| OAI-SearchBot search eligibility |
If a public site wants a chance to appear in ChatGPT search features, review robots.txt so OAI-SearchBot is not accidentally blocked. |
Check robots.txt, avoid broad user-agent blocks that catch OAI-SearchBot, then recheck after publishing. |
Allowed crawling does not guarantee ranking, citations, clicks, or traffic. |
| GPTBot training-use policy |
No. GPTBot is a training-use crawler signal. It should be reviewed separately from OAI-SearchBot so training policy and search visibility policy do not get mixed. |
Document whether GPTBot is allowed or disallowed based on training-use preference. |
GPTBot log hits are not proof that a page appears in ChatGPT Search. |
| ChatGPT-User user-triggered visits |
No. ChatGPT-User is for user-triggered actions and should not be treated as automatic search crawling or Search appearance proof. |
Separate ChatGPT-User log events from OAI-SearchBot crawler checks and from human traffic reporting. |
A ChatGPT-User visit is not a ranking guarantee or automatic crawler hit. |
| Robots.txt, sitemap, and public discovery files |
Start with public discovery files: homepage, robots.txt, sitemap.xml, and crawlable target pages. If these fail, AI search visibility work is premature. |
Run the live report and fix 404, 5xx, accidental Disallow, missing Sitemap, or private-path exposure issues. |
A sitemap submission or robots.txt edit is a discovery hint, not traffic proof. |
| llms.txt context map |
Treat llms.txt as a concise public context map. It can make important pages easier to inspect, but it is not a ranking or citation guarantee. |
Publish a short root /llms.txt file with public URLs only, then validate it. |
A generated llms.txt draft is not proof that the file is live or useful. |
| Answer pages and proof links |
Use one-query answer pages, answer packs, proof lookup rows, and citation snippets so assistants can get a concise answer before scanning the whole site. |
Expose proof-linked answer packs and include at least one proof link with every claim. |
A broad AI SEO article without specific proof links is weaker than a short answer with verifiable sources. |
| Measurement, not fake traffic |
Use Search Console clicks, qualified referrals, sessions, conversions, or tool activations. Keep impressions, crawler hits, fake searches, and self-clicks out of traffic claims. |
Measure query-level clicks and tool events after changes; do not manipulate searches or engagement. |
No fake searches, no self-clicks, and no crawler hits counted as human traffic. |