Records · Verification table
AI Do page: "What happens after" section, verification, 6 November 2026
The working record behind the pages, published as written. Source file source/verify/ai-after-2026-11-06.md. All records.
Scope. Every factual sentence in the new <h2 id="after">What happens after</h2> section of content/ai-do.html and the four new source entries s24 to s27. Nothing else on the page was changed.
Method. Sources fetched on 2026-11-06 between 10:49 and 10:54 EST (date stamps 10:49:59, 10:51:27, 10:54:47). Default: curl -sL --compressed with a browser user agent, HTML stripped to text; PDFs read with pdftotext -layout; abstracts through the Europe PMC REST API and OpenAlex. Pages that refused curl were tried in headless Chrome ([local working file]). Two web searches were used only to find URLs (the Gemini verification help page, the 988 pages, a study of state legislative offices); no sentence rests on a search summary. Nothing from memory.
Verdicts. CONFIRMED: the source's words support the sentence as written. LEFT OUT: the source could not be read, so the item is not on the page.
Fetch log (all 2026-11-06).
- privacy.claude.com, "Is my data used for model training?" (dated March 16, 2026) and "How long do you store my data?" (dated July 1, 2026): 200.
- support.google.com/gemini/answer/13594961 (Gemini Apps privacy hub, "Last updated: October 26, 2026"): 200.
- help.openai.com/en/articles/7730893-data-controls-in-chatgpt: curl 403; headless Chrome returned only "Just a moment..." (Cloudflare). UNREADABLE; ChatGPT left out.
- support.google.com/gemini/answer/16722517 (Verify AI-generated images, videos, and audio): 200.
- deepmind.google/technologies/synthid/ (redirects to /models/synthid/): 200 (first fetch came back gzip-compressed; refetched with
--compressed). - contentauthenticity.org/how-it-works and c2pa.org/faqs/: 200. contentcredentials.org/verify: 200 but a JavaScript app with no text; not used.
- aeaweb.org, Allcott et al. 2020 article page: 200 (abstract). academic.oup.com (Castelo et al.): curl 403, headless Chrome Cloudflare check; read instead as full text from Europe PMC, PMC11834938 (fullTextXML 200). Allcott et al. 2024 (PMC11126999): abstract via Europe PMC API, 200.
- cep.lse.ac.uk/pubs/download/dp1350.pdf (Beland and Murphy): 200, PDF. OpenAlex, doi:10.3368/jhr.0224-13403r2 (Abrahamsson): 200, abstract.
- waituntil8th.org (home) and /faqs: 200.
- faunalytics.org, ASPCA Legislative Alerts Study Interview Analysis PDF: 200 (pdfinfo CreationDate Sun Dec 12 2004; the text carries no date).
- 988lifeline.org: /get-help/what-to-expect/, /confidentiality/, /privacy-policy/, FAQ "Does the 988 Lifeline trace callers, texters, or chatters?", FAQ on call recording, FAQ on "confidential", FAQ on geolocation: all 200.
Working files are in the session scratchpad, [local working file]. They are not kept in the repo.
Rows
| # | Claim on the page | URL | Fetch | Exact quote | Verdict |
|---|---|---|---|---|---|
| A1 | Claude: Anthropic "won't use your past or new chats to train future models" after training is turned off | https://privacy.claude.com/en/articles/10023548-how-long-do-you-store-my-data | 200 (curl) | "If you decide to turn off the model improvement setting, we will not use your previous or new chats or coding sessions for future model training." | CONFIRMED |
| A2 | "Chats already in a training run, or in a model already trained, stay there." | same | 200 | "Your data will still be included in model training runs that are already in progress, or in models that have been trained." | CONFIRMED |
| A3 | "A chat you delete leaves its servers within 30 days." | same | 200 | "When you delete a conversation it's: Removed from your chat history immediately; Deleted from our back-end storage systems within 30 days" | CONFIRMED |
| A4 | "Chats its systems flag for breaking its rules are kept up to two years." | same | 200 | "We retain inputs and outputs for up to 2 years and trust and safety classification scores for up to 7 years if your chat or session is flagged by our automated trust and safety systems as violating our Usage Policy." | CONFIRMED |
| A5 | Gemini: with Keep Activity off, future chats aren't used for training unless you send feedback | https://support.google.com/gemini/answer/13594961 | 200 (curl) | "Your future chats won't appear in your Activity, and won't be used to train our AI models, unless you choose to send Google feedback." | CONFIRMED |
| A6 | "Each chat is still kept for 72 hours." | same | 200 | "Future chats are still saved for 72 hours so Gemini can respond to you, process your feedback, and protect Google, its users, and the public." | CONFIRMED |
| A7 | "Chats human reviewers already saw are kept up to three years, even if you delete your activity." | same | 200 | "If you delete your activity, past chats already reviewed by service providers aren't deleted because they aren't connected to your Google Account. Instead, they're retained for up to 3 years." | CONFIRMED |
| A8 | "Some features stop working while it's off, such as Google Workspace apps." | same | 200 | "At this time, other Connected Apps like Google Workspace are unavailable if you turn off Keep Activity." | CONFIRMED |
| A9 | "OpenAI's help page refused our reader, so we can't say what ChatGPT keeps." | https://help.openai.com/en/articles/7730893-data-controls-in-chatgpt | curl 403; headless Chrome "Just a moment..." | (none) | LEFT OUT (the ChatGPT item is not on the page; the sentence states the refusal) |
| B1 | A SynthID watermark "means Google's AI made or edited all or part of it" | https://support.google.com/gemini/answer/16722517 | 200 (curl) | "If a SynthID watermark is detected, it means all or part of the image or video was created or edited by Google's AI models." | CONFIRMED |
| B2 | "No watermark means Google's AI didn't make it; another company's AI still could have." | same | 200 | "If a SynthID watermark isn't detected, it means the image or video wasn't created or edited by Google AI, but it could have been created by other AI systems." | CONFIRMED |
| B3 | "Sometimes the answer is unclear, as with very simple images or very small edits." | same | 200 | "In some cases it isn't clear if the image or video was created or edited by Google AI or not. Some common reasons for this are: \"Not enough details to watermark\" (e.g., very simple or abstract content). \"Likely too small an edit\"" | CONFIRMED |
| B4 | "You get about 10 image checks in any 24 hours." | same | 200 | "There's a usage quota of approximately 10 image checks, 10 video checks (up to 5 minutes total), and 10 audio checks (up to 3 hours total) within any 24-hour rolling window." | CONFIRMED |
| C1 | Content Credentials: "you see which company signed the information and the history it records" | https://support.google.com/gemini/answer/16722517 ; https://contentauthenticity.org/how-it-works | 200; 200 | Gemini help: "Show the company that provided and cryptographically signed the information." CAI: "Content Credentials provide information about the origin and history of a piece of content" | CONFIRMED |
| C2 | "If the file or its record changed after signing, the check shows that." | https://contentauthenticity.org/how-it-works | 200 | "When information in Content Credentials has been improperly changed, or when the associated content has been changed after Content Credentials were applied, the Content Credentials will identify and display this information." | CONFIRMED |
| C3 | "It doesn't tell you whether the picture is real" | same; Gemini help | 200 | CAI: "they aren't intended to prescriptively indicate whether a piece of content is \"real.\"" Gemini help, "What Content Credentials don't do": "Tell you if an image, video, or audio file is definitely AI-generated" | CONFIRMED |
| C4 | "and the name shown may not be the owner's" | https://contentauthenticity.org/how-it-works | 200 | "The name displayed in Content Credentials generated with the CAI's tools might not always be the same as the individual or organization that owns the content." | CONFIRMED |
| D1 | People who left Facebook for four weeks in 2018 "reported more wellbeing, knew less factual news and were less polarised" | https://www.aeaweb.org/articles?id=10.1257/aer.20190658 | 200 (abstract) | "deactivating Facebook for the four weeks before the 2018 US midterm election … (ii) reduced both factual news knowledge and political polarization; (iii) increased subjective well-being" | CONFIRMED |
| D2 | "Afterwards they used Facebook a lot less, the drop lasted, and they valued it less." | same | 200 | "(iv) caused a large persistent reduction in post-experiment Facebook use. Deactivation reduced post-experiment valuations of Facebook" | CONFIRMED |
| D3 | "In a bigger six-week study before the 2020 election, the effect on polarisation was close to zero, and people may have known less general news." | Europe PMC, PMC11126999 (Allcott et al., PNAS 2024) | 200 (API abstract) | "randomizing a subset of 19,857 Facebook users and 15,585 Instagram users to deactivate their accounts for 6 wk before the 2020 U.S. election"; "Facebook deactivation had no significant effect on an index of knowledge, but secondary analyses suggest that it reduced knowledge of general news"; "the effects … on affective and issue polarization … were all precisely estimated and close to zero." | CONFIRMED (contest kept: 2018 study found less polarisation, 2020 study near zero) |
| D4 | Mobile-internet study: "wellbeing and mental health were still better two weeks after the block ended, and screen time stayed lower" | Europe PMC fullTextXML, PMC11834938 (Castelo et al., PNAS Nexus 2025) | 200 | "All measures were taken three times: at baseline (T1), 2 weeks later (T2), and another 2 weeks later (T3)." "in the Intervention group, SWB and mental health remained significantly higher at T3 than they were at T1. These persistent benefits may be attributable to the fact that screen time in the Intervention group remained lower at T3 than at T1, despite the fact that we did not require participants in this group to continue blocking mobile internet after T2." | CONFIRMED |
| D5 | "Only a quarter of the people who signed up kept the block on for ten of the fourteen days." | same | 200 | "of the 467 who committed to blocking mobile internet for 2 weeks, 266 set up the app required to do so and 119 (25.5% of those who committed) met our preregistered definition of \"compliant\" (having the block active for at least 10 of the 14 intervention days" | CONFIRMED |
| E1 | Wait Until 8th: "nothing is made public" | https://www.waituntil8th.org/faqs | 200 (curl) | "Your pledge is not posted anywhere or made known publicly." "Your personal information will not be made public on this site." | CONFIRMED |
| E2 | "Within one to two business days you get an email with tips for recruiting other parents." | same | 200 | "After you sign the pledge, you will receive an email from Wait Until 8th in one to two business days with tips and suggestions on how to rally other parents from your child's school to sign the pledge." | CONFIRMED |
| E3 | Once ten families from the grade and school sign, the pledge is active; one group email with the list of families and parents' emails; then updates when more join | same | 200 | "Once 10 parents from your child's grade and school sign the pledge, Wait Until 8th will send a group email to the families who have signed. … You will receive a list of families who are delaying from your child's grade and the emails for the parents. Then moving forward, Wait Until 8th will notify the group of any new additions to the pledge." | CONFIRMED |
| E4 | "Until then your pledge stays anonymous." | same | 200 | "The pledge is anonymous until the pledge becomes active." | CONFIRMED |
| F1 | Phone-free school: "In England, schools that enforced it weakly saw no significant change in test scores." | https://cep.lse.ac.uk/pubs/download/dp1350.pdf | 200 (PDF) | "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." | CONFIRMED |
| F2 | "In Norway, bans changed no averages but cut bullying." | OpenAlex, doi:10.3368/jhr.0224-13403r2 | 200 (abstract) | "banning smartphones in middle schools has no average effect on education or mental health but masks important gender differences. Bans … lower bullying for both genders" | CONFIRMED |
| G1 | "In interviews with 77 state legislative offices in ten states" | https://faunalytics.org/wp-content/uploads/2014/08/Faunalytics-ASPCA-Legislative-Alerts-Study-Interview-Analysis.pdf | 200 (PDF) | "HRC conducted 77 phone interviews with legislative offices in ten states." | CONFIRMED |
| G2 | "constituents came first, and many offices, though not most, threw away messages from outside the district" | same | 200 | "Legislators clearly place the highest priority on messages from constituents." "In many legislative offices, but not the majority, communications from non-constituents are immediately discarded." | CONFIRMED |
| G3 | "Roughly half the states kept tallies; most offices did so only for major bills." | same | 200 | "Roughly half of states say they do keep tallies, but probably the majority of offices overall say they only do so for major issues or bills." | CONFIRMED (the source's "probably" is softened into "most"; the page labels the study's age and funder) |
| G4 | "Most offices were annoyed by form letters but often counted them the same as personal ones if a constituent sent them." | same | 200 | "The vast majority of interviewees expressed significant annoyance with form letters, form emails, and blast faxes. However, in a surprising number of cases these messages were counted and weighed equally to personal messages, as long as they came from constituents." | CONFIRMED |
| G5 | "The study is undated (its file is from 2004) and was done for the ASPCA." | same | 200 | Title: "ASPCA LEGISLATIVE ALERTS STUDY – INTERVIEW ANALYSIS"; no date in the text; pdfinfo: "CreationDate: Sun Dec 12 20:50:37 2004 EST". | CONFIRMED (date from file metadata only) |
| G6 | "No study we read says how fast state offices reply." | the sources read this run | — | The Faunalytics study gives no reply times; the CMF reports cited elsewhere on the site cover Congress. | CONFIRMED (a statement about this run's reading) |
| H1 | 988: "a live counselor says hello, asks whether you're safe, then listens and shares resources" | https://988lifeline.org/get-help/what-to-expect/ | 200 (curl) | "A counselor will say hello and introduce themselves. Your skilled counselor will ask you if you are safe. After they ask about your safety, your counselor will listen to you, understand how your problem is affecting you, provide support, and share any resources that may be helpful." | CONFIRMED |
| H2 | "988 tries to route calls to the nearest center by area code." | https://988lifeline.org/faq/about-us/faq-does-the-988-lifeline-trace-callers-texters-or-chatters/ | 200 | "The 988 Lifeline does attempt to route callers to the nearest center to them in the network using their area code to determine the most proximal center." | CONFIRMED |
| H3 | "You don't have to say who you are or where you are." | https://988lifeline.org/confidentiality/ | 200 | "You do not have to share any personal information with the 988 Lifeline counselors about who you are or where you are to receive support." | CONFIRMED |
| H4 | "Confidential isn't anonymous: the counselor takes notes, a center may record the call, and centers keep records at least three years." | confidentiality statement; recording FAQ | 200; 200 | "\"Confidential\" and \"anonymous\" are not considered to be the same"; "When you are connected to a 988 Lifeline counselor, that person will take notes on the conversation."; "Crisis centers in the 988 Lifeline network may independently utilize call recordings for quality assurance and training purposes"; "The crisis contact centers must maintain all contact records for a minimum period of three years." | CONFIRMED |
| H5 | "Your details can be used to arrange further care, to look into complaints or deaths, for court orders and, in some states, to bill insurance." | https://988lifeline.org/confidentiality/ | 200 | "SAMHSA, Vibrant, and crisis contact centers use this information to investigate deaths after someone has contacted the 988 Lifeline, to investigate grievances, and in response to court orders. States and crisis contact centers may use this information to coordinate care required beyond your call, text, or chat. In some states, your personal information is used to submit claims to government and private insurance companies." | CONFIRMED |
| H6 | "Otherwise nothing is shared without your consent unless the law requires it or someone is at imminent risk of harm." | same | 200 | "Sharing of information for any other reason is prohibited unless there is imminent risk of harm to you or to someone else, or the sharing of information is required by law or we have your informed consent." | CONFIRMED |
| H7 | "988 can't trace you the way 911 can." | trace FAQ | 200 | "The 988 Lifeline does not currently have the capability to directly \"trace\" callers, chat, or text users in the same way that 911 providers do." | CONFIRMED |
| H8 | "In the rare case emergency services must be called and you can't or won't say where you are, the counselor gives 911 your phone number, or your IP address in a chat." | trace FAQ | 200 | "In the atypical situations where emergency services must be contacted to prevent persons from seriously or fatally harming themselves, and the person is unwilling or unable to share their location information, 988 Lifeline crisis counselors must provide what information they have to 911 operators–the caller's/text user's phone number or the chat user's IP address" | CONFIRMED |
New sources and grades
| id | Source | Grade | Why |
|---|---|---|---|
| s24 | Google, Gemini Apps Help, Verify AI-generated images, videos, and audio | primary | the provider's own document |
| s25 | Wait Until 8th, FAQs | charity's own figures | an organisation describing its own process |
| s26 | HRC for the ASPCA, interview analysis (Faunalytics PDF) | industry research | a commissioned study by a research firm, not peer-reviewed |
| s27 | 988 Lifeline: What to expect, Confidentiality statement, tracing FAQ, recording FAQ | primary | the service's own documents |
Re-used, re-read today: s7 (Anthropic and Google pages), s9 (CAI How it works), s6 (Allcott 2020, Allcott 2024, Castelo 2025), s17 (Beland and Murphy; Abrahamsson).
Counts
| Verdict | Rows |
|---|---|
| CONFIRMED | 40 (A1–A8, B1–B4, C1–C4, D1–D5, E1–E4, F1, F2, G1–G6, H1–H8) |
| LEFT OUT | 1 (A9, ChatGPT: OpenAI's help page unreadable) |
Found while reading, not changed (outside this section)
- Google DeepMind's SynthID page (read today) says: "With the new SynthID Detector, anyone can easily check if an image, video, or audio file was made with AI from Google or our partners, including OpenAI, NVIDIA, Kakao, and soon, Apple." The Gemini help page (read today) says: "While other companies have started to adopt SynthID watermarks, Gemini can currently only recognize content created by Google AI tools." The page's existing "Ask Gemini whether Google made it" line ("other companies' AI output won't be flagged") matches the Gemini help page for the Gemini check; the separate Detector may now read partners' watermarks. Flagged for the next AI review; not edited here.
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