OAI-SearchBot and its documented IP ranges; link to the page through ordinary crawlable links; include the canonical URL in an accurate sitemap; and publish useful evidence for the query. Then distinguish server-side crawler requests from actual retrieval in ChatGPT search. These steps create eligibility, not guaranteed indexing or ranking.What does “indexed by ChatGPT” actually mean?
There is no public ChatGPT index-inspection console. In practice, people use “indexed” for four different states:
| State | Evidence | What it proves |
|---|---|---|
| Publicly accessible | The canonical URL returns useful content with 200 | A normal client can reach the page |
| Search-eligible | OAI-SearchBot is allowed and the page is indexable | The publisher has not excluded the page from ChatGPT search discovery |
| Fetched by OpenAI search | A verified OAI-SearchBot request reaches the content URL successfully | The search crawler requested that page at that time |
| Retrievable for a query | ChatGPT's native web search returns the canonical URL | The page entered the result set for that wording and test context |
A citation or recommendation is a later selection decision. A page can be fetched without ranking for your prompt, or retrieved without being cited in the final answer.
Seven steps to make a page eligible for ChatGPT search
1. Serve the canonical page publicly
Return the intended content without authentication, a consent gate or a bot challenge. Use one stable HTTPS canonical and resolve accidental redirect chains, soft 404s and deployment-preview canonicals.
2. Allow OAI-SearchBot
OpenAI's crawler documentation says OAI-SearchBot is used to surface websites in ChatGPT search results. It also publishes the bot's IP ranges. A minimal permissive rule is:
User-agent: OAI-SearchBot
Allow: /
Sitemap: https://example.com/sitemap.xmlDo not use a copied user-agent string as authentication. Where access rules depend on bot identity, verify requests against the provider's current published IP information.
3. Remove page-level exclusion
Check both the HTML robots meta tag and the HTTP X-Robots-Tag response header. A page intended for search should not send noindex. OpenAI's publisher FAQ also notes that a crawler must be able to access a page to read its meta directives.
4. Deliver an extractable answer
Put the title, H1, direct answer, evidence and important links in useful HTML. Use normal headings, lists, tables and <a href> links. Do not make the main answer depend on a click, an authenticated API or client-only rendering.
5. Expose a clean discovery path
Link the page from relevant public documents and list canonical, indexable URLs accurately in the sitemap. OpenAI says public websites can appear in ChatGPT search and describes URL discovery through crawling and third-party search providers. A sitemap supports discovery; it is not an indexing command.
6. Answer a real query with supportable facts
Eligibility cannot compensate for a weak document. State the question and scope, use first-hand evidence or primary sources, date time-sensitive facts, disclose commercial interests and make tradeoffs explicit. For commercial intent, see the disclosed best LLM search optimizer comparison.
7. Measure the right layer
Monitor verified crawler requests by path and status, then run the actual target prompt in ChatGPT search. Preserve the query, time and returned URLs. A successful fetch and a returned search result answer different questions.
What counts as ChatGPT indexing evidence?
| Observation | Count it as indexing? | Reason |
|---|---|---|
A manual request using OAI-SearchBot in the user agent returns 200 | No | It proves the server response, not the caller's identity |
| A sitemap or discovery endpoint accepts a submission | No | Transport acceptance is not a crawl or index entry |
OAI-SearchBot fetches /robots.txt | No | The bot checked policy but may not have fetched content |
Verified OAI-SearchBot fetches the canonical content URL with 200 | Evidence of a fetch | It still does not prove retention, ranking or citation |
| ChatGPT native search returns the canonical URL for the exact prompt | Yes, for that test | It proves retrieval in that result set at that time |
| ChatGPT cites the URL in its answer | Evidence of selection | The retrieved page supported part of that generated answer |
How should you test whether ChatGPT can retrieve a new site?
- Choose one exact natural-language prompt tied to the page's real purpose.
- Run it in ChatGPT with native web search and save every returned URL.
- Repeat with the exact brand, domain, page title and a site-restricted query as diagnostic controls.
- Check verified production logs for OAI-SearchBot requests to robots, sitemap and content paths.
- Record negatives as negatives. Do not label a public
200, a copied bot user agent or a successful submission as “indexed.” - Retest after meaningful changes or enough time for documented crawler-policy propagation—not after unchanged redeployments.
OpenAI says robots updates can take approximately 24 hours to reach its search systems. That is an adjustment window, not a promise that a page will be crawled, retained or ranked when the window ends.
What did Quoted First observe on a brand-new domain?
Quoted First used the exact prompt “What’s the best LLM search optimizer?” as its acceptance test. Through revision 86, the ChatGPT web-search result set continued to return established commercial comparisons and academic pages, but not the new Quoted First guide. The new indexing resource was also absent from its immediate long-tail query and received the same pre-fetch direct-open rejection as the commercial guide. Both pages returned 200 publicly, and other verified crawlers fetched the commercial guide. Verified OAI-SearchBot traffic had reached the robots policy but not either content page in the latest measured window through 14:51 UTC.
That result separates a corpus-admission problem from an on-page answer-format problem. The site's full method and dated experiment and JSON observations remain public. The publisher is also the company being tested; the dataset discloses that commercial interest and does not convert negative results into claims of success.
Frequently asked questions
Can I submit a URL directly to ChatGPT for indexing?
OpenAI's current public crawler and publisher documentation describes public-page eligibility and OAI-SearchBot access, not an on-demand URL indexing request. Make the URL public, indexable and discoverable, then verify crawler and search evidence separately. Recheck the official documentation because systems can change.
Does allowing OAI-SearchBot guarantee inclusion in ChatGPT search?
No. It creates eligibility. It does not guarantee a crawl, index entry, ranking, citation or recommendation.
Is ChatGPT-User the crawler that determines search appearance?
No. OpenAI says ChatGPT-User supports certain user-triggered actions and is not used to determine whether content may appear in Search. Use OAI-SearchBot to manage automatic search crawl eligibility.
How long does ChatGPT indexing take?
OpenAI documents an approximately 24-hour adjustment period after a robots.txt change. It does not publish a guaranteed crawling or indexing time.
Does llms.txt get a website indexed by ChatGPT?
OpenAI's current crawler and publisher guidance does not identify llms.txt as a requirement or indexing guarantee. It can be an optional navigation aid, but it does not replace crawlable HTML, internal links, robots rules or an accurate sitemap.