ChatGPT Search visibility: OAI-SearchBot, robots.txt and what website owners should check
A technical checklist for making public website content accessible to ChatGPT Search without confusing search access with model-training permissions.
AI-search visibility starts with access. A public page cannot become a useful source if the relevant search crawler cannot fetch it reliably. Website owners should review robots.txt, CDN and firewall rules, response codes, canonical URLs and whether important pages are actually reachable without client-side obstacles.
OpenAI separates search crawling from model-training preferences. That distinction matters operationally: a business can manage search discovery and training permissions independently rather than treating every AI crawler as the same thing.
Technical access is only eligibility, not a ranking guarantee. Clear service definitions, original evidence, strong entity consistency, useful answers, internal links and trustworthy source pages still determine whether a page is useful enough to surface.
Measurement should include more than generic organic traffic. Referral traffic, branded searches, landing pages reached from AI-assisted discovery and the quality of resulting enquiries are more useful commercial signals.
For multilingual sites, crawler access must be consistent across language folders and alternate URLs. Hreflang, canonical rules and localised content should agree rather than sending conflicting signals.