How to Rank in ChatGPT Search: 5 Steps That Work in 2026
Last verified: 2026-09-24. ChatGPT search does not publish a ranking algorithm, so this guide only contains mechanisms that are documented, observable, or measurable — plus what we verify weekly on our own domain.
How ChatGPT search discovers content
Section titled “How ChatGPT search discovers content”ChatGPT search combines real-time retrieval with its model knowledge. The discovery side is crawlable and observable:
- OAI-SearchBot — OpenAI’s search crawler. If it cannot fetch your page, ChatGPT search cannot cite the live version of it. OpenAI documents separate user agents for search (
OAI-SearchBot), training (GPTBot), and user-triggered fetches (ChatGPT-User). - Bing’s index — Bing is one of the discovery surfaces feeding ChatGPT search. Sites that Bing indexes quickly tend to surface in ChatGPT answers sooner. This is observable: notify Bing and watch.
- Third-party sources — ChatGPT answers frequently cite review sites, directories, and comparison pages rather than product homepages. Your own page is one candidate source among many.
ChatGPT search ranking factors
Section titled “ChatGPT search ranking factors”OpenAI publishes no ranking specification, so anyone claiming a definitive list is guessing. What follows is our working model, built from watching which pages get cited across our fixed query sets — treat it as engineering notes, not gospel.
- Being in Bing’s index. ChatGPT search is served by Bing’s index. If Bing has not indexed a page, ChatGPT cannot cite it. This is the most common failure we see, and the cheapest to fix.
- Allowing
OAI-SearchBot. The search crawler is separate from the training crawler (GPTBot). Blocking both because you “don’t want AI training on your content” also removes you from search answers. - Citable content. Pages get cited when they contain passages that answer a question standing alone — a definition, a number, a step-by-step. Marketing copy that requires context converts visitors but gives the model nothing to quote.
- Structured data. FAQ, HowTo and Article markup make the citable passages explicit. It is not a ranking boost in the classic sense; it is reducing the model’s parsing risk.
- Authority you don’t control. Mentions and links from sites ChatGPT already trusts. For newer domains this is the slowest factor — borrow it (directories, communities, guest posts) while you build it.
- Freshness. Visible dates and real updates. Answers favor pages that look maintained over pages that look published-and-forgotten.
None of these guarantee a citation. All of them are verifiable in an afternoon, which is more than can be said for most SEO advice about AI search.
Step 1 — Allow the search crawler (not necessarily the training one)
Section titled “Step 1 — Allow the search crawler (not necessarily the training one)”In robots.txt:
User-agent: OAI-SearchBotAllow: /OAI-SearchBot (search citations) and GPTBot (model training) are separate tokens. Blocking GPTBot does not block search citations. Also check your CDN or WAF: bot-fighting rules can silently block search crawlers even when robots.txt allows them — verify with a real fetch, not just the file.
Step 2 — Accelerate Bing discovery
Section titled “Step 2 — Accelerate Bing discovery”- Submit your sitemap in Bing Webmaster Tools.
- Use IndexNow: host a key file at your root and POST new or updated URLs to
api.indexnow.org. An HTTP 200 means “URL received”, not “indexed” — measure the outcome separately. - E-GEO’s own site does both automatically on every deploy; the script is in the repository (
site/indexnow.sh).
How to get indexed by ChatGPT
Section titled “How to get indexed by ChatGPT”“Indexed by ChatGPT” really means two things, and conflating them is where most setups go wrong: being fetchable by OpenAI’s crawlers, and being present in the Bing index that ChatGPT search draws from. The mechanics are Steps 1 and 2 above; as a checklist:
- robots.txt allows the search crawler — and you can treat search and training independently:
# Allow ChatGPT search citationsUser-agent: OAI-SearchBotAllow: /
# Optionally decline training useUser-agent: GPTBotDisallow: /Also leave ChatGPT-User allowed — that is the fetcher used when someone asks ChatGPT about your page directly in a conversation.
- Bing has the URL — sitemap submitted, IndexNow ping sent, and the URL actually indexed (check in Bing Webmaster Tools, not just Google Search Console). A page can rank in Google and still be invisible to ChatGPT search because Bing never picked it up.
- Verified before optimizing —
OAI-SearchBotnot blocked, URL indexed in Bing, page returning 200 without a login wall. Ten minutes of verification beats a week of rewriting pages that ChatGPT was never able to see.
On naming: SearchGPT was the prototype; the shipped product is ChatGPT search, and the infrastructure carried over — advice written for SearchGPT still applies. On llms.txt: it is a proposal, not an adopted standard, and no major AI search engine has committed to reading it; we ship one because it costs nothing (see what is llms.txt), but it replaces none of the steps above.
Step 3 — Make pages citable
Section titled “Step 3 — Make pages citable”Answer engines lift self-contained passages:
- Open each page with a direct 50–170 word answer to the query it targets.
- Use descriptive headings, tables with dates, and FAQ sections with
FAQPageJSON-LD. - Add
Article/TechArticleschema withdatePublishedanddateModified— freshness is a selection signal. - Keep one canonical entity description identical across your site, README, and listings.
Step 4 — Borrow authority you don’t have yet
Section titled “Step 4 — Borrow authority you don’t have yet”ChatGPT often cites whoever already ranks for the question. Getting referenced by those sources (directories, curated lists, niche blogs) moves you into the candidate set faster than waiting for direct citation.
Step 5 — Measure, don’t assume
Section titled “Step 5 — Measure, don’t assume”To track rankings in ChatGPT over time instead of trusting anecdotes:
- Ask ChatGPT a fixed set of category questions weekly and record whether you appear and which sources get cited.
- Watch Bing Webmaster Tools and your server logs for
OAI-SearchBotfetches. - Treat any single answer as a sample, not ground truth — answers vary run to run.
Does llms.txt help with ChatGPT search? No evidence that any major AI search engine reads it yet. Keep one if it’s free to maintain, but crawler access and Bing indexation do the real work.
How long until ChatGPT cites my site? Bing indexation can take days to weeks for a newer domain; citations follow indexation. If you’re indexed and still uncited after a month, the problem is content citability or authority, not plumbing.
Is ChatGPT search the same as SearchGPT?
SearchGPT was the prototype name; the shipped product is ChatGPT search. Advice written for SearchGPT applies — the infrastructure (OAI-SearchBot, Bing’s index) carried over.
What not to trust
Section titled “What not to trust”- No tool can guarantee a ChatGPT citation.
- Blocking AI training crawlers is a rights choice; it is not a search-visibility strategy.
- “AI SEO” services promising guaranteed placement are selling something that does not exist.
In short: allow OAI-SearchBot, get Bing to index you fast (sitemap + IndexNow), publish citable dated content with schema, earn references from the sources ChatGPT already cites, and measure with a fixed query set. E-GEO automates the content and schema side of this — the analyze → rewrite → JSON-LD pipeline plus a monitoring loop — open source, MIT: github.com/mverab/eGEOagents. Optimizing for Perplexity too? See how to rank in Perplexity.