1. Primary sources first
A claim about dots should trace to an OpenAI publication — the launch post, the safety write-up, the help center, a release note, or the model page — or to another source we name. Screenshots of social posts are not sufficient on their own.
2. Date everything
Facts get a “last updated” date at the page level and a verification date in our source index. If we cannot say when something was true, we do not present it as true.
3. Label the gaps
Where OpenAI has published nothing — credits for a second dot, add-on pricing, a full list of eligible markets, whether dots come to the EEA and UK — we write “not confirmed” and explain what would confirm it. Guessing is not a style choice; it is a correctness failure.
4. Separate reporting from analysis
Inside an article, a paragraph that is our judgement says so. We quote OpenAI's wording verbatim where the exact phrasing matters, including their own qualifiers such as “our most capable model” or “may still make mistakes”, and attribute vendor-reported benchmark numbers as vendor-reported.
5. Correct in public
Substantive corrections update the page, change the “last updated” date, and are noted in the page's change note where the error could have misled a reader. We do not delete and repost to hide a correction.
6. No pay-for-coverage
We do not accept payment for coverage, for a positive assessment, or for a link in our editorial recommendations. Display advertising is served programmatically and labelled; it has no influence on what we publish, and no advertiser sees an article before publication.
7. Independence from AI tooling
This site covers AI agents, so we say this plainly: research and drafting here make use of AI tooling, but every published claim is checked against the cited source by a human editor, who is accountable for it. If you find an error, that editor got it wrong — not the tool.
See also: source index, disclaimer, and contact.