ChatGPT Search Eligibility Audit: Practical Checklist

ChatGPT Search Eligibility Audit: A Practical Checklist

A website can be technically healthy in Google and still be inaccessible to ChatGPT Search. The reason is simple: ChatGPT Search has its own crawler, OAI-SearchBot, and your robots.txt, CDN, WAF, firewall, or bot-protection system can block it independently of Googlebot.

OpenAI’s current guidance identifies two core technical requirements for search eligibility: allow OAI-SearchBot to crawl the site, and make sure your hosting or CDN does not block traffic from OpenAI’s published SearchBot IP ranges. Passing those checks makes a site eligible to appear; it does not guarantee that ChatGPT will select or rank it for a particular query. (help.openai.com)

Direct Answer

To audit a website for ChatGPT Search eligibility, verify that:

1. the content is publicly reachable;
2. OAI-SearchBot is not blocked by robots.txt;
3. your CDN, firewall, WAF, or bot protection accepts legitimate OAI-SearchBot traffic from OpenAI’s published IP ranges;
4. pages return usable HTTP responses rather than authentication, CAPTCHA, challenge, 403, or rate-limit barriers;
5. you have not confused GPTBot controls with OAI-SearchBot controls.

OpenAI says search placement uses multiple factors related to relevance and reliability, and placement is not guaranteed. Eligibility is therefore the first technical gate, not a ranking strategy. (help.openai.com)

Eligibility Checklist

Check OAI-SearchBot

OpenAI identifies OAI-SearchBot as the crawler used to surface websites in ChatGPT search features. Its documentation explicitly recommends allowing the crawler in robots.txt and allowing requests from OpenAI’s published SearchBot IP ranges. (developers.openai.com)

A minimal configuration is:

User-agent: OAI-SearchBot
Allow: /

If you want ChatGPT Search access but do not want your site crawled by GPTBot for potential model training, the two controls can be separated:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

OpenAI documents these settings as independent. Allowing Search does not require allowing GPTBot. (developers.openai.com)

This distinction matters because GPTBot and OAI-SearchBot have different purposes. GPTBot relates to crawling that may be used for training OpenAI’s generative AI foundation models. OAI-SearchBot controls automatic crawling for Search. ChatGPT-User, meanwhile, represents certain user-triggered visits and is not the crawler used to determine Search inclusion. (developers.openai.com)

INTERNAL LINK: OAI-SearchBot Explained

Audit robots.txt Carefully

Do not only search your robots.txt file for the literal text OAI-SearchBot. Evaluate the rules that actually apply to it.

The Robots Exclusion Protocol defines crawler rules through user-agent groups and Allow or Disallow path rules. When no specific group matches a crawler, a User-agent: * group may apply. More-specific path rules take precedence over less-specific matching rules. (rfc-editor.org)

For example:

User-agent: *
Disallow: /

User-agent: OAI-SearchBot
Allow: /

A dedicated OAI-SearchBot group can therefore express a different policy from the general crawler policy.

Also verify that /robots.txt itself works reliably. RFC 9309 distinguishes an unavailable robots file from one that is unreachable because of server or network errors. A persistent 5xx or connectivity problem can therefore have very different consequences from simply having no robots file. (rfc-editor.org)

Check the Infrastructure Layer

A correct robots.txt does not prove that OpenAI can actually fetch the page.

OpenAI specifically tells publishers to confirm that their host or CDN accepts traffic from its published SearchBot IP addresses. The current ranges are maintained in OpenAI’s searchbot.json file rather than being a fixed list you should copy permanently into documentation. (help.openai.com)

This matters when a site uses Cloudflare, Akamai, a managed WAF, custom firewall rules, rate limiting, or anti-bot systems. OpenAI notes that crawler traffic can be blocked by 403 responses, automated bot mitigation, JavaScript challenges, CAPTCHAs, authentication checks, or rate limiting. (help.openai.com)

A useful server-side audit therefore checks both:

curl -I https://example.com/page

and:

curl -A "OAI-SearchBot" -I https://example.com/page

The second request can expose user-agent-based application rules, but it is not proof that genuine OpenAI traffic is allowed. Your request does not originate from an OpenAI SearchBot IP. Confirm the real infrastructure behavior through CDN/WAF configuration and logs using OpenAI’s published IP ranges.

Diagnose Failed Access

OpenAI notes that Search systems may take approximately 24 hours after a robots.txt update to adjust. That is an operational estimate, not a promise that the page will appear in search after 24 hours. (developers.openai.com)

Eligibility Is Not Ranking

A successful audit establishes crawl eligibility. It does not establish visibility.

OpenAI says ChatGPT Search ranks results using multiple factors intended to help users find relevant and reliable information, and explicitly states that placement is not guaranteed. OpenAI has not publicly documented a complete Search ranking algorithm or a fixed set of ranking-factor weights. (help.openai.com)

This means several commonly discussed tactics should not be misrepresented as formal ChatGPT Search eligibility requirements. OpenAI’s current public eligibility guidance does not list XML sitemaps, llms.txt, schema markup, GPTBot permission, or a special “AI SEO” meta tag as mandatory conditions for Search inclusion.

Those technologies may have other legitimate uses, but they should not be presented as requirements without supporting evidence.

INTERNAL LINK: ChatGPT Search vs Traditional Search

Measure Real Traffic

Publishers that allow OAI-SearchBot can also look for downstream evidence in analytics. OpenAI states that ChatGPT referral URLs automatically include:

utm_source=chatgpt.com

This makes referral traffic from ChatGPT easier to isolate in analytics platforms. (help.openai.com)

Referral traffic and crawler logs answer different questions. Logs tell you whether OpenAI can reach the site. Referral data tells you whether users actually clicked through from ChatGPT. Neither alone proves how often a page was considered, ranked, quoted, or omitted.

Common Mistakes

The most important mistake is treating ChatGPT Search access as identical to Google indexing. It is not. A website can permit Googlebot while blocking OAI-SearchBot at robots.txt or infrastructure level.

Another mistake is allowing the user agent in robots.txt while a WAF silently rejects the same crawler. The reverse can also happen: the network permits the request, but robots.txt says not to crawl the content.

Finally, blocking GPTBot does not automatically opt a site out of ChatGPT Search. OpenAI intentionally separates the training and search controls. (developers.openai.com)

SeoNest Recommendation

Treat the audit as two separate gates:

Crawler policy → infrastructure access.

First establish that OAI-SearchBot is permitted by robots.txt. Then establish that a legitimate request from OpenAI’s published SearchBot infrastructure can actually reach the page without security or application barriers.

Only after these checks pass should you investigate content relevance, authority, page quality, entity clarity, or other visibility questions. Otherwise, you may spend time “optimizing for ChatGPT” while the crawler cannot reliably access the site at all.

FAQ

Does allowing OAI-SearchBot guarantee ChatGPT visibility?

No. It establishes eligibility for crawling and inclusion, but OpenAI says placement is not guaranteed. (help.openai.com)

Do I need to allow GPTBot?

Not for ChatGPT Search eligibility. OpenAI documents GPTBot and OAI-SearchBot as independent controls. (developers.openai.com)

Can I block training but allow Search?

Yes. You can allow OAI-SearchBot while disallowing GPTBot. (developers.openai.com)

Should I hard-code SearchBot IP addresses?

Prefer OpenAI’s published searchbot.json as the source of truth because crawler infrastructure can change. (openai.com)

How quickly do robots.txt changes apply?

OpenAI says its Search systems may take approximately 24 hours to adjust after a robots.txt update. (developers.openai.com)

Final Takeaway

A ChatGPT Search eligibility audit is primarily an access audit. The essential question is not “Have we added enough AI SEO markup?” It is: Can OAI-SearchBot legally and technically reach the content?

Check robots.txt, verify OpenAI’s current SearchBot IP ranges against your infrastructure policy, inspect WAF and server responses, keep GPTBot controls separate, and measure both crawler activity and referral traffic. Once those foundations work, search visibility becomes a relevance and selection problem rather than an access problem.

Sources

1. OpenAI — “Searching the web with ChatGPT.” Updated August 2026. Eligibility guidance, ranking disclaimer, and SearchBot infrastructure requirement. OpenAI: Searching the web with ChatGPT (help.openai.com)
2. OpenAI — “Overview of OpenAI Crawlers.” Current developer documentation. Definitions and controls for OAI-SearchBot, GPTBot and ChatGPT-User; approximately 24-hour robots.txt adjustment note. OpenAI Crawlers documentation (developers.openai.com)
3. OpenAI — “Publishers and Developers – FAQ.” Updated August 2026. Publisher discovery guidance and ChatGPT referral tracking. OpenAI Publishers and Developers FAQ (help.openai.com)
4. OpenAI — SearchBot published IP ranges. Current machine-readable source for OAI-SearchBot network prefixes. OpenAI searchbot.json (openai.com)
5. OpenAI — “Advertiser Guidance for Allowing OpenAI Web Crawlers.” Updated September 20, 2026. Practical crawler troubleshooting covering WAFs, CDNs, bot protection, CAPTCHAs, 403 responses and rate limiting. OpenAI crawler troubleshooting guidance (help.openai.com)
6. IETF — RFC 9309: Robots Exclusion Protocol. September 2022. Standards-track definition of robots.txt parsing, user-agent matching, Allow / Disallow, access behavior and caching. RFC 9309 (rfc-editor.org)