goalkeepers 2025 PDF: 200.
- ALIMA financial transparency, 2025 annual report page, donate page: 200.
- Taimaka donate: 200. Every.org checkout (
every.org/taimaka): curl 429; headless Chrome "Vercel Security Checkpoint … Failed to verify your browser" (BLOCKED). PayPal hosted donate button: 403 (BLOCKED). - Charity Navigator, EIN 843964208: 200. ProPublica Nonprofit Explorer, EINs 843964208 and 133327220: 200.
- Action Against Hunger: Lifeline page 200 (curl); donate page "Client Challenge" to curl, read in headless Chrome (it redirects to
/take-action/donate/); 2025 audited financial statements PDF 200. - Semafor, 22 September 2026 (IRC staffing): 200. The New Humanitarian, 19 February 2025: 403 (not needed; Semafor carries the figures).
- Feeding America: home, "about our claims" FAQ, 2025 annual report PDF, FY2026 audited statements PDF: 200. Donate page (give.feedingamerica.org): 200, no meal line in its static HTML.
- NBER w24652: 200. USDA FNS "Measuring the Effect of SNAP Participation on Food Security" PDF: 200. USDA WIC 2021 estimates page (→ fna.usda.gov): 200. USDA SEBTC summary report PDF: 200. USDA Summer EBT participating page and New York page (→ fna.usda.gov): 200. FRAC
frac.org/summer-ebt: curl 403, headless Chrome 403, WebFetch 200 (quotes are the tool's copy). - GAO-09-570 and GAO-15-666: product pages and PDFs 403 to curl; product page read through WebFetch; PDFs saved by WebFetch and read with
pdftotext. - AEA, Nunn and Qian article page: 200. Crossref, 10.1093/ej/uead091 and 10.1007/s11109-014-9277-1: 200. OpenAlex records for both: 200. OUP article page: 403. EconStor copy of the Christian and Barrett working paper (I4R DP 1): curl got a challenge page; saved by WebFetch, read with
pdftotext. Springer, Bergan and Cole: abstract read in headless Chrome. Yale ISPS, Butler and Nickerson: 200 (old URL redirects to/resource/can-learning-constituency-opinion-affect-how-legislators-vote-results-from-a-field). - ReThink Orphanages (→ bettercarenetwork.org/rethink-orphanages): 200. 211.org about-us, mental-health, home, food, and five other 211.org pages: 200. United Way, "The facts about 211" page: 200. IRC simplified protocol research summary PDF: 200.
- Repo files read:
content/child-hunger-do.html(and its diff),source/verify/child-hunger-2026-11-05.md,source/impact/needs-2026-11-06.md,source/impact/child-hunger-2026-11-05.md,source/verify/child-hunger-2026-11-06-additions.md(the sibling Learn/Practice table),content/about.html,content/child-hunger-learn.html,content/child-hunger-practice.html,source/local/copy.json,tools/app.js,vercel.json,source/stale.tsv.
Working files (raw downloads and extracted text) are in the session scratchpad, [local working file]. They are not kept in the repo.
---
A. Description, lede, help-now box, "Near you"
| # | Page said | Source | Fetch | Source says (verbatim) | Verdict | Grade | Fix applied |
|---|---|---|---|---|---|---|---|
| A1 | Description: "and the one US action with trial evidence" | https://www.nber.org/papers/w24652; the page's own Summer EBT item | OK (curl) | "a randomized field experiment in which 30,000 elderly individuals not enrolled in – but likely eligible for – the Supplemental Nutrition Assistance Program (SNAP) …" The paper does not say it is the only US action with trial evidence, and the same page cites a randomised evaluation of Summer EBT (F10). | NOT IN SOURCE | research | "the one US action with trial evidence" → "a US action tested in a randomised trial" |
| A2 | Lede: step one's programs "prevent the illnesses that kill malnourished children" | https://www.givewell.org/charities/top-charities; UNICEF/UN IGME release (5 Nov row L7) | OK (curl) | "Medicine to prevent malaria"; "Nets to prevent malaria"; "Vitamin A deficiency leaves children vulnerable to infections and can lead to death"; vaccines "prevent disease and reduce child mortality". 5 Nov L7: "malnutrition weakens children's immunity and increases their risk of dying from common childhood diseases." | CONFIRMED (a synthesis of the two) | evaluator / primary | none |
| A3 | Help-now: "dial 211, free and at any hour" | https://211.org/about-us; https://www.unitedway.org/the-facts-about-211-2025-impact-snapshot; needs file A1.2, A1.3 | OK (curl) | 211.org: "Thousands of caring, local experts are available to help, 24/7. Calls to 211 are confidential and can be anonymous." United Way: "211 is the federally designated 3-digit dialing code that connects to a free, confidential service". 211.org's footer: "© Copyright 2026 United Way Worldwide." | CONFIRMED (the "free" is United Way's page, not the About page s17 cited; see S17) | primary / charity's own figures | none here; s17 fixed |
| A4 | Help-now: "find a food bank by ZIP" | needs file A3.1 (https://www.feedingamerica.org/find-your-local-foodbank) | needs file | "Enter your ZIP code to find nearby food banks in the Feeding America network" | CONFIRMED | primary | none |
| A5 | Help-now: "Apply for SNAP through your state: you hear back within 30 days" | needs file A3.3, A3.4 (https://www.fna.usda.gov/snap/recipient/eligibility) | needs file | "a member of your household must contact your state agency directly to apply"; "will process it and send you a notice telling you whether or not you are eligible for benefits within 30 days." | CONFIRMED | primary | none |
| A6 | Help-now: "or within 7 if your household has almost no money" | needs file A3.4 | needs file | "You may be eligible to receive SNAP benefits within 7 days of your application date if you meet additional requirements. For example, if your household has less than $100 in liquid resources and $150 in monthly gross income …" | CORRECTED (the source says "may", and it is benefits, not a reply) | primary | → "and may get benefits within 7 if your household has almost no money"; s17 "within 7" → "benefits possibly within 7" |
| A7 | Help-now: "Enter your ZIP for your state's application" | source/local/copy.json, child-hunger | read | "text": "If your family needs food: dial 211, {foodbank:find a food bank near you}, or {snap:apply for SNAP in [state]}. …" | CONFIRMED | internal | none (copy.json has the same 7-day wording: proposal 1) |
| A8 | "Near you": the ZIP adds "your state's own programs and laws, and the finders that take a ZIP. The code stays in your browser" | tools/app.js; vercel.json; source/local/copy.json | read | vercel.json: "connect-src 'none'". app.js has no fetch(, XMLHttpRequest, sendBeacon or WebSocket. copy.json lists per-state items (Summer EBT, SNAP) and finders. | CONFIRMED | internal | none |
B. The three steps
| # | Page said | Source | Fetch | Source says (verbatim) | Verdict | Grade | Fix applied |
|---|---|---|---|---|---|---|---|
| B1 | Step 1 evidence label strong | GiveWell top charities; Cochrane abstracts (B2 to B4) | OK | "Exceptionally strong. Many high-quality studies of nets have consistently found strong impact." (also SMC, New Incentives); vitamin A: "Very strong. Many high-quality studies of vitamin A supplementation show strong impact, though results vary by study." Nets and vitamin A: "high-certainty evidence". | CONFIRMED | evaluator / research | none |
| B2 | "Bed nets cut child deaths 17% across trials" | Pryce et al., E-utilities PMID 30398672 | OK (curl) | "Insecticide-treated nets reduce child mortality from all causes by 17% compared to no nets (rate ratio 0.83, 95% CI 0.77 to 0.89; 5 trials, 200,833 participants, high-certainty evidence)." "The included studies were conducted between 1987 and 2001." | CONFIRMED | research | none |
| B3 | "seasonal malaria medicine prevents about three-quarters of cases" | Meremikwu et al., PMID 22336792 | OK (curl) | "IPTc prevents approximately three quarters of all clinical malaria episodes (rate ratio 0.26; 95% CI 0.17 to 0.38; 9321 participants, six trials, high quality evidence)". Mortality: "the trials were underpowered to reach statistical significance (risk ratio 0.66, 95% CI 0.31 to 1.39". | CONFIRMED | research | none (s4 now notes the effect) |
| B4 | "vitamin A cut deaths 12% across trials" | Imdad et al., PMID 35294044 | OK (curl) | "A meta-analysis for all-cause mortality included 19 trials (1,202,382 children). At longest follow-up, there was a 12% observed reduction in the risk of all-cause mortality … (risk ratio (RR) 0.88, 95% confidence interval (CI) 0.83 to 0.93; high-certainty evidence)." | CONFIRMED | research | none |
| B5 | "though the largest single trial found nothing" | Awasthi et al. (DEVTA), PMID 23498849 | OK (curl) | "mortality ratio 0·96, 95% CI 0·89-1·03, p=0·22"; "DEVTA contradicts the expectation from other trials that vitamin A supplementation would reduce child mortality by 20-30%, but cannot rule out some more modest effect." | CORRECTED ("nothing" overstates a non-significant 4% estimate the authors say cannot rule out a modest effect) | research | → "found no clear effect" |
| B6 | Step 2: "a child's course of peanut paste and a nurse" | IRC research summary PDF; needs file C4.3 (Taimaka) | OK (curl) | IRC: "treatment by CHWs resulted in similar (94%) recovery as treatment by formal health care workers." Taimaka: "These CHWs provide therapeutic foods, medicines, and vaccinations". The page's own lede says "a health worker". | CORRECTED (precision; community health workers, not nurses, do most of it) | charity's own figures | "a nurse" → "a health worker" |
| B7 | Step 2 evidence label moderate | B8, B9 | OK | Trials for recovery; observational data for deaths (B8, B9). | CONFIRMED | research / evaluator | none |
| B8 | "The treatment protocols are trial-tested for recovery" | ComPAS (PLOS Medicine page); OptiMA-DRC, PMID 35303461 | OK (curl) | ComPAS: "Recovery in the combined protocol arm was non-inferior to that in the standard protocol arm, with 76.3% recovering in the combined protocol arm and 73.5% recovering in the standard protocol arm". OptiMA-DRC: "an open-label, non-inferiority, randomised controlled trial". | CONFIRMED | research | none |
| B9 | "the claim that treatment prevents death rests on older observational data" | GiveWell IRC and Taimaka grant pages | OK (curl) | "We do not have any direct evidence of the mortality rates of untreated children with malnutrition. … We instead rely on historical observational studies to inform our estimate." "based on several observational studies conducted in the 1980s and 1990s." | CONFIRMED | evaluator | none |
| B10 | "GiveWell rates these programs a little below its top charities" | GiveWell IRC and Taimaka pages | OK (curl) | IRC: "around eight times as cost-effective as unconditional cash transfers ("8x cash") … This is close to 10x cash, our cost-effectiveness bar". Taimaka: "approximately nine times as cost-effective as unconditional cash transfers. At the time we recommended this grant, our cost-effectiveness bar for recommending grants was 10x cash." | CONFIRMED | evaluator | none |
| B11 | "Per life saved, the two steps cost about the same." | GiveWell top charities; IRC and Taimaka pages | OK (curl) | Top charities: "$3,500 per life saved" (vitamin A) to "$5,500 per life saved" (nets). Taimaka: "$4,787,985" grant, "Total number of deaths averted among malnourished children 1,128". IRC: "Cost per under-five death averted due to malnutrition treatment $5,487". | CONFIRMED (our arithmetic: $4,245 and $5,487 sit inside $3,500 to $5,500) | our arithmetic | none (H6 fixes the Q&A sentence that said otherwise) |
| B12 | Step 3 evidence label weak: moves some votes | E1 sources; CMF 2017 (5 Nov D30) | OK | Bergan and Cole: "being targeted by constituent contacts increases the probability of supporting the relevant legislation by about 12 percentage points." Butler and Nickerson: "much more likely to vote in line with constituent opinion". CMF: staff opinion survey. | CONFIRMED | research / primary | none |
C. Where to give
| # | Page said | Source | Fetch | Source says (verbatim) | Verdict | Grade | Fix applied |
|---|---|---|---|---|---|---|---|
| C1 | "GiveWell published its own modelling corrections in March 2026" | https://www.givewell.org/about/our-mistakes | OK (curl) | "Last updated: March 2026"; "How we addressed our mistake (updated March 2026): … we've adjusted our vitamin A supplementation model to account for its overlap with azithromycin distribution." | CONFIRMED | evaluator | sentence split for the audit's length limit |
| C2 | "(its vitamin A estimate had been about a fifth too optimistic)" | as C1 | OK (curl) | "This oversight could be leading us to overestimate VAS cost-effectiveness by approximately 20%." | CORRECTED (the source hedges: "could be") | evaluator | → "its vitamin A estimate may have been about a fifth too high" |
| C3 | "the biggest vitamin A trial, a million children in India, found no effect on deaths while the pooled trials do" | DEVTA PMID 23498849; Imdad PMID 35294044 | OK (curl) | DEVTA: "under-5 population 1 million"; "mortality ratio 0·96, 95% CI 0·89-1·03"; "cannot rule out some more modest effect". Imdad: "12% observed reduction". | CORRECTED | research | → "found no clear effect on deaths, while the pooled trials found one" |
| C4 | ALIMA "runs its own treatment trials" (and Q4 "ALIMA runs the trials") | ClinicalTrials.gov NCT03751475; OptiMA-DRC PMID 35303461 | OK (API, curl) | "leadSponsor": {"name": "Alliance for International Medical Action" for "OptiMA-DRC". Authors include "(2)The Alliance for International Medical Action (ALIMA), Paris, France." | CONFIRMED | primary / research | none |
| C5 | ALIMA "publishes audited accounts" | https://alima.ngo/en/about-us/financial-transparency/ | OK (curl) | "Our accounts are audited annually by an independent Statutory Auditor"; link "Certified accounts 2025" → ASSOCIATION-ALIMA-Comptes-annuels-2025-anglais.pdf. | CONFIRMED | charity's own figures | none |
| C6 | ALIMA: "GiveWell funds its malnutrition work" | GiveWell Metrics Report 2025; GiveWell Cameroon June 2025 page | OK (curl) | "ALIMA USA — Malnutrition Treatment in Mirriah and Dakoro, Niger and Ngouri, Chad (October 2025) $4,711,130"; "ALIMA USA — Primary Healthcare and Malnutrition Treatment in Far North Cameroon (June 2025) $1,900,000". | CONFIRMED | evaluator | s5 now names these GiveWell pages |
| C7 | ALIMA: "and has cut its estimate for one country" (cited to s5 and s7, neither of which covered ALIMA's GiveWell pages) | https://www.givewell.org/research/grants/ALIMA-malnutrition-Ngouri-NDjamena-February-2024; Metrics Report 2025 | OK (curl) | Feb 2024 page (published July 2024): "At the time we recommended the grant, we estimated that the program's cost-effectiveness was below our funding threshold of ten times as cost-effective as unconditional cash transfers". Metrics Report: the October 2025 grant above includes "Ngouri, Chad". | NEWER DATA (the cut is dated 2024; GiveWell funded the same Chad site again in October 2025; the cited sources did not cover it) | evaluator | → "GiveWell funds its malnutrition work: in 2024 it rated the Chad program below its bar, then funded Chad again in 2025."; both GiveWell documents named in s5 (no link added; see proposal 4) |
| C8 | Taimaka "$94 to $132 per child depending on which of its pages you read" | https://taimaka.org/donate; needs file C4.3 | OK (curl) | Donate: "Every $94 in our program treats a child with acute malnutrition, a life-threatening condition." Home (needs file): "$132 Cost-Per-Child". | CONFIRMED | charity's own figures | none |
| C9 | Taimaka "One-time gifts only" | https://taimaka.org/donate | OK (curl) | Options listed: "Donate with credit or debit card", "Donate with bank transfer, stocks, etc.", "Donate via Paypal"; no "monthly" or "recurring" on the page. The card and PayPal checkouts are third parties (C9b, C9c) that could not be read. | NOT IN SOURCE (the page shows no monthly option; it does not say gifts are one-time only) | charity's own figures | → "No monthly option on its page" |
| C9b | (sub-check) Every.org card checkout | https://www.every.org/taimaka#/donate/card | curl 429; headless Chrome | "Vercel Security Checkpoint … Failed to verify your browser" | BLOCKED | — | none |
| C9c | (sub-check) PayPal donate button | https://www.paypal.com/donate/?hosted_button_id=FMPFBD55QX4J8 | curl | HTTP 403 | BLOCKED | — | none |
| C10 | Taimaka "three stars from Charity Navigator" | https://www.charitynavigator.org/ein/843964208 | OK (curl) | "Taimaka Project is a Charitable Organization that has earned a 3/4 Star rating on Charity Navigator."; "Three -Star Charity"; "Financial Statements : No" | CONFIRMED | evaluator | none |
| C11 | Taimaka: "GiveWell's estimate puts it just under its own funding bar" | GiveWell Taimaka page | OK (curl) | "We estimate the cost-effectiveness of Taimaka's program to be close to our funding bar (~9x)"; "our current best guess is that this program is below our funding bar" | CONFIRMED | evaluator | none |
| C12 | Taimaka: "with a renewal decision pending"; footer: "GiveWell's renewal decision on Taimaka was pending when we checked" | GiveWell Taimaka page; https://www.givewell.org/sitemap.xml | OK (curl) | "We expect to make a decision on whether to renew our support for Taimaka's program in 2026." The sitemap's only Taimaka page is the June 2025 grant page (lastmod 2025-06-26); no 2026 decision page. | CONFIRMED (as of 6 Nov 2026) | evaluator | none |
| C13 | IRC: "the organisation cut a third of its staff after the 2025 aid cuts" | https://www.semafor.com/article/09/22/2026/world-has-entered-age-of-impunity-amid-global-aid-cuts-miliband-says | OK (curl) | "The fallout from US President Donald Trump's aid cuts have led the IRC itself to shrink dramatically in the past year and a half, from 18,000 employees to 12,000". | CONFIRMED (12,000 is two-thirds of 18,000); the attribution in s7 was wrong (S7d) | news | none here |
| C14 | Action Against Hunger: "it came off because the gift is unrestricted" | https://www.actionagainsthunger.org/take-action/give-monthly-lifeline/; donate page | OK (curl; donate page in headless Chrome) | Lifeline: "As a Lifeline Sustainer, you support Action Against Hunger's lifesaving, rapid-response programs in over 50 countries." Neither page uses "restrict" or "unrestricted" (also 5 Nov row D18). | NOT IN SOURCE | charity's own figures | → "the gift supports its "programs in over 50 countries"" |
| C15 | AAH: "the page reports no outcomes" | Lifeline page | OK (curl) | The only figures on the page are gift amounts ("$15 $0.50/DAY" … "$1000"), "Your first three monthly gifts are matched 2X" and "more than 50 countries". | CONFIRMED | charity's own figures | none |
| C16 | AAH: "its "90 cents of every dollar" line" | https://www.actionagainsthunger.org/take-action/donate/ (rendered) | headless Chrome | "90% of your donation goes to help those in need." | CORRECTED (the page's quotation marks held words AAH does not use) | charity's own figures | → its line that "90% of your donation goes to help those in need" |
| C17 | AAH: "doesn't match its audited 2025 accounts (82 to 87 cents)" | https://www.actionagainsthunger.org/app/uploads/2026/10/ACF-USA-2025-FS.pdf | OK (pdftotext) | Year ended 31 December 2025: "Total program services 106,477,760"; "Program Support 6,240,738"; "Total expenses 129,820,850". 2024: "Total program services 136,837,579"; "Program Support 7,775,508"; "Total expenses 160,034,763". | CONFIRMED (our arithmetic: 82.0%, or 86.8% with program support; 2024 was 85.5% and 90.4%) | primary / our arithmetic | "(82 to 87 cents)" → "(82% to 87%)"; s7 labels the arithmetic |
D. "What it costs, honestly"
| # | Page said | Source | Fetch | Source says (verbatim) | Verdict | Grade | Fix applied |
|---|---|---|---|---|---|---|---|
| D1 | "About $4,200 to $5,500 is the cost of one death prevented, by GiveWell's figures for Taimaka and the IRC" | GiveWell Taimaka and IRC pages | OK (curl) | Taimaka: "Grant size to charity $4,787,985"; "Total number of deaths averted among malnourished children 1,128". IRC: "Cost per under-five death averted due to malnutrition treatment $5,487" ("using annualized projections for a three-year grant renewal in Burkina Faso as an example"). | CONFIRMED (our arithmetic: 4,787,985 ÷ 1,128 = $4,245) | evaluator / our arithmetic | none |
| D2 | "it takes about 40 treatments at Taimaka's price, and about 25 at the IRC's, to prevent one death" | as D1 | OK (curl) | Taimaka: "Cost per child $105"; "Number of severely malnourished children treated 45,424"; "1,128" deaths averted. IRC: "Cost per malnourished child reached $215"; "Number of malnourished children reached 11,798"; "Total number of deaths averted among malnourished children 463". | CORRECTED (the numbers hold two ways: 45,424 ÷ 1,128 = 40.3 and 4,245 ÷ 105 = 40.4; 11,798 ÷ 463 = 25.5 and 5,487 ÷ 215 = 25.5. But "Taimaka's price" is GiveWell's $105, not either of Taimaka's own figures) | our arithmetic | → "about 40 children treated at Taimaka, and about 25 at the IRC" |
| D3 | "Most treated children would have survived anyway" | as D1 | OK (curl) | Taimaka: "Annual mortality rate from all causes among children 6-59 months with untreated malnutrition - best guess 4.3%". IRC: "8.2%". | CONFIRMED | evaluator | none |
E. The call
| # | Page said | Source | Fetch | Source says (verbatim) | Verdict | Grade | Fix applied |
|---|---|---|---|---|---|---|---|
| E1 | "two field experiments found constituent contact moves legislators on low-profile votes" | Bergan and Cole (Springer abstract); Butler and Nickerson (https://isps.yale.edu/research/publications/isps11-016) | headless Chrome; OK (curl) | Bergan and Cole: "a field experiment in which Michigan state legislators are randomly assigned to be contacted by their constituents about a specific bill or to a control group … being targeted by constituent contacts increases the probability of supporting the relevant legislation by about 12 percentage points." Butler and Nickerson: "We surveyed 10,690 New Mexicans about the Governor's spending proposals … District-specific survey results were then shared with a randomly selected half of the legislature. The legislators receiving their district-specific survey results were much more likely to vote in line with constituent opinion". | CORRECTED (only one experiment tested contact; the other gave legislators survey results; neither abstract says "low-profile") | research | → "In two field experiments, state legislators who heard from constituents, or were shown their district's views, voted more in line with them." |
| E2 | "Food for Peace ships American-grown food on American ships." | GAO-15-666 PDF; GAO-09-570 PDF | WebFetch-saved PDFs, pdftotext | GAO-15-666 (August 2015): "Under U.S. law, a minimum share of U.S. food aid must be shipped on U.S.-flag vessels. … from the original 50 percent to 75 percent in 1985, and back to 50 percent in 2012." GAO-09-570: "U.S. legal requirements to procure U.S.-grown agricultural commodities for food aid". | CORRECTED (the rule is at least half, not all) | primary | → "Food for Peace buys American-grown food and must ship at least half of it on American ships." |
| E3 | "Government auditors found that buying food near the crisis costs about half as much for grains" | GAO-09-570 PDF | WebFetch-saved PDF, pdftotext | "GAO found that local procurement in sub-Saharan Africa cost about 34 percent less than similar in-kind food aid purchased and shipped from the United States"; "the average cost of WFP's local procurements in sub-Saharan Africa and Asia was 34 percent and 29 percent lower"; Latin America "comparable". The "over 50 percent" for grains is Lentz, Passarelli and Barrett (World Development, 2013) in source/impact/child-hunger-2026-11-05.md, not GAO, and is not cited on the page. | CORRECTED | primary | → "buying food in the region cost about a third less in Africa and Asia" |
| E4 | "and arrives months faster" | GAO-09-570 PDF | as E3 | "in-kind international food aid delivery to 10 sub-Saharan African countries took an average of 147 days, while local procurement only took about 35 days and regional about 41 days." | CONFIRMED (tense matched to the 2009 finding: "arrived months sooner") | primary | tense only |
| E5 | "one study links in-kind US food aid to longer conflicts" | https://www.aeaweb.org/articles?id=10.1257/aer.104.6.1630 | OK (curl) | "an increase in U.S. food aid increases the incidence and duration of civil conflicts, but has no robust effect on inter-state conflicts or the onset of civil conflicts." | CONFIRMED | research | none |
| E6 | "a finding another study disputes" | Crossref 10.1093/ej/uead091; I4R DP 1 (EconStor PDF) | OK (API); WebFetch-saved PDF | Crossref abstract: "We demonstrate these concerns by revisiting recent studies of the causal origins of conflict." Working paper: "The most compelling cross-country studies, such as Nunn and Qian (2014, hereafter NQ), use a panel IV strategy"; "An earlier version circulated with the title "Revisiting the Effect of Food Aid on Conflict: A Methodological Caution."" | CONFIRMED (it disputes the method behind the finding; s9 now says so) | research | s9 wording (S9d) |
F. Closer to home
| # | Page said | Source | Fetch | Source says (verbatim) | Verdict | Grade | Fix applied |
|---|---|---|---|---|---|---|---|
| F1 | "What fixes that is income, SNAP, WIC and school meals" ("WIC" added after 5 Nov) | the page's sources s10 to s13 | read | No source on the page measures WIC's effect on food insecurity; s11's WIC source is coverage only ("WIC served an estimated 51.2 percent of those eligible"). None was read this run. | NOT IN SOURCE (for WIC) | — | "WIC" dropped from this sentence (it stays in the action, where only enrolment and reach are claimed) |
| F2 | "and the one action with trial evidence is helping a family sign up" | NBER w24652 | OK (curl) | As A1: a trial with "30,000 elderly individuals", not families; the paper does not say it is the only such action. | NOT IN SOURCE | research | → "Helping people apply for SNAP raised enrolment in a randomised trial." |
| F3 | Evidence label moderate: strong per hour | NBER w24652; source/impact/child-hunger-2026-11-05.md §2.1 item 3 | OK | Impact review: "Grade moderate-to-strong per hour. Information plus application help raised SNAP enrolment from 6% to 18% among 30,000 likely-eligible people in a randomised trial". | CONFIRMED | research / internal | none |
| F4 | "In a trial with 30,000 older adults, information plus help with the application tripled SNAP enrolment, from 6% to 18%." | NBER w24652 | OK (curl) | "Only 6 percent of the control group enrolls in SNAP over the next 9 months, compared to 11 percent of the Information Only group and 18 percent of the Information Plus Assistance group." | CONFIRMED | research | none |
| F5 | "SNAP cuts children's food insecurity by about a tenth of a point per family, across thousands of households" | USDA FNS, Measuring the Effect of SNAP Participation on Food Security (Mabli et al., August 2013) | OK (pdftotext) | "Participating in SNAP for about six months was associated with a decrease in the percentage of households with children in which children were food insecure by 8.6 percentage points in the cross-sectional sample, from 33.3 percent … to 24.8 percent … In the longitudinal sample … by 10.1 percentage points, from 32.3 percent … to 22.2 percent". "9,811 households interviewed in 30 States". "should not be considered indicative of the causal effects of SNAP." | CORRECTED (about 9 to 10 percentage points, not a tenth of a point; an association, not a cut) | primary | → "In a USDA study of 9,811 households, six months on SNAP went with about 9 to 10 points less food insecurity among children."; s11 notes the authors' caveat |
| F6 | "WIC reaches only about half of the people eligible" | https://www.fns.usda.gov/research/wic/eligibility-and-program-reach-estimates-2021 (→ fna.usda.gov/research/wic/eer/2021) | OK (curl) | "In the average month of 2021, WIC served an estimated 51.2 percent of those eligible for WIC" | CONFIRMED | primary | none |
| F7 | "Food banks run enrolment help: Feeding America's network assisted 312,000 SNAP applications last year." | Feeding America 2025 Annual Report PDF | OK (pdftotext) | "312,000 SNAP APPLICATIONS assisted by the network to help people access the food they need"; "*These numbers reflect fiscal year 2025 (July 1, 2024 through June 30, 2025)." | CORRECTED ("last year" read in November 2026 points to the wrong year) | charity's own figures | → "in the year to June 2025"; s11 dated |
| F8 | Summer EBT evidence label moderate | SEBTC report | OK | Randomised evaluation of the benefit (F9); the action is a state opt-in ask. | CONFIRMED | primary | none |
| F9 | "The summer grocery benefit for school-age children cut the worst child food insecurity by a third in its randomised evaluation" | USDA FNS SEBTC Demonstration: Summary Report (May 2016) | OK (pdftotext) | "SEBTC decreased the prevalence of the most severe food insecurity among children by one-third. Without SEBTC, 9.1 percent of households experienced VLFS-C; in contrast, with a $60 per child per month SEBTC benefit, 6.1 percent"; "benefits per summer month for each eligible school-age child"; chapter "Randomly Assigning and Notifying Households". | CONFIRMED | primary | none |
| F10 | "a dozen states still don't take part" | https://frac.org/summer-ebt | WebFetch (curl and headless Chrome 403) | "Yet children in 12 states are missing out on the nutrition they need to fuel their minds and bodies" … "because their state has chosen not to participate." (WebFetch's copy) | CONFIRMED | charity's own figures | s12 notes "12 states" |
| F11 | "New York does." | FRAC (as F10); https://www.fna.usda.gov/sebt/new-york | WebFetch; OK (curl) | FRAC's page names no states. USDA's New York page: "Good News! Summer EBT is Available in Your Location"; "Website: NY Summer EBT". | CORRECTED (true, but not in the cited FRAC page; USDA's page named in s12, no link added) | primary | s12 adds the USDA page by name |
| F12 | Evidence label weak: modest, short-term | An et al., PMID 30834852 | OK (curl) | "Interventions were modest in scale and usually short in follow-up duration." | CORRECTED ("modest" describes the studies' scale, not the size of the effect) | research | → weak: small, short studies |
| F13 | "Pantry programs improve diets modestly in the studies that exist." | An et al., PMID 30834852 | OK (curl) | "Fourteen articles evaluating twelve distinct interventions … Five were RCT and the remaining seven were pre-post studies. All studies found that food pantry-based interventions were effective in improving participants' diet-related outcomes." | CORRECTED (effective, in small and short studies) | research | → "Programs run in pantries improved clients' diets in the small, short studies that exist." Citation fixed (S13b) |
| F14 | "A dollar goes further than a can because the network buys in bulk" | https://www.feedingamerica.org/ways-to-give/faq/about-our-claims | OK (curl) | "Our partners donate truckloads of food they are unable to sell." "Our generous supporters help cover the cost of sorting, storing, and transporting that food". Nothing on bulk buying. | NOT IN SOURCE | charity's own figures | → "A dollar goes further than a can: most of what the network moves is donated, and money pays to sort, store and truck it." |
| F15 | "and most of what it moves is donated anyway" | Feeding America FY2026 audited statements | OK (pdftotext) | "Contributions of nonfinancial assets 4,714,444" of "Total public support and revenue … 5,334,035" (thousands); "Feeding America donated 2,669 million pounds … of food product to its network members." | CONFIRMED | primary | none |
| F16 | "“$1 provides 10 meals” is a leverage ratio, not a purchase" | https://www.feedingamerica.org/; about-our-claims FAQ | OK (curl) | Home: "Every $1 provides at least 10 meals to neighbors in need." Claims FAQ: "We divide the total number of pounds of donated food by what it costs to keep Feeding America running smoothly … Each meal is roughly 1.2 pounds, giving us our meal per dollar figure." Also on the home page today: "Make More Thanksgiving Meals Possible … Every $1 provides 20 meals." | CORRECTED (the quotation was not Feeding America's words; the leverage reading is confirmed). The 20-meal Thanksgiving line is a seasonal appeal, noted, not added | charity's own figures | → "“Every $1 provides at least 10 meals” is a leverage ratio, not a purchase."; s13 explains the conversion |
G. "What happens after" and the languages line (against source/impact/needs-2026-11-06.md)
| # | Page said | Source | Fetch | Source says (verbatim) | Verdict | Grade | Fix applied |
|---|---|---|---|---|---|---|---|
| G1 | Top Charities Fund: "GiveWell grants it on a rolling basis … and emails you once your money is committed" | https://www.givewell.org/top-charities-fund (needs C1.1, C1.2; re-read) | OK (curl) | "make grants on a rolling basis from the Top Charities Fund"; "We send a follow-up email to all Top Charities Fund donors after their donations are committed". | CONFIRMED | charity's own figures | none |
| G2 | "all of it to programs" | as G1 | OK (curl) | "GiveWell grants 100% of the value of donations designated for the Top Charities Fund (minus any fees charged by payment processors)". | CORRECTED | charity's own figures | → "all of it except payment fees" |
| G3 | "Every grant from the fund so far has gone to malaria nets, malaria medicine, vitamin A or vaccination incentives" | as G1 (full table, 61 rows, re-read today) | OK (curl) | Rows include "February 2022 … $7.8 million … Sightsavers — Deworming in Nigeria and Cameroon", "February 2022 … $4.8 million … Evidence Action's Deworm the World Initiative", "June 2021 … $1.5 million to Sightsavers", Schistosomiasis Control Initiative (2014 to 2018) and GiveDirectly (2014 to 2015). Every row from "May 2022" onward is Against Malaria Foundation, Malaria Consortium, Helen Keller Intl or New Incentives. | CORRECTED (the needs file's "meaning check" read only the newest rows) | charity's own figures | → "Since May 2022 every grant from the fund has gone to …" |
| G4 | "not to malnutrition treatment" | as G1 | OK (curl) | No row names a malnutrition-treatment grantee. | CONFIRMED | charity's own figures | split into its own sentence |
| G5 | "a health worker measures each child's upper arm and hands the family one or two sachets of peanut paste a day" | IRC research summary PDF (needs C4.1; re-read) | OK (pdftotext) | "a Combined Protocol is proposed that admits children with MUAC<125mm and/or oedema and treats as follows: MUAC <115mm - 2 sachets RUTF/d; MUAC 115mm- <125mm -1 sachet RUTF/d)." That dose is the IRC's combined protocol; GiveWell's Taimaka page gives no dose. | CORRECTED (attributed to the protocol it comes from) | charity's own figures | → "hands the family sachets of peanut paste, one or two a day under the IRC's simplified protocol" |
| G6 | "Taimaka describes six to ten weekly visits" | needs C4.3 | needs file | "These CHWs provide therapeutic foods, medicines, and vaccinations to help a child recover over the course of 6-10 weekly visits." | CONFIRMED | charity's own figures | none |
| G7 | "In the IRC's study in Mali, 92% of children recovered, and 26% of those relapsed within six months." | IRC research summary PDF | OK (pdftotext) | "IRC and the Mali Ministry of Health treated more than 27,500 children with a simplified, combined protocol. Overall, 92% of the children treated recovered." Separately: "the 6-month incidence of relapse … among 420 children in western Mali who recovered from acute malnutrition … Cumulative relapse incidence to MUAC < 125 mm was relatively high (26%)". | CORRECTED (two different studies; "26% of those" applied the 420-child figure to the 27,500) | charity's own figures | → "In the IRC's work in Mali, 92% of more than 27,500 children treated recovered; in a separate study of 420 who recovered, 26% relapsed within six months." |
| G8 | "You won't get a child's name or a photo. None of GiveWell, ALIMA, Taimaka or the IRC runs a sponsorship scheme; the money pays for a program." | needs C3 | needs file | "What the page can say accurately: no sponsorship program and no promise of a child's name or photo was found for any of the four." | CONFIRMED (an absence, as read on each site's pages and sitemap addresses) | primary | none |
| G9 | "A $47 "Inspired Gift" is unrestricted too: it doesn't buy that carton for one child." | needs C4.2 | needs file | "Inspired Gifts represent lifesaving aid. Your donation is unrestricted, allowing UNICEF to help children where the need is greatest." | CONFIRMED | charity's own figures | none |
| G10 | Stopping GiveWell: "in its donor portal under "My Recurring Donations" or by writing to donations@givewell.org" | needs C2.1 | needs file | "From the "My Recurring Donations" tab in your account, you can change your donation date, amount, and payment information, or cancel at any time."; "You can also send us an email at donations@givewell.org". | CONFIRMED | primary | none |
| G11 | Stopping IRC: "by calling 1-855-973-7283 or writing to partners@rescue.org" | needs C2.3 | needs file | "You can change or cancel your recurring donation any time by contacting us at 1-855-973-7283."; "contact our Supporter Care team at partners@rescue.org or 855-973-7283." | CONFIRMED | primary | none |
| G12 | "ALIMA doesn't publish steps; write to info.usa@alima.ngo." | needs C2.2 | needs file | "ALIMA's pages do not say how to cancel."; "Please reach out to us at info.usa@alima.ngo if you have any questions". | CONFIRMED | primary | none |
| G13 | "After you call, the office records it." | needs C5.1 | needs file | "Congressional staff process and record large volumes of identical form emails"; one House legislative director: "Their comments are recorded, and the data is shared with the entire staff". | CONFIRMED (one staffer's words plus CMF's description) | primary | none |
| G14 | "Offices mostly answer their own constituents" | needs C5.2 | needs file | "Most Members answer only those from their districts or states." | CONFIRMED | primary | "only" added to match |
| G15 | "a stock reply in under a week, a new one in two to four weeks, and sometimes none at all. That's normal" | needs C5.2 | needs file | "acceptable and achievable average turnaround goals are less than one week for responses using pre-approved language and two to four weeks for those requiring new text"; "many messages go unanswered for three months, six months, or even longer"; the needs file: "None describes a reply process specific to phone calls." | CORRECTED (goals for written messages, stated as what happens after a call) | primary | → "Their reply goals are for letters and emails: under a week for a stock answer, two to four weeks for a new one, and many wait months. The reports say nothing about replies to calls." |
| G16 | "your call still counted" | needs C5.1 | needs file | "Their comments are recorded, and the data is shared with the entire staff for immediate action." | CONFIRMED | primary | none |
| G17 | Languages: "211 answers in more than 180" | https://www.211.org/get-help/mental-health (needs B; re-read); United Way facts page | OK (curl) | 211.org: "211 conversations are confidential, can be made anonymously, and are available in 180 languages upon request." United Way: "Available 24/7 in 180+ languages". | CORRECTED (the operator's own page says 180, on request) | primary | → "211 answers in 180 on request" |
| G18 | "Feeding America's food-bank finder has a Spanish version" | needs A3.2 | needs file | "Ingresa tu código postal para encontrar bancos de alimentos cercanos en la red de Feeding America." | CONFIRMED | primary | none |
H. Questions people ask
| # | Page said | Source | Fetch | Source says (verbatim) | Verdict | Grade | Fix applied |
|---|---|---|---|---|---|---|---|
| H1 | Q1: "$240 a year is about 4% to 7% of one life saved, whichever step you choose" | GiveWell top charities, Taimaka, IRC | OK (curl) | "$3,500 per life saved" … "$5,500 per life saved"; Taimaka $4,245 and IRC $5,487 per death (D1). | CONFIRMED (our arithmetic: 240 ÷ 5,500 = 4.4%, 240 ÷ 3,500 = 6.9%; 240 ÷ 5,487 = 4.4%, 240 ÷ 4,245 = 5.7%) | our arithmetic | none |
| H2 | Q1: "GiveWell puts … treatment at about $4,200 to $5,500 a death prevented" | GiveWell Taimaka page | OK (curl) | GiveWell gives the grant and the deaths averted (D1), not a Taimaka cost per death; the $4,245 is our division. | CORRECTED (attribution) | our arithmetic | "GiveWell puts" → "GiveWell's figures put" |
| H3 | Q1: "about two children's treatment at its $105 price" | GiveWell Taimaka page; Taimaka donate | OK (curl) | GiveWell: "Cost per child $105". Taimaka's own pages say $94 and $132. 240 ÷ 105 = 2.3. | CORRECTED ("its" read as Taimaka's) | evaluator | → "at GiveWell's $105 cost per child" |
| H4 | Q3: "and the field experiments agree for low-profile votes" | as E1 | as E1 | As E1. | CORRECTED | research | → "and two field experiments with state legislators agree" |
| H5 | Q4: "ALIMA runs the trials and publishes its accounts" | as C4, C5 | OK | As C4, C5. | CONFIRMED | primary | none |
| H6 | Q4: "GiveWell's top charities save more children's lives per dollar than malnutrition treatment does" (5 Nov row D31; the line changed) | as B10, B11 | OK (curl) | GiveWell rates treatment at 8x and 9x against a 10x bar (B10), but its own figures put a death averted at $4,245 and $5,487 against $3,500 to $5,500 for the top charities (B11), which is the page's step 2. | CORRECTED (the 5 Nov confirmation did not survive the arithmetic now on the page; D31 is superseded) | evaluator | → "GiveWell rates its top charities a little higher than malnutrition treatment. By its figures a life saved costs about the same either way; the evidence behind the top charities is stronger." |
| H7 | Q6: "Then donor health aid fell sharply in 2025" (s15 now cites only the Goalkeepers PDF; Fortune's 26.9% is no longer cited) | Goalkeepers 2025 PDF | OK (pdftotext) | "And this year, reductions to health funding have put achieving the SDGs even further out of reach." The projections assume "a 20% reduction to Development Assistance for Health in 2026 compared to 2024 funding levels"; no size is given for 2025. | NOT IN SOURCE ("sharply"; the cut itself is confirmed) | primary | → "Then donors cut health aid in 2025" (same wording as the Learn and Practice fixes) |
| H8 | Q7 question: "Should I volunteer abroad, or support an orphanage?" | Goldman et al., Europe PMC PMC7311356 | OK (curl) | The commission's evidence is about institutions: "volunteering at institutions for children carries great risks"; it also says "Some forms of volunteering can have beneficial outcomes". | CORRECTED (the "No" was broader than the evidence) | research | → "Should I volunteer at an orphanage abroad, or fund one?" |
| H9 | Q7: "orphanage donations and volunteering keep children in institutions" | as H8 | OK (curl) | "we consider volunteers, visitors, and private donations, which are all important drivers of institutionalisation." | CONFIRMED | research | worded "donations and volunteers help keep children in institutions" |
| H10 | Q7: "when most have a living parent" | as H8 | OK (curl) | "Submissions indicated that many children in institutions do have a living parent". ReThink Orphanages says "an estimated 80% are not actually orphans and have at least one living parent" with no source on its page. | CORRECTED | research | "most" → "many" |
| H11 | Q7: "short volunteer trips have no measured benefit for the children" | as H8; https://rethinkorphanages.org/ | OK (curl) | Goldman: "Volunteers can unintentionally add to this neglectful and fragmented care, especially in situations in which visitors stay in institutions for only a few days, weeks, or months, thus increasing the instability of the care arrangement." ReThink: "the transient nature of orphanage volunteering is harmful to children". Neither says "no measured benefit". | NOT IN SOURCE | research / charity's own figures | → "and that short volunteer stays add to the instability of their care" |
| H12 | Q7: "Money to a vetted program does more than your presence." | as H8 | OK (curl) | "Donors and volunteers should redirect their funding and efforts to community-based and family-based programmes". | NOT IN SOURCE (as worded) | research | → "It asks donors and volunteers to fund family and community care instead." |
| P1 | Prompt: "the malnutrition-treatment programs GiveWell funds (ALIMA, Taimaka, IRC)" | GiveWell Metrics Report 2025 | OK (pdftotext) | "ALIMA USA – Malnutrition treatment $6,611,130"; "The Taimaka Project – Malnutrition treatment $614,133"; "International Rescue Committee – Malnutrition treatment $16,825,525". | CONFIRMED | evaluator | none |
S. The source list
| # | Page said | Source | Fetch | Source says (verbatim) | Verdict | Grade | Fix applied | |
|---|---|---|---|---|---|---|---|---|
| S1a | s1: Black et al., link text "The Lancet", graded primary (coordinator finding 5) | E-utilities PMID 23746772; content/about.html | OK (curl) | "Lancet. 2013 Aug 3;382(9890):427-451." "Maternal and child undernutrition and overweight in low-income and middle-income countries." "Black RE(1), Victora CG(2) …" About page: "research is peer-reviewed work". | CORRECTED (grade) | research | primary → research; the title added as link text | |
| S1b | s1: Madewell et al. (BMJ Global Health), graded primary | 5 Nov rows L8, L39; content/about.html; Learn page s2 | read | A peer-reviewed article; the Learn page grades the same paper research. | CORRECTED (grade, to match the site's scale and the Learn page) | research | primary → research | |
| S3 | s3: GiveWell, "how it estimates cost-effectiveness" | https://www.givewell.org/how-we-work/our-criteria/cost-effectiveness | OK (curl) | Title "Cost-Effectiveness \ | GiveWell"; "Published: November 2017; Last Updated: April 2025"; "Estimates are often based on limited information and are therefore extremely rough." | CONFIRMED | evaluator | none |
| S4a | s4: Pryce et al., "Insecticide-treated nets for preventing malaria", Cochrane, 2018 (child mortality down 17%) | PMID 30398672 | OK (curl) | "Cochrane Database Syst Rev. 2018 Nov 6;11(11):CD000363." "Insecticide-treated nets for preventing malaria." "Pryce J(1), Richardson M, Lengeler C." | CONFIRMED | research | none | |
| S4b | s4: Meremikwu et al., "Seasonal malaria chemoprevention", Cochrane, 2012 | PMID 22336792 | OK (curl) | "Cochrane Database Syst Rev. 2012 Feb 15;2012(2):CD003756." "Intermittent preventive treatment for malaria in children living in areas with seasonal transmission." "Meremikwu MM(1), Donegan S, Sinclair D, Esu E, Oringanje C." | CORRECTED (title written from memory) | research | title corrected; effect noted, "(clinical malaria down about three-quarters)" | |
| S4c | s4: Imdad et al., "Vitamin A supplementation for preventing morbidity and mortality in children", Cochrane, 2022 (deaths down 12%) | PMID 35294044 | OK (curl) | "Cochrane Database Syst Rev. 2022 Mar 16;3(3):CD008524." "Vitamin A supplementation for preventing morbidity and mortality in children from six months to five years of age." "Imdad A(1), Mayo-Wilson E(2) …" | CORRECTED (title cut short; authors, year, venue, effect confirmed) | research | full title | |
| S4d | s4: Awasthi et al., the DEVTA trial, The Lancet, 2013, one million children | PMID 23498849 | OK (curl) | "Lancet. 2013 Apr 27;381(9876):1469-77." "Vitamin A supplementation every 6 months with retinol in 1 million pre-school children in north India: DEVTA, a cluster-randomised trial." "Awasthi S(1), Peto R, Read S, Clark S, Pande V, Bundy D". | CONFIRMED | research | none | |
| S4e | s4: DEVTA "(… no effect on deaths)" | as S4d | OK (curl) | "mortality ratio 0·96, 95% CI 0·89-1·03, p=0·22"; "cannot rule out some more modest effect." | CORRECTED | research | → "no significant effect on deaths" | |
| S4f | s4: GiveWell, Our mistakes (the March 2026 corrections) | as C1 | OK (curl) | "Last updated: March 2026". | CONFIRMED | evaluator | none | |
| S5a | s5: Taimaka "about nine times as cost-effective as cash, against a bar of ten" | GiveWell Taimaka page | OK (curl) | "approximately nine times as cost-effective as unconditional cash transfers. At the time we recommended this grant, our cost-effectiveness bar for recommending grants was 10x cash." | CONFIRMED | evaluator | none | |
| S5b | s5: Bailey et al., the ComPAS trial, PLOS Medicine, 2020 | https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1003192 | OK (curl) | Title: "A simplified, combined protocol versus standard treatment for acute malnutrition in children 6–59 months (ComPAS trial): A cluster-randomized controlled non-inferiority trial in Kenya and South Sudan"; first author "Jeanette Bailey"; "Published: July 9, 2020". | CONFIRMED | research | none | |
| S5c | s5: Cazes et al., the OptiMA trial in the DRC, linked to pubmed/35090531 | E-utilities 35090531 and 35303461 | OK (curl) | 35090531: "Trials. 2022 Jan 28;23(1):89." "Simplifying and optimising management of acute malnutrition in children aged 6 to 59 months: study protocol for a 3 arms community-based individually randomised controlled trial in decentralised Niger." "Daures M(1) …". 35303461: "Lancet Glob Health. 2022 Apr;10(4):e510-e520." "… (OptiMA-DRC): a non-inferiority, randomised controlled trial." "Cazes C(1), Phelan K(2) …". | CORRECTED (the link went to a different paper: the Niger protocol, first author Daures) | research | URL → https://pubmed.ncbi.nlm.nih.gov/35303461/; venue The Lancet Global Health added | |
| S5d | s5: OptiMA "(recovery 72% against 63% at six months)" | PMID 35303461 | OK (curl) | "The primary endpoint was a favourable outcome at 6 months, defined as being alive, not acutely malnourished … and with no further episodes of acute malnutrition throughout the 6-month observation period"; "282 (63%) of 446 children in the standard group and 325 (72%) of 450 children in the OptiMA group had a favourable outcome". | CORRECTED (the 72% is alive, recovered and not relapsed, not recovery alone) | research | → "(alive, recovered and not relapsed six months on: 72% against 63%)" | |
| S5e | s5 arithmetic: "$4,787,985 for an estimated 1,128 deaths averted, is about $4,245 a death, or about 40 treatments at $105; its IRC figures, about $5,500 a death at $215 a child, are about 25 treatments. $240 a year against $3,500 to $5,500 is 4% to 7%." | GiveWell Taimaka, IRC, top charities | OK (curl) | As D1, D2, H1. | CONFIRMED (4,787,985 ÷ 1,128 = 4,244.7; ÷ 105 = 40.4; 45,424 ÷ 1,128 = 40.3; 5,487 ÷ 215 = 25.5; 11,798 ÷ 463 = 25.5; 240 ÷ 3,500 = 6.9%; 240 ÷ 5,500 = 4.4%) | our arithmetic | none | |
| S7a | s7: ALIMA financial transparency (audited 2025 accounts) and 2025 annual report | ALIMA pages | OK (curl) | As C5; annual report page: "From Resilience to Hope: ALIMA Publishes Its 2025 Annual Report", "July 17, 2026". | CONFIRMED | charity's own figures | none | |
| S7b | s7: Charity Navigator, Taimaka (three stars) | as C10 | OK (curl) | As C10. | CONFIRMED | evaluator | none | |
| S7c | s7: "Taimaka's and Action Against Hunger's latest filings and audited statements, read through ProPublica's Nonprofit Explorer" | ProPublica EIN 843964208 and 133327220; Charity Navigator; AAH PDF | OK (curl) | ProPublica Taimaka: "Fiscal Year Ending Dec. 2024 … Revenue $682,142", no audit listed; Charity Navigator: "Financial Statements : No". ProPublica AAH, 2025: "Form 990 documents are not available for this tax period, but audit documents are available for download." The statements quoted in C17 were read on AAH's own site. | CORRECTED (Taimaka has no audited statements on file) | primary | → "Taimaka's latest filing (Form 990 for 2024; no audited statements on file), read through ProPublica's Nonprofit Explorer, and Action Against Hunger's audited 2025 statements from its own site"; AAH donate and Lifeline pages and the 82%/87% arithmetic named | |
| S7d | s7: "IRC's staff reduction from 18,000 to about 12,000 since February 2025 is from its own statements", graded charity's own figures | Semafor (C13); one web search for an IRC statement | OK (curl) | Semafor (as C13). The search found no IRC statement of these figures; The New Humanitarian (19 Feb 2025, 403 to curl) is the early-2025 report. | CORRECTED (the figures are Semafor's reporting, not an IRC statement read) | news | → "the IRC's fall from 18,000 to 12,000 staff in the year and a half after the 2025 aid cuts, reported by Semafor, 22 September 2026" news (no link added) | |
| S8a | s8: Bergan and Cole, "Call your legislator", Political Behavior, 2015 | Crossref 10.1007/s11109-014-9277-1; Springer (rendered) | OK (API); headless Chrome | "Call Your Legislator: A Field Experimental Study of the Impact of a Constituency Mobilization Campaign on Legislative Voting"; "Daniel E. Bergan", "Richard T. Cole"; "Political Behavior" vol 37 issue 1 pages 27-42; print 2015-03, online 2014-06-01. | CONFIRMED | research | subtitle restored | |
| S8b | s8: Butler and Nickerson, "Can learning constituency opinion affect how legislators vote?", 2011 | ISPS page | OK (curl) | "Daniel M. Butler and David W. Nickerson (2011) "Can Learning Constituency Opinion Affect How Legislators Vote? Results from a Field Experiment", Quarterly Journal of Political Science: Vol. 6:No 1, pp 55-83." | CONFIRMED | research | subtitle and venue added | |
| S8c | s8: "(field experiments on state legislators)" | as E1 | as E1 | "Michigan state legislators"; "a randomly selected half of the legislature" (New Mexico). | CONFIRMED | research | none | |
| S9a | s9: GAO, local and regional procurement, 2009 | GAO-09-570 (WebFetch of the product page; PDF) | WebFetch; pdftotext | "International Food Assistance: Local and Regional Procurement Can Enhance the Efficiency of U.S. Food Aid" … "but Challenges May Constrain Its Implementation"; "Published May 29, 2009". | CONFIRMED | primary | none | |
| S9b | s9: GAO, cargo preference, 2015 | GAO-15-666 PDF | pdftotext | "August 2015 … INTERNATIONAL FOOD ASSISTANCE Cargo Preference Increases Food Aid Shipping Costs, and Benefits Are Unclear". | CONFIRMED | primary | none | |
| S9c | s9: Nunn and Qian, "US food aid and civil conflict", American Economic Review, 2014 | AEA page | OK (curl) | "US Food Aid and Civil Conflict"; "Nathan Nunn", "Nancy Qian"; "vol. 104, no. 6, June 2014". | CONFIRMED | research | none | |
| S9d | s9: Christian and Barrett, "a critique of that finding", Economic Journal, 2024 | Crossref 10.1093/ej/uead091 | OK (API) | "Spurious Regressions and Panel IV Estimation: Revisiting the Causes of Conflict"; "Paul Christian", "Christopher B Barrett"; "The Economic Journal" vol 134 issue 659 pages 1069-1099; published online 2023-10-25, print 2024-03-18. | CONFIRMED (authors, venue, 2024 issue) | research | link text → the paper's title, plus "which questions the method behind that finding" | |
| S11a | s11: Finkelstein and Notowidigdo, "Take-up and targeting: experimental evidence from SNAP", 2018 (30,000 people; 6% to 18%) | NBER w24652 | OK (curl) | "Take-up and Targeting: Experimental Evidence from SNAP"; "Amy Finkelstein & Matthew J. Notowidigdo"; "Issue Date May 2018"; published "The Quarterly Journal of Economics, vol 134(3), pages 1505-1556" (2019); "30,000 elderly individuals". | CONFIRMED | research | "30,000 people" → "30,000 older adults" (coordinator finding 2) | |
| S11b | s11: USDA, Measuring the effect …, 2013 (9,811 households) | as F5 | OK (pdftotext) | "August 2013"; "9,811 households interviewed in 30 States". | CONFIRMED | primary | effect and the authors' causal caveat added | |
| S11c | s11: USDA, WIC eligibility and program reach, 2021 data | as F6 | OK (curl) | "National and State Level Estimates of WIC Eligibility and Program Reach in 2021". | CONFIRMED | primary | none | |
| S12a | s12: USDA, Summer EBT demonstration evaluation (very low food security among children down about a third) | as F9 | OK (pdftotext) | As F9. | CONFIRMED | primary | none | |
| S12b | s12: FRAC, Summer EBT state participation | as F10 | WebFetch | As F10. | CONFIRMED | charity's own figures | "(12 states not taking part)"; USDA New York page named (F11) | |
| S13a | s13: Feeding America, "10 meals" conversion its own; audited statements confirm pounds and values, not meals; donated goods most of what it moves | as F15, F16 | OK | As F15, F16; the FY2026 audit mentions meals only in "meal programs". | CONFIRMED | charity's own figures | conversion method spelled out | |
| S13b | s13: "the pantry-program studies are summarised in the Learn pages' sources and in the site's impact review" (link /learn/) | content/child-hunger-learn.html, content/learn.html; An et al. PMID 30834852 | read; OK (curl) | No Learn page cites a pantry study; the impact review is a repo file, not a page. The study behind the claim (impact review S89): "Public Health Nutr. 2019 Jun;22(9):1704-1716." "A systematic review of food pantry-based interventions in the USA." "An R(1), Wang J(2), …". | CORRECTED (the citation pointed at nothing a reader could check) | research | → An et al., linked at https://pubmed.ncbi.nlm.nih.gov/30834852/ (a host already on the page) | |
| S16a | s16: Goldman et al. linked to pubmed/32589867 (coordinator finding 1) | E-utilities 32589867, 32589873 | OK (curl) | 32589867: "Lancet Psychiatry. 2020 Aug;7(8):703-720." "Institutionalisation and deinstitutionalisation of children 1: a systematic and integrative review of evidence regarding effects on development." "van IJzendoorn MH(1) …". 32589873: "Lancet Child Adolesc Health. 2020 Aug;4(8):606-633." "Institutionalisation and deinstitutionalisation of children 2: policy and practice recommendations for global, national, and local actors." "Goldman PS(1) …". | CORRECTED | research | URL → https://pubmed.ncbi.nlm.nih.gov/32589873/ | |
| S16b | s16: title "Institutionalisation and deinstitutionalisation of children", venue "the Lancet Group Commission, 2020" | as S16a; Europe PMC full text | OK (curl) | As S16a; "The work of this Lancet Group Commission has been motivated by the evidence of such damaging effects marshalled in our accompanying review meta-analysis." | CORRECTED | research | full title; venue The Lancet Child & Adolescent Health, 2020, "part of a Lancet Group Commission" | |
| S16c | s16: ReThink Orphanages, on volunteering and donations | https://rethinkorphanages.org/ (→ bettercarenetwork.org/rethink-orphanages) | OK (curl) | "Well-meaning support for orphanages – through donations, volunteering, tourist trips and faith-based mission – is challenging care reform efforts." | CONFIRMED | charity's own figures | none | |
| S17 | s17: "211, About 211 (24 hours, free and confidential, 180+ languages)" | 211.org about-us; 211.org mental-health; United Way facts page | OK (curl) | About: "available to help, 24/7. Calls to 211 are confidential and can be anonymous." (no "free", no languages). Mental-health page: "available in 180 languages upon request." United Way: "a free, confidential service"; "180+ languages". | CORRECTED (the About page says less than the parenthesis claimed) | primary / charity's own figures | → "About 211 (24/7, confidential) and its mental-health help page (180 languages on request)" primary; "United Way, which runs 211.org, on its 211 facts page (a free service), read 6 Nov 2026" charity's own figures (no link added) | |
| S18 | s18: GiveWell, Top Charities Fund (granted on a rolling basis; the grants listed), read 6 Nov 2026 | as G1 | OK (curl) | "Published: June 2021; Last updated: September 2026". | CONFIRMED | charity's own figures | none | |
| S19 | s19: Taimaka home (six to ten weekly visits); GiveWell, Donor illusions, 2009; the four organisations' pages, none offering sponsorship | needs C3, C4.3 | needs file | As G6, G8; Donor illusions: "Your sponsorship contributions are not given directly to a child." (GiveWell quoting Save the Children, 2009). | CONFIRMED | charity's own figures / primary | none | |
| S20 | s20: GiveWell more information; IRC monthly giving; ALIMA US donation FAQ | needs C2 | needs file | As G10 to G12. | CONFIRMED | primary | none | |
| S21 | s21: CMF, Citizen-Centric Advocacy (2017) and Building Trust … (2022), graded research; s8 grades the same 2017 report primary | content/about.html; CMF PDFs (5 Nov D37; needs C5) | read | About page: "primary is the original document. research is peer-reviewed work". The CMF reports are not peer-reviewed; 5 Nov row D37 graded the 2017 report primary. | CORRECTED (one report carried two grades on one page) | primary | research → primary | |
| X1 | Coordinator finding 4: Shang and Croson; Caviola and Greene; Karlan and List | content/child-hunger-do.html | read | None of the three is cited on the Do page. | n/a | — | none |
Changed lines already covered by 5 November rows (not re-verified)
- Lede "about $100 to $200 a child" (D1, O4); "about 2.2 million" (D2); "a carton of peanut paste and a health worker who measures the child's arm each week" (O3).
- Step 1 "about $3,500 to $5,500 a life" and "save the same children malnutrition kills, in the same places" (D10); "from $5 a month" (O8, D11).
- Step 3 "Government funding runs out on 11 December 2026 … House bill cuts it by a quarter. The Senate hasn't acted." (D5 to D8, D9 for R49343 in s6).
- GiveWell form path and "$5 minimum" (D11); "Its four top charities …" (D10); "GiveWell funds these programs too, and rates them a little below …" (D12); ALIMA "choose United States, then Recurring donation" (D15); Taimaka "(northern Nigeria) treats malnutrition and nothing else" (D13); "GiveWell's figure is $105" (L41, D35); IRC "a third less therapeutic food" and "Monthly gifts go wherever IRC needs them" (D17); UNICEF USA "$47 Inspired Gift … unrestricted" (D19's own fix).
- Costs list "$47 buys one carton …" and "$100 to $200" (D19, D1: markup change only); "Taimaka's own estimate is lower, $1,500 to $4,500" (L14, L43).
- Call script figures "$165 million", "$1.2 billion to $900 million" (D23); "Your House member votes until 3 January" (D24); switchboard (D21); "a specific request and a local reason" (D22).
- Closer to home "751,000 American children" (D25); City Harvest and Feeding America finder (D26).
- Q2 and Q5 (citation renumbered only: D28, D29); Q6 "fell 63% between 1990 and 2023" and the Gates projection (D32; s15 now cites the Goalkeepers PDF directly, as L17 proposed).
- s6 (D9), s10 (D38), s14 (L45), s15 (L15, L17, L29).
- Not claims: the chips,
data-untilanddata-expiredattributes, the.localand.sharepagemarkup, the third AI prompt.
Summary
136 rows: A 8, B 12, C 19 (with the two checkout sub-checks), D 3, E 6, F 16, G 18, H 13 (with P1), S 41 (with X1).
| Verdict | Count | Rows |
|---|---|---|
| CONFIRMED | 81 | A2–A5, A7, A8; B1–B4, B7–B12; C1, C4–C6, C8, C10–C13, C15, C17; D1, D3; E4–E6; F3, F4, F6, F8–F10, F15; G1, G4, G6, G8–G14, G16, G18; H1, H5, H9, P1; S3, S4a, S4d, S4f, S5a, S5b, S5e, S7a, S7b, S8a–S8c, S9a–S9d, S11a–S11c, S12a, S12b, S13a, S16c, S18–S20 |
| CORRECTED | 42 | A6; B5, B6; C2, C3, C16; D2; E1–E3; F5, F7, F11–F13, F16; G2, G3, G5, G7, G15, G17; H2–H4, H6, H8, H10; S1a, S1b, S4b, S4c, S4e, S5c, S5d, S7c, S7d, S13b, S16a, S16b, S17, S21 |
| NOT IN SOURCE | 9 | A1, C9, C14, F1, F2, F14, H7, H11, H12 (each reworded to what the source says, or removed) |
| NEWER DATA | 1 | C7 (GiveWell funded ALIMA in Ngouri, Chad again in October 2025, after rating Chad below its bar in 2024) |
| BLOCKED | 2 | C9b (Every.org checkout, Vercel checkpoint), C9c (PayPal donate button, 403) |
| UNREACHABLE · UNFETCHABLE | 0 | — |
| n/a | 1 | X1 |
Hosts that refused tools but whose content was read another way: gao.gov (PDFs via WebFetch's saved file), frac.org (WebFetch's copy only), academic.oup.com (Crossref and the I4R working paper), link.springer.com (headless Chrome), pubmed.ncbi.nlm.nih.gov (E-utilities), actionagainsthunger.org donate page (headless Chrome).
Where the walk stops
- FRAC's "12 states" is WebFetch's copy of a page that refused curl and headless Chrome; it matches the 5 Nov reading and the impact review's list.
- Every per-death figure for malnutrition treatment rests on GiveWell's use of 1980s–90s observational mortality data (B9); the IRC's $5,487 is a Burkina Faso example.
- The IRC's 92% and 26% are the IRC's own summaries of two studies; the staff figures are Semafor's reporting.
- The "many have a living parent" sentence in Goldman et al. reports submissions to a committee, not a count.
- Whether Taimaka's Every.org or PayPal checkouts allow a monthly gift is unknown: both blocked automated reads.
- The needs-file lines rest on that file's quotes (read 08:55 to 09:21 EST today); 211, United Way, the IRC PDF and GiveWell's fund page were re-read for this table.
What would overturn these, and when they go stale
- A GiveWell page deciding Taimaka's renewal (expected "in 2026"), or a page for its October 2025 ALIMA grant, changes C7, C11, C12.
- Every.org or PayPal showing a monthly option for Taimaka would make C9's "no monthly option on its page" incomplete.
- Action Against Hunger publishing 2026 accounts, or changing its "90%" line, changes C16 and C17.
- USDA's SNAP eligibility page is mid-update for new legislation (needs A3.5); re-read before quoting the 30- and 7-day rules.
- FRAC and USDA Summer EBT participation change each year;
source/stale.tsvhas 2027-03-15 for this. - Feeding America's Thanksgiving "Every $1 provides 20 meals" appeal is seasonal; the standing line is "at least 10 meals".
Proposals for files outside this page (not applied)
source/local/copy.json, child-hunger,"if": "has_snap": "You hear back within 30 days, or within 7 if your household has almost no money." → "You hear back within 30 days, and may get benefits within 7 if your household has almost no money." (USDA: "You may be eligible to receive SNAP benefits within 7 days … if you meet additional requirements.")content/index.html: repeats none of the claims corrected here; no change proposed.source/impact/needs-2026-11-06.md, C1.1 "Meaning check": "Every Top Charities Fund recipient in the table is a malaria, vitamin A or vaccination-incentive program" is true only from May 2022; the table also lists deworming grants (Sightsavers and Deworm the World, February 2022; Sightsavers, June 2021; SCI, 2014 to 2018) and GiveDirectly (2014 to 2015).- Links not added, because the brief allowed no new links: GiveWell's ALIMA Chad page (same host) and its Metrics Report (files.givewell.org, a new host), Semafor (new host), United Way's 211 facts page (new host), USDA's Summer EBT New York page and 211's mental-health page (hosts already on the page). Each is named in the source list as text. A person decides; the gate locks on new hosts.
checked: 5 Nov 2026was left as is, as on the Learn and Practice pages; once the topic is complete, "5 and 6 Nov 2026" would be accurate.- Unchanged line on this page, outside the brief's scope: s2 grades UNICEF USA's gift page
primary; on the site's scale an organisation describing its own gift ischarity's own figures(the needs file grades it so). - 5 November row D31 is superseded by H6 above; the 5 Nov file is left unedited.
Contest a fact, suggest something, or tell us what’s broken
If someone is in danger right now, don’t use this form. Call your local emergency number (911 in the US). For trafficking: 1-888-373-7888, or text 233733.
No name, no account. Your note goes into the site’s private records, nothing about your browser or connection is stored with it, and the email box is the only way we’d know who you are. The weekly review reads every note, re-reads the original source, and fixes the page if it should. A contested fact and its outcome are listed on Contested, in the review’s words, never yours.
No tracking on any link.Ticks stay on this device.What changedContested