How to rank in ChatGPT (2026)
ChatGPT has no position one. It rewrites your question into a few searches, reads a short list of results and cites some of them. To rank, be in those results, with the answer in the first 200 characters of your page.
- Rev.
- November 9, 2026
- Docs checked
- November 8, 2026
- Our data
- Nov 7 run, 28 ChatGPT answers
- Part of
- How to get cited by ChatGPT
Our pick
Want to see where you rank on your buyers' real questions today? supereo's free audit ours is our pick: it writes 20 questions from your site, asks them on ChatGPT and 6 engines in all, and shows every answer and its sources, in about 2 minutes, no card. When you're ready to move, Copilot tells you exactly what to change on each page. Run the free audit.
What ranking means in ChatGPT
A ChatGPT answer has two places you can rank, and they're measured differently.
| Where you appear | What it looks like | What decides it |
|---|---|---|
| Cited | Your page is one of the linked sources | Whether your page is in the results ChatGPT reads for its searches, and whether it answers them |
| Named | Your brand appears in the answer text | What the cited pages say, and what ChatGPT already knows |
| Recommended | The answer tells the reader to pick you | Whether those pages name you as the answer for the reader's situation |
OpenAI doesn't publish a ranking formula. Its help page says only that "ChatGPT ranks search results using multiple factors intended to help users find relevant, reliable information. Placement is not guaranteed." What it does publish is how ChatGPT searches, and that's where ranking starts.
How ChatGPT searches before it answers
OpenAI says that when ChatGPT uses a search partner, it "typically rewrites your query into one or more targeted queries", and gives an example: a question about drugs that target CCR8 becomes the search "CCR8 immunotherapy drug development 2025". In our November 7, 2026 run of real buyer questions, we captured ChatGPT's own searches on 18 answers, and Claude's on 16 for comparison.
| How it searched | ChatGPT | Claude |
|---|---|---|
| Searches per answer (median) | 2, up to 4 | 1, up to 4 |
| Words per search (median) | 12 | 6 |
| Searches with a year in them | 24 of 41 | 4 of 19 |
| Searches with the word "official" | 10 of 41 | 0 of 19 |
| Answers that added brand names the question didn't contain | 6 of 18 | 2 of 16 |
Our capture of each engine's search queries, November 7, 2026. ChatGPT's six brand-adding answers were all questions asking for the best tool or alternative.
Three things follow. ChatGPT writes long, specific searches, so a page whose title and opening use the same specific words matches more of them. It adds the year and "official" on its own, so dated titles and clearly official pages (your docs, your pricing) fit its searches. And it adds names it already knows, so being named on the lists in your category puts you into searches you never typed. Examples of its searches from the run, with the brand names left out:
OpenAI ChatGPT search publisher crawler OAI-SearchBot official
OpenAI official guidance how websites appear in ChatGPT search recommendations brands SEO
AEO agency pricing answer engine optimization cost monthly retainer 2026
Bing Webmaster Tools AI Performance report citations officialAcross all 28 of ChatGPT's answers in the run, it was handed a median of 16 search results and cited a median of 9.5. Ranking means being one of those 16 and then one of the cited.
Where ChatGPT's results come from
Three sources feed those results. OpenAI's own index, built by OAI-SearchBot: OpenAI says sites opted out of it "will not be shown in ChatGPT search answers". Search partners: OpenAI says ChatGPT "sometimes partners with other search providers" and cites Microsoft's and Shopify's privacy policies for them. And, per the 2024 launch post, "content provided directly by our partners".
How they mix depends on the account. Resoneo, which reverse-engineered ChatGPT's traffic in July and August 2026, found free accounts using Think drew about three quarters of their results from OpenAI's own index, while paid accounts in thinking mode drew about three quarters from Google results. Of the pages ChatGPT used, one in three ranked on Google's first page and one in twenty on Bing. So ranking in ChatGPT for paid users leans on ranking in Google for ChatGPT's own searches, and for free users on being in OpenAI's index.
What ChatGPT sees of your page
When ChatGPT reads a result list, it gets a title, a snippet and a URL per page. Resoneo measured the snippet that OpenAI's index stores: about 200 characters "taken from the start of the page's rendered body, not from your metadata. Your meta description is ignored." In instant mode, that's your whole pitch. It also found the snippet "is frozen at crawl time, and it ages": 13% of snippets ran more than a month behind the crawl, and only an OAI-SearchBot visit updates it.
| What sits above the H1 | Share of pages | Characters lost |
|---|---|---|
| Kicker, section or category label | 29% | 18 |
| Publication or update date | 11% | 25 |
| Alt text of the first image | 9% | 50 |
| Site or brand name | 7% | 44 |
| Breadcrumb | 7% | 23 |
| Author byline | 3% | 45 |
Source: Resoneo, ChatGPT retrieval, 534 pages ChatGPT cited, July 2026. One page in seven had no H1 at all, and when a page had one, the snippet contained it 84% of the time.
The fix is mechanical: an H1 that states the topic, nothing between it and the first paragraph but navigation, and a first sentence that answers the question with a number or a name. Put the date and the byline after the answer, as this page does.
Ranking signals you can act on
| Signal | Evidence | What to do |
|---|---|---|
| Crawler access | OpenAI: opted-out sites "will not be shown in ChatGPT search answers" | Allow OAI-SearchBot in robots.txt and at your CDN (step 1) |
| Freshness | Pages ChatGPT cited were updated a median 6 days before, against 19 for pages it read and dropped (our October 2026 panel) | Update pages when facts change, and let the crawler back in |
| Answer first | 54% of cited docs pages answer up front, against 36% of never-cited ones (same panel) | Answer in the sentence after the H1 |
| Official page types | ChatGPT sent 62% of its citations to vendor pages and docs (same panel, 1,345 answers) | Docs, pricing and comparison pages on your own domain |
| Google rank for its searches | One page in three ChatGPT used ranked on Google's first page (Resoneo, July 2026) | Rank for the long, specific searches ChatGPT writes |
| Being on the lists | ChatGPT added known brand names to its searches in 6 of 18 answers (our November 7 run) | Get onto the roundups and directories it cites (step 4) |
Why ChatGPT can't find a page
OpenAI's help page on its index lists why a page can be missing or stale. It's the closest thing OpenAI publishes to a list of ranking blockers:
- "The site blocks crawling through robots.txt or similar controls."
- "The site uses CDN or bot-blocking behavior."
- "The page requires login or personalization."
- "The content depends heavily on scripts or dynamic loading."
- "The page is new, rarely accessed, or low-signal."
- "The site structure makes deeper pages harder to discover."
Source: OpenAI, Offline web search for ChatGPT workspaces, read November 8, 2026. It adds: "there is no refresh SLA for a specific URL. A particular page may be refreshed on a different schedule depending on the site, crawl access, caching, popularity, and other technical factors."
"Popularity" and "rarely accessed" are the two you can't fix on your own site. Readers sent from the lists in step 4 are how you move them.
How to check your ChatGPT rankings
Answers change between runs, so a single check misleads. In supereo's October 16, 2026 variance run (20 buyer questions, 10 runs each), 23% of question-and-brand pairs flipped between recommended and not on ChatGPT from one run to the next. Check like this:
- Write the questions your buyers ask, in their words, including the ones that name a competitor or a budget.
- Ask each one with search on, several times, logged out as well as logged in, and note whether it searched at all.
- Open Sources and record which of your pages are cited, which competitor pages are, and whether the answer recommends you.
- Repeat weekly on the same questions, worded the same way, and compare rates, not single answers.
That's the loop supereo runs for you: every question asked again and again on all six engines, a range on every rate, and a test against control questions for every change you ship.
Questions
How do I rank in ChatGPT search?
Let OAI-SearchBot crawl your site, rank in Google and Bing for the specific searches ChatGPT writes, and publish pages whose first 200 characters answer the question. In our November 7 run, ChatGPT cited a median of 9.5 of the 16 results its searches returned.
Is there a list of ChatGPT ranking factors?
No official one. OpenAI says ChatGPT "ranks search results using multiple factors intended to help users find relevant, reliable information" and lists why pages go missing from its index. The signals above come from that page, our data and Resoneo's study.
Do I need to rank on Google to rank in ChatGPT?
It helps. Resoneo found one in three pages ChatGPT used on Google's first page, and paid thinking mode drew about three quarters of its results from Google. Free accounts lean on OpenAI's own index, which only OAI-SearchBot fills.
How often does ChatGPT update what it knows about my page?
OpenAI gives no schedule: "there is no refresh SLA for a specific URL." Resoneo found 13% of stored snippets more than a month behind the crawl. A robots.txt change takes about 24 hours, per OpenAI.
Why does ChatGPT cite my competitor instead of me?
Usually because their page is in the results for ChatGPT's searches and yours isn't, or because theirs answers in the first sentence. Check crawler access first, then compare the two pages' first 200 characters. More in how to get recommended by ChatGPT.
Related
How we compiled this
OpenAI's help pages were read November 8, 2026. ChatGPT's and Claude's searches were captured in supereo's November 7, 2026 run of real buyer questions; the panel figures come from supereo's October 2026 page panel and the variance figure from its October 16, 2026 run. Resoneo's figures are quoted from its published study. Full method.