Records · Verification table
AI pages: the lines added after the 5 November read, verified 2026-11-06
The working record behind the pages, published as written. Source file source/verify/ai-2026-11-06-additions.md. All records.
Pages: content/ai.html, content/ai-learn.html, content/ai-practice.html (the Do page is a separate pass). Scope: every claim on a line added or changed between ee6f7b8 and HEAD (8551ff8; no commit after 244a844 touched these three files) that has no row in source/verify/ai-2026-11-05.md. Lines whose wording was applied from a 5 Nov row were checked against that row, and re-read where the applied words went beyond its quote (A-L12, A-L14, A-L15).
Method
- Fetches ran 09:24–09:36 EST on 6 Nov 2026. Each page was saved with curl (browser user agent, cookie jar) into its own folder and turned to text locally. PDFs were read with
pdftotext. Quotes below are copied from those files, not from a fetch summary. Three pages were read through WebFetch: the FTC alert (curl gets FTC's 404 script response), the AEA trial registry, and the Europe PMC refusal. Their quotes are as WebFetch returned them. - Nothing came from memory. On 6 Nov the meta-analysis contest and the "With others" section were written from notes in
source/impact/ai-2026-11-05.md, which marks Wu and Bartoš[record](index only). Every author, title, year and venue below was read on the paper's own page today. - Two web searches were used only to find pages (a 2026 Bartoš manuscript; the NHH working paper) and to test two absence claims. Every claim rests on a page that was then read. Where only a search summary exists, the row says so.
- Hosts that refused today: nature.com sends WebFetch through a login redirect (303), and curl with cookies read it. europepmc.org returns a Cloudflare "Just a moment..." 403 to every automated request, the homepage included. The article record was read through the Europe PMC REST API on ebi.ac.uk instead. cetas.turing.ac.uk HTML 403 (the PDF downloaded). sciencedirect.com 403. openaccess.nhh.no failed DNS lookup; its handle resolves to nva.sikt.no, and that API served the PDF. consumer.ftc.gov 404 to curl (read via WebFetch).
Verdicts: CONFIRMED, CORRECTED (fix given), NOT IN SOURCE, NEWER DATA (a later version or newer work exists), UNREACHABLE (DNS/proxy/timeout), BLOCKED (403/405/script-only 404), UNFETCHABLE (the host says the page is gone). Grades use the About page's words: primary · research · preprint · charity's own figures · news · industry research.
---
1. Overview: content/ai.html
| # | Claim (words on the page) | Page | Source URL | Fetch | Exact quote | Verdict | Grade | Fix applied |
|---|---|---|---|---|---|---|---|---|
| A-O1 | Origin line: "Towards a more loving, hopeful and faithful humanity, an essay from 2025" | Overview | https://medium.com/feed/@isaac.holze | 200 (curl, RSS) | item "Towards a more Loving, Hopeful and Faithful Humanity", <pubDate> "Mon, 19 May 2025 13:54:30 GMT" | CONFIRMED | primary (author's feed) | none |
| A-O2 | "its page lists what has changed since it was written" | Overview | source/corrections.json → corrections.ai; site/ai/essay/index.html | internal | three correction entries; the built essay page contains "Since this was written" | CONFIRMED (internal) | — | none |
2. Learn: content/ai-learn.html
2a. The meta-analysis contest and source 11 (written from memory on 6 Nov)
| # | Claim (words on the page) | Page | Source URL | Fetch | Exact quote | Verdict | Grade | Fix applied |
|---|---|---|---|---|---|---|---|---|
| A-L1 | Question: "Do the meta-analyses show AI tutoring helps?" | Learn, contest | nature.com/articles/s41599-026-07019-z; nature.com/articles/s41599-025-04787-y; api.osf.io/v2/preprints/8vs32/ | 200 / 200 / 200 | Wu title: "ChatGPT’s impact on student learning outcomes: a meta-analysis of 35 experimental studies"; Wu inclusion rule: "the research must use ChatGPT either as a direct learning tool or as a supported learning application". Wang and Fan title: "The effect of ChatGPT on students’ learning performance, learning perception, and higher-order thinking: insights from a meta-analysis". Bartoš title: "Adjusting for Publication Bias Reveals No Evidence for the Effect of ChatGPT on Students’ Learning Performance…" | CORRECTED. All three pool any use of ChatGPT in learning, not tutoring. "Tutoring" blurs the hints-vs-answers line the page is built on | research / preprint | → "Do the meta-analyses show ChatGPT helps students learn?" |
| A-L2 | "One 2026 pooling of 35 experiments" | Learn, contest | https://www.nature.com/articles/s41599-026-07019-z | 200 (curl) | "The analysis included 35 studies published between 2022 and 2024, involving 4193 participants."; methods: "limited to experimental designs, including both experimental and quasi-experimental approaches" | CORRECTED (minor): studies, some quasi-experimental | research | → "A 2026 pooling of 35 studies"; s11 "35 experimental studies" |
| A-L3 | "finds a large average gain"; s11 "a large average effect" | Learn, contest + s11 | https://www.nature.com/articles/s41599-026-07019-z | 200 (curl) | "The results indicated a moderately positive effect of ChatGPT on student learning outcomes ( g = 0.670)" | CORRECTED: the authors call it moderate | research | → "finds a moderate average gain"; s11 "the authors call the effect “moderately positive”" |
| A-L4 | "another finds the gain disappears once publication bias is corrected" | Learn, contest | https://api.osf.io/v2/preprints/8vs32/ ; https://api.osf.io/v2/preprints/h529e/ ; https://ejwagenmakers.com/papers.html | 200 / 200 / 200 | 2025 preprint (8vs32): "In a recent article, Wang and Fan1 conducted a comprehensive meta-analysis … featuring 51 studies." … "Here we show that these effects greatly diminish once publication bias is accounted for, and the evidence in favor of the benefits disappears." 2026 manuscript (h529e): "Publication bias-adjusted analyses reveal model-averaged effects approximately one-third the magnitude of unadjusted estimates (standardized mean difference, SMD = 0.196, 95% credible interval from 0.000 to 0.323)" | CORRECTED. The "gain disappears" finding is a re-analysis of the retracted Wang and Fan paper, not "another" pooling. The 2026 pooling by the same group finds about a third, not zero | preprint | → "Earlier, a preprint re-analysing its data found no evidence of a benefit once publication bias was accounted for." |
| A-L5 | s11: "Bartoš et al., 2026 (the effect vanishes after publication-bias adjustment; abstract read)" | Learn, s11 | https://api.osf.io/v2/preprints/8vs32/ (+ /bibliographic_contributors/, /versions/); https://ejwagenmakers.com/papers.html | 200 | OSF: "date_published": 2025-05-12, one version (8vs32_v1); contributors "František Bartoš", "Patrícia Martinková", "Eric-Jan Wagenmakers". Wagenmakers' list: "Bartoš, F., Martinková, P. & Wagenmakers, E.-J. (2025). Adjusting for publication bias reveals no evidence for the effect of ChatGPT on students’ learning performance, learning perception, and higher-order thinking. Manuscript rejected for publication." | CORRECTED. The citation fuses two papers: the finding and "abstract read" are the 2025 preprint's (the source named in the impact file, line 228); the year 2026 is not | preprint | → "Bartoš, Martinková and Wagenmakers, PsyArXiv, May 2025, a re-analysis of Wang and Fan’s data (once publication bias is accounted for, “the evidence in favor of the benefits disappears”; abstract read)" |
| A-L6 | (newer work by the same group, not on the page) | — | https://api.osf.io/v2/preprints/h529e/ ; https://ejwagenmakers.com/papers.html | 200 | "Effect of Artificial Intelligence on Learning: A Meta-Meta-Analysis", posted 2026-01-28: "a study-level meta-meta-analysis of 1,840 effect sizes from 67 meta-analyses. The results show strong evidence of severe publication bias and extreme between-study heterogeneity." … "broad claims of generalized learning gains resulting from AI/LLMs appear premature". List: "Bartoš, F., Bujak, O. Z., Martinková, P., & Wagenmakers, E.-J. (2026) … Manuscript submitted for publication." | NEWER DATA | preprint | none (a new source and link; proposed, see §5) |
| A-L7 | "a widely cited 2025 meta-analysis saying it helps" | Learn, contest | https://www.nature.com/articles/s41599-025-04787-y | 200 (curl; WebFetch got a 303 to idp.nature.com) | "Published: 06 May 2025"; "566k Accesses 486 Citations 1106 Altmetric" (as of 6 Nov 2026); "ChatGPT has a large positive impact on improving learning performance ( g = 0.867)" | CONFIRMED | research | wording only: "that found a large gain" |
| A-L8 | "was retracted in April 2026"; s11 "retracted 22 April 2026" (link s41599-026-07310-z) | Learn, contest + s11 | https://www.nature.com/articles/s41599-025-04787-y ; https://www.nature.com/articles/s41599-026-07310-z | 200 / 200 | "This article was retracted on 22 April 2026". Notice: "Retraction Note: The effect of ChatGPT on students’ learning performance…", "Published: 22 April 2026"; "The Editor has decided to retract this article due to concerns relating to discrepancies in the meta‑analysis. These concerns were initially raised by Magnus Ingebrigtsen and Marko Lukic. … The authors have not responded to correspondence regarding this retraction." | CONFIRMED | research | none |
| A-L9 | s11: "Wang and Fan, the 2025 meta-analysis, Humanities and Social Sciences Communications" | Learn, s11 | https://www.nature.com/articles/s41599-025-04787-y | 200 | "Jin Wang 1 & Wenxiang Fan 1 , 2"; "Humanities and Social Sciences Communications volume 12 , Article number: 621 ( 2025 )"; dc.type ReviewPaper | CONFIRMED | research (retraction stated beside it) | none |
| A-L10 | s11: "Wu et al., Humanities and Social Sciences Communications 13, 2026" | Learn, s11 | https://www.nature.com/articles/s41599-026-07019-z | 200 | dc.creator "Wu, Xinning", "Zhu, Pei", "Zhang, Jinliang", "Yin, Mengwei", "Wang, Yingxi"; "Published: 26 March 2026"; citation_volume 13; citation_firstpage 684 | CONFIRMED | research | none |
| A-L11 | s11: Wu "4,193 participants" | Learn, s11 | as A-L10 | 200 | "involving 4193 participants" | CONFIRMED | research | none |
| A-L12 | s11 grade span for Wu and Bartoš: "research, unread" | Learn, s11 | as A-L3, A-L5 | 200 | Wu read today: abstract and its bias test ("the publication bias test revealed no significant publication bias"). Bartoš: PsyArXiv preprint, "Manuscript rejected for publication" | CORRECTED (grade) | — | Wu → research ("abstract and bias tests read"); Bartoš → preprint |
2b. Other added or changed lines
| # | Claim (words on the page) | Page | Source URL | Fetch | Exact quote | Verdict | Grade | Fix applied | |
|---|---|---|---|---|---|---|---|---|---|
| A-L13 | Words, cheapfake: "In the 2024 European elections, viral disinformation leaned on old methods…" | Learn, words | https://cetas.turing.ac.uk/sites/default/files/2024-09/cetas_briefing_paper_-_ai-enabled_influence_operations_-_threat_analysis_of_the_2024_uk_and_european_elections.pdf | 200 (PDF; the HTML page 403) | Title: "AI-Enabled Influence Operations: Threat Analysis of the 2024 UK and European Elections", Sam Stockwell, September 2024. "A prominent category of AI-enabled disinformation which emerged in all three election contexts"; "In all three cases however, the articles which achieved virality were often ones which were either human-crafted or recirculated by social media influencers" | CORRECTED (scope): the finding covers three elections (UK, EU, French). "The 2024 European elections" reads as the EU vote alone, or all of Europe | primary | → "In the 2024 UK, French and EU elections" (fact 5's wording) | |
| A-L14 | "…leaned on old methods, bot comments and influencers, more than on generative AI" | Learn, words | as A-L13 | 200 | "Generative AI played less of a role in boosting the virality of disinformation compared to traditional interference methods and human influencers. Malicious actors continued to rely on well-established disinformation tactics, such as flooding comments sections on social media with bot accounts (astroturfing), and exploiting politically-aligned social media influencers to spread disinformation" | CONFIRMED | primary | none | |
| A-L15 | Words, Content Credentials: "Credentials can be separated from the file and are often stripped when it's re-uploaded" | Learn, words | https://c2pa.org/faqs/ | 200 (curl) | "What happens if C2PA metadata is stripped from the media it represents? While C2PA Manifests are typically embedded in the asset, they can be separated."; "providing resilience in cases where metadata is accidentally or intentionally stripped."; "While metadata can be removed inadvertently or intentionally" (the FAQ's one "often" is in "often referred to as “Durable Content Credentials”") | NOT IN SOURCE ("often", "when it's re-uploaded"); "can be separated" CONFIRMED | primary | → "Credentials can be separated from the file or stripped, by accident or on purpose, so their absence proves nothing." (The Learn page cites no source for this entry; see §5) | |
| A-L16 | Contest 1: "one 2026 preprint found a gain from tutoring-style use that still held a week later" | Learn, contest | https://arxiv.org/abs/2607.08849 | 200 (curl) | "[Submitted on 9 Jul 2026 ( v1 ), last revised 4 Oct 2026 (this version, v2)]"; "complete unaided assessments immediately and one week later. AI access raises test scores, and a smaller gain persists a week later. Gains remain for students who use AI as a tutor ("augmentation") and fade for students who have AI write for them ("automation")." | CONFIRMED | preprint | none | |
| A-L17 | s1: new link https://europepmc.org/article/PMC/PMC12232635 | Learn, s1 | that URL; also europepmc.org/articles/PMC12232635 and europepmc.org/ | 403 to curl and WebFetch; body "Just a moment..." (Cloudflare), the homepage too | — | BLOCKED (the record behind it is confirmed in A-L18; the URL is Europe PMC's own article path; tools/links.py: 0 problems) | — | none | |
| A-L18 | s1: "PNAS 122(26), 2025 … open access via Europe PMC" | Learn, s1 | https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=PMCID:PMC12232635&resultType=core&format=json | 200 (REST API) | "title": "Generative AI without guardrails can harm learning: Evidence from high school mathematics."; journal "Proceedings of the National Academy of Sciences of the United States of America", "volume": "122", "issue": "26"; "isOpenAccess": "Y"; "inEPMC": "Y"; "firstPublicationDate": "2025-06-25"; full text served at …/rest/PMC12232635/fullTextXML (200) | CONFIRMED | research | none | |
| A-L19 | s8: "Common Sense's institute says it is funded by philanthropy and by industry, including makers of some products it rates." (link …/risk-assessments/methodology) | Learn, s8 | https://institute.commonsensemedia.org/risk-assessments/methodology | 200 (curl) | page title "How We Review \ | Youth AI Safety Institute"; "The Youth AI Safety Institute is funded by both philanthropy and industry, including the makers of some of the technologies we evaluate. The Institute is solely responsible for its standards, research, and evaluations, and maintains complete editorial independence over published results." | CONFIRMED | charity's own statement (no grade span on this sentence) | none |
3. Practice: content/ai-practice.html
| # | Claim (words on the page) | Page | Source URL | Fetch | Exact quote | Verdict | Grade | Fix applied |
|---|---|---|---|---|---|---|---|---|
| A-P1 | Calendar: "One reminder on the first of each month that opens this page. It holds no dates or figures" | Practice | tools/app.js (§5) | internal | 'DTSTART:' + y + pad(m) + '01T100000', 'RRULE:FREQ=MONTHLY', 'DESCRIPTION:' + icsText('Open ' + url + ' and do what is due this month.'), 'URL:' + url | CONFIRMED (code; the event's own start date is its only date) | — | none |
| A-P2 | With others 1: "Enforced bans raised test scores in England" | Practice | https://cep.lse.ac.uk/pubs/download/dp1350.pdf | 200 (PDF) | "We find that following a ban on phone use, student test scores improve by 6.41% of a standard deviation. Our results indicate that there are no significant gains in student performance if a ban is not widely complied with."; "four English cities (Birmingham, London, Leicester and Manchester)"; "We also created an alternative measure of compliance using punishment at the school after a phone is used. Results were similar" | CONFIRMED | research (see A-P20) | none |
| A-P3 | "a ban on paper did nothing" | Practice | as A-P2 | 200 | "we find larger effects in schools that report a high level of compliance with a ban compared to schools where compliance is weak, where there is no statistically significant impact."; "A school was considered to have a high-compliance ban if the response was greater than four." | CORRECTED: "did nothing" overstates a non-significant estimate. The measure is pupils' compliance as rated by the head | research | → "where pupils didn’t keep to the ban, there was no significant gain." |
| A-P4 | "cut bullying in Norway" | Practice | https://doi.org/10.3368/jhr.0224-13403r2 (→ jhr.uwpress.org/content/early/2026/03/04/jhr.0224-13403R2); working paper via https://hdl.handle.net/11250/3119200 → nva.sikt.no (PDF) | 200 / 200 (openaccess.nhh.no: DNS failure, so the handle was used) | JHR abstract: "Bans reduce psychological health care use among girls, lower bullying for both genders". The JHR abstract names no country. Author's working paper (NHH SAM 01/24, February 2024): "by studying the effects of banning smartphones from Norwegian middle schools" | CONFIRMED | research | own sentence: "Bans cut bullying in Norway." |
| A-P5 | "Enforced" carried to Norway by "Enforced bans … and cut bullying in Norway" | Practice | as A-P4 | 200 | The JHR abstract does not mention enforcement or compliance. It does say "banning smartphones in middle schools has no average effect on education or mental health but masks important gender differences" | NOT IN SOURCE | research | split so "enforced" applies to England only |
| A-P6 | "two wellbeing studies found no difference" | Practice | https://pmc.ncbi.nlm.nih.gov/articles/PMC11984610/ ; https://www.frontiersin.org/journals/developmental-psychology/articles/10.3389/fdpys.2026.1796471/full | 200 / 200 | Goodyear: "Mean WEMWBS score was 47 (SD = 9) with no evidence of a difference between groups (adjusted mean difference −0.48, 95% CI −2.05 to 1.06, p = 0.62)." Vereecke: "No short-term changes emerged in social loneliness, friend support, or cyberbullying victimization, although cyberbullying perpetration declined slightly. Over time, social loneliness increased modestly"; "Within the specific conditions of this school, a total smartphone ban was not associated with clear improvements in adolescents' social wellbeing" | CORRECTED: true of Goodyear; Vereecke found changes, just no clear improvement | research | → "Two wellbeing studies found no clear benefit." |
| A-P7 | With others 2: "Delay the smartphone with ten families in a grade." | Practice | https://www.waituntil8th.org/ | 200 (curl) | "By signing the online pledge, you promise not to give your child a smartphone until at least the end of 8th grade as long as at least 10 families total from your child’s grade and school pledge." | CONFIRMED | charity's own figures | none |
| A-P8 | "The pledge has no studies behind it" | Practice | https://www.waituntil8th.org/why-wait ; /faqs ; /our-story | 200 ×3 | Why wait cites studies of harms: "The early results of a landmark study on brain development by the National Institute…"; "a new UC San Francisco-led study published in the Journal of Child Psychology and Psychiatry found"; "One study showed that 42% of online youth users have been exposed to online pornography." Its links include cep.lse.ac.uk/pubs/download/dp1350.pdf ("Another study") and ncbi PMC7954067 ("One study"). None tests the pledge. FAQ: "How will the wait until 8th pledge work?" gives no evaluation. Today's search found no evaluation either (a search summary, used only to test the absence) | CORRECTED: the site does put studies behind the pledge, but none tests it | charity's own figures | → "Nobody has tested the pledge itself" |
| A-P9 | "it's a way of not being the only parent" | Practice | https://www.waituntil8th.org/ | 200 | "Banding together helps decrease the pressure to have a phone at an early age."; "Parents feel powerless in this uphill battle and need community support" | CONFIRMED (paraphrase) | charity's own figures | none |
| A-P10 | s6: "Wait Until 8th, the pledge (no studies cited)" | Practice, s6 | as A-P8 | 200 | as A-P8 | CORRECTED | charity's own figures | → "(its site cites studies of phone harms, none of the pledge itself)" |
| A-P11 | With others 3: "A tutor designed to make students work helped" | Practice | https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12232635/fullTextXML (Bastani et al., full text) | 200 | "Student performance in the GPT Tutor arm was statistically indistinguishable from that of the control arm"; "These negative learning effects are largely mitigated by the safeguards in GPT Tutor."; "While the guardrails implemented in GPT Tutor appear to largely mitigate these negative effects, substantial work is required to enable generative AI to positively enhance rather than diminish education."; with the tool: "48% improvement in grades for GPT Base and 127% for GPT Tutor" | CORRECTED: it helped only while students had it. On the exam without it, the tutor group matched the no-AI group | research | → "A tutor designed to make students work did no harm; the answer bot did." |
| A-P12 | "nobody has evaluated a policy on paper" | Practice | not in s1 (Bastani). Tested at https://www.socialscienceregistry.org/trials/15594 | 200 (WebFetch) | Registry: "AI Policy in Education - A Randomized Controlled Trial", registered March 24, 2025; status "In development"; "Is the intervention completed?" "No"; no papers or reports; "Participating classes are randomly assigned to varying AI policy conditions" (China, secondary). A search found no published evaluation | NOT IN SOURCE (an absence claim; a trial is registered with no results) | — | → "Nobody has published an evaluation of a policy on paper." |
| A-P13 | With others 4: "In trials, “prebunking” raised people's recognition of manipulation techniques" | Practice | https://pmc.ncbi.nlm.nih.gov/articles/PMC9401631/ | 200 (curl) | "In seven preregistered studies, i.e., six randomized controlled studies ( n = 6464) and an ecologically valid field study on YouTube ( n = 22,632), we find that these videos improve manipulation technique recognition"; "prebunking has gained prominence as a means to preemptively build resilience against anticipated exposure to misinformation" | CONFIRMED | research | none |
| A-P14 | "how long the effect lasts isn't settled" | Practice | as A-P13 | 200 | "although our field study found that the videos were effective within a 24-hour window (with a median time of 18.4 hours between watching the video and responding to the survey question), we were unable to study how long the inoculation effect remains significant ( 46 , 47 )." (46 = Maertens et al. 2021, longitudinal; unread) | CONFIRMED | research | none |
| A-P15 | s4: "FTC, family emergency scams, March 2023" (link consumer.ftc.gov/consumer-alerts/2023/03/scammers-use-ai-enhance-their-family-emergency-schemes) | Practice, s4 | that URL | curl 404 (1,301 bytes; FTC's script response, as tools/links.py notes); WebFetch OK | headline "Scammers use AI to enhance their family emergency schemes"; "By Alvaro Puig Consumer Education Specialist March 20, 2023"; "Call the person who supposedly contacted you and verify the story."; "Use a phone number you know is theirs." | CONFIRMED | primary | none |
| A-P16 | s6: "Beland and Murphy, Ill communication, LSE, 2015" (link cep.lse.ac.uk/pubs/download/dp1350.pdf) | Practice, s6 | that URL; https://api.crossref.org/works?query.bibliographic=… ; https://doi.org/10.1016/j.labeco.2016.04.004 | 200 (PDF) / 200 (Crossref record) / sciencedirect 403 | "CEP Discussion Paper No 1350 May 2015 Ill Communication: Technology, Distraction & Student Performance Louis-Philippe Beland Richard Murphy"; "Published by Centre for Economic Performance London School of Economics and Political Science". Crossref: 10.1016/j.labeco.2016.04.004, "Ill Communication: Technology, distraction & student performance", Labour Economics, 2016-08, journal-article (record only; publisher page 403) | NEWER DATA (status): the citation is right; a peer-reviewed version exists | research | none (proposed, §5) |
| A-P17 | s6: "Abrahamsson, Journal of Human Resources, 2026, doi:10.3368/jhr.0224-13403r2 (abstract read)" | Practice, s6 | https://doi.org/10.3368/jhr.0224-13403r2 | 200 (resolves; DOIs ignore case) | "Smartphone Bans, Student Outcomes and Mental Health"; "Sara Abrahamsson" (sole author); "Published online before print March 11, 2026, 0224-13403R2"; "This article requires a subscription to view the full text." | CONFIRMED | research | none |
| A-P18 | s6: "Goodyear et al., school phone policies and adolescent wellbeing, 2025" (PMC11984610) | Practice, s6 | https://pmc.ncbi.nlm.nih.gov/articles/PMC11984610/ | 200 (curl) | "School phone policies and their association with mental wellbeing, phone use, and social media use (SMART Schools): a cross-sectional observational study"; The Lancet Regional Health - Europe 51:101211, "2025 Feb 4"; first author "Victoria A Goodyear"; "adolescents from 30 English secondary schools"; "We recruited 1227 participants (age 12–15)" | CONFIRMED (the link text describes the paper; it is not its title) | research | none |
| A-P19 | s6: "Vereecke et al., one school's ban, 2026" (Frontiers 10.3389/fdpys.2026.1796471) | Practice, s6 | https://www.frontiersin.org/journals/developmental-psychology/articles/10.3389/fdpys.2026.1796471/full | 200 (curl) | "More than disconnecting? School smartphone bans and adolescents' social wellbeing"; Frontiers in Developmental Psychology 4:1796471, online 2026/09/01; first author "Vereecke, Jana" (KU Leuven, Belgium); "Secondary school students ( N = 960 …) were followed across eight waves over one academic year"; "Within the specific conditions of this school" | CONFIRMED | research | none |
| A-P20 | s7: "Roozenbeek et al., Psychological inoculation improves resilience against misinformation on social media, Science Advances, 2022" (PMC9401631) | Practice, s7 | https://pmc.ncbi.nlm.nih.gov/articles/PMC9401631/ | 200 (curl) | citation_title "Psychological inoculation improves resilience against misinformation on social media"; authors Jon Roozenbeek, Sander van der Linden, Beth Goldberg, Steve Rathje, Stephan Lewandowsky; Science Advances 8(34) eabo6254; "2022 Aug 24"; "PMCID: PMC9401631" | CONFIRMED | research | none |
4. Grade labels on changed lines (checked against the About page's definitions)
About: "primary is the original document. research is peer-reviewed work, preprint when it isn't yet. … charity's own figures is an organisation reporting on itself. news is press coverage." tools/audit.py has no "secondary" or "report" label.
| # | Label on the page | Page | Basis | Verdict | Fix applied |
|---|---|---|---|---|---|
| A-G1 | IASR 2026 primary "(an expert report, not peer-reviewed)" | Learn s4 | The report is the original document of its own summary. Its 77% is "In one study, participants misidentified AI-generated text as human-written 77% of the time." (5 Nov read, row L10) | CONFIRMED as a label for the document. The 77% traces to a study the report cites, unread (§5) | none |
| A-G2 | CETaS primary | Learn s5 | Read today: "We identified just 16 confirmed viral cases of AI-enabled disinformation or deepfakes during the UK general election, while only 11 viral cases were identified…"; "there is no evidence that AI-enabled disinformation or deepfakes meaningfully impacted" (its own count) | CONFIRMED | none |
| A-G3 | Vaccari and Chadwick research "(abstract read)"; Schiff, Schiff and Bueno research | Learn s5 | Peer-reviewed journals (Social Media + Society; APSR 119(1)), as read 5 Nov (row L44); not re-read today | CONFIRMED (label) | none |
| A-G4 | Inside Privacy news (was "secondary") | Learn s9 | A law firm's blog (Covington) reporting the vote (row L27). It is secondary, and "news" is the nearest label the audit allows | CONFIRMED (nearest label) | none |
| A-G5 | research over four phone-ban studies | Practice s6 | JHR, Lancet Regional Health – Europe and Frontiers are peer-reviewed. Beland and Murphy's linked 2015 file is a discussion paper; its peer-reviewed version is Labour Economics 2016 (A-P16) | CONFIRMED (label fits the work) | none |
| A-G6 | Wait Until 8th charity's own figures | Practice s6 | The pledge's terms are the organisation's own words | CONFIRMED | none |
| A-G7 | Roozenbeek research; FBI and FTC primary | Practice s7, s4 | Science Advances, peer-reviewed; the agencies' own alerts | CONFIRMED | none |
5. Noticed outside this pass's scope (proposals only, not applied)
- Learn contest: the newer Bartoš, Bujak, Martinková and Wagenmakers meta-meta-analysis (PsyArXiv h529e, 28 Jan 2026; 1,840 effect sizes from 67 meta-analyses; bias-adjusted effect about a third, SMD 0.196, credible interval 0.000 to 0.323) is the broadest evidence on the question. Adding it means a new source, a new link and a new host (osf.io).
- Learn contest: Wu et al. report "no significant publication bias", which makes the dispute legible. It was not added (no new claims).
- Learn s11: Wu and Bartoš have no links. Candidates: https://www.nature.com/articles/s41599-026-07019-z (host already on the page) and https://osf.io/preprints/psyarxiv/8vs32_v1 (new host).
- Learn words, Content Credentials: the entry has no footnote on the Learn page. The C2PA FAQ is cited on the Practice page (s2) and could be cited here too.
- Learn fact 4: the IASR says "In one study" before the 77%. The page states it as a general fact (row L10 noted it). Suggested: "In one study, people judged AI text to be human 77% of the time." Row L11's point also stands: Deepfake-Eval measures AUC, not accuracy.
- Learn s9: row L36's recommended sources for Colorado and the state chatbot count (Arnold & Porter; Multistate) are still missing.
- Practice s6: name the journal version: "Beland and Murphy, Ill communication, LSE discussion paper 2015; Labour Economics, 2016". Read the ScienceDirect page first (403 today).
- Practice "With others" item 1: Abrahamsson's abstract also finds "no average effect on education or mental health". The page's "contested on wellbeing" label covers it. A future pass could say so in the line.
- The pages' "checked: 5 Nov 2026" stays: this pass re-read only the additions. A dated note like "additions checked 6 Nov 2026" would need a template change.
- Readers lose any tick on the three reworded "With others" items, because ticks are keyed by item text.
6. Summary by verdict
One verdict per numbered claim row in §1–§3 (41 rows; A-L15's confirmed half, "can be separated", is counted under its NOT IN SOURCE verdict):
| Verdict | Count | Rows |
|---|---|---|
| CONFIRMED | 23 | A-O1, A-O2, A-L7, A-L8, A-L9, A-L10, A-L11, A-L14, A-L16, A-L18, A-L19, A-P1, A-P2, A-P4, A-P7, A-P9, A-P13, A-P14, A-P15, A-P17, A-P18, A-P19, A-P20 |
| CORRECTED | 12 | A-L1, A-L2, A-L3, A-L4, A-L5, A-L12, A-L13, A-P3, A-P6, A-P8, A-P10, A-P11 |
| NOT IN SOURCE | 3 | A-L15 ("often … re-uploaded"), A-P5, A-P12 |
| NEWER DATA | 2 | A-L6, A-P16 |
| BLOCKED | 1 | A-L17 (europepmc.org link; record confirmed via the REST API) |
| UNREACHABLE | 0 | (openaccess.nhh.no DNS failure was a supporting fetch, routed around) |
| UNFETCHABLE | 0 | — |
Grade labels (§4): 7 rows, all CONFIRMED; one relabel made, in A-L12 (Wu research, Bartoš preprint).
After the edits: python3 tools/build.py then python3 tools/audit.py --quiet → "0 problems"; python3 tools/links.py → "34 pages, 346 external URLs, 0 blocked, 0 problems".
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