Home

How to appear in ChatGPT answers: what OpenAI actually documents

What OpenAI documents about how ChatGPT search finds, ranks and cites websites: OAI-SearchBot, robots.txt, query rewriting, local and product results.

Published Sep 28, 2026 · 13 min read

Can you pay to appear in ChatGPT answers?

No. OpenAI states that ads do not influence the answers ChatGPT gives, that ads run on separate systems from the chat model, and that advertisers have no ability to shape, rank or alter responses.

Ads are a different surface. They can appear below the end of a response, labeled as sponsored and visually separated from it. OpenAI's January 16, 2026 post on its advertising principles calls this answer independence: answers are optimized for what is most helpful to the user.

The same holds for products. OpenAI states that product results are selected independently by ChatGPT, are not ads, and are not influenced by any OpenAI partnerships.

So there are two separate jobs. Earning a citation is unpaid work on your own site. Buying reach is covered in how to advertise on ChatGPT and how ChatGPT Ads work. Our comparison of ChatGPT Ads vs. organic mentions sets the two side by side.

What does OpenAI document, and what does it leave out?

OpenAI documents access and process. It does not publish a ranking formula.

TopicDocumented by OpenAI?
Which crawler controls search visibilityYes
How a question becomes search queriesYes
How location and memory change a searchYes
The ranking factors and their weightsNo

On ranking, the help page says only this much: ChatGPT ranks search results using multiple factors intended to help users find relevant, reliable information, and placement is not guaranteed.

Advice you will read elsewhere often goes further: a named search index that matters most, an ideal word count for an opening paragraph, a number of weeks until the first citation. None of those figures appear in the OpenAI pages listed in our sources, so this guide does not repeat them. Where we give our own judgment below, we say so.

Which OpenAI crawler decides whether you appear in search?

OAI-SearchBot decides it. OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links.

OpenAI runs four user agents, and each does a different job.

User agentWhat it is forAffects search answers?
OAI-SearchBotSurfacing websites in ChatGPT's search featuresYes
GPTBotCrawling content that may be used to train OpenAI's modelsNo
ChatGPT-UserVisiting a page when a user's request triggers itNo
OAI-AdsBotValidating landing pages submitted as adsNo

Three details in the crawler documentation are often misread:

  • The settings are independent. You can allow OAI-SearchBot and disallow GPTBot. That means: show my site in search, but do not use my content for training.
  • ChatGPT-User is not a search switch. OpenAI states it is not used to determine whether content may appear in Search, and it is not used for crawling the web automatically. Because a user starts those visits, robots.txt rules may not apply to them.
  • OAI-AdsBot only visits pages submitted as ads. OpenAI states the data it collects is not used to train its models.

If a site allows both OAI-SearchBot and GPTBot, OpenAI says it may use the results of one crawl for both purposes, to avoid crawling twice.

Blocking every AI crawler in one rule, a common default in security plugins, also removes you from search answers.

How do you allow OAI-SearchBot in robots.txt?

Add a rule that names the crawler and allows the paths you want found. OpenAI's own example uses two lines:

  • User-agent: OAI-SearchBot
  • Allow: /

To stay out of model training while staying in search, add a second group:

  • User-agent: GPTBot
  • Disallow: /

After you change robots.txt, OpenAI says it can take about 24 hours for its systems to adjust for search results. Do not judge the change on the same day.

To confirm the crawler is reaching you, look for its name in your server logs. OpenAI's example user-agent string ends with OAI-SearchBot/1.4; +https://openai.com/searchbot, and OpenAI notes the version number may change. When it fetches the robots.txt file itself, the string may carry an extra robots.txt marker, which helps when your logs do not record paths.

Why is your site still blocked after you fix robots.txt?

Because robots.txt is only the first of three layers. OpenAI's crawler guidance names them in order:

  1. robots.txt. OpenAI's crawlers respect it. If access is disallowed, crawling stops immediately.
  2. Web protection and bot mitigation. Services such as Cloudflare or Akamai can mistake a legitimate crawler for unwanted traffic, often returning a 403 Forbidden error.
  3. Human verification. CAPTCHAs, JavaScript challenges, behavioral analysis and session validation can block an automated crawler even after it passes the first two layers.

OpenAI recommends allowlisting by crawler user agent where possible. It also advises against relying only on IP addresses seen in your logs over a short period, because crawler infrastructure may change.

If your security system needs a fixed list, OpenAI publishes one per crawler:

CrawlerPublished IP list
OAI-SearchBothttps://openai.com/searchbot.json
GPTBothttps://openai.com/gptbot.json
ChatGPT-Userhttps://openai.com/chatgpt-user.json
OAI-AdsBothttps://openai.com/adsbot.json

The search help page repeats the point from the other side: to be eligible, allow the crawler and confirm that your host or content delivery network allows traffic from the published searchbot addresses.

Rate limiting is the last thing to check. OpenAI suggests reviewing responses with the code 429 Too Many Requests, plus firewall and CDN logs, around the time the crawler tried to reach you.

This guidance was written for advertisers, and the same layers block ad review. If you also run ads, see why ChatGPT ads are not serving.

How does ChatGPT turn a question into a search?

It rewrites the question. OpenAI states that ChatGPT search typically rewrites a query into one or more targeted queries, and that after reviewing the initial results it may send additional, more specific queries.

OpenAI's example is a researcher asking a long, conversational question about drug development. The first search is a short string of keywords and a year. A later search names one specific compound and a conference.

Three consequences follow for anyone planning content:

  • The query that finds you is not the sentence the person typed. It is shorter, more specific, and written by the model.
  • One conversation can produce several searches. A page that answers a narrow follow-up can be retrieved even if it would never rank for the broad topic.
  • Memory changes the query. If memory is enabled, ChatGPT may use relevant saved memories when rewriting. Two people asking the same question can trigger different searches.

ChatGPT also decides when to search. OpenAI states it may search the web automatically when a question would benefit from current information, and users can start a search manually. When no search runs, there are no citations to win.

Does ChatGPT use other search engines?

Sometimes. OpenAI states that ChatGPT search sometimes partners with other search providers, and its help page links to the privacy policies of Microsoft and Shopify.

OpenAI does not say which provider handles which question, or how much weight any provider's ranking carries. Claims that one index is the main source are not supported by the pages we read.

Our judgment: being properly indexed by conventional search engines is sensible groundwork, because partner providers are part of the process. It does not replace allowing OAI-SearchBot, which is the only requirement OpenAI states outright.

How do local businesses appear in ChatGPT answers?

Through location-aware queries. ChatGPT may use an approximate location based on the user's IP address, and OpenAI's example is a request for good restaurants nearby being rewritten into a search that names the city.

The documented details:

  • ChatGPT may estimate general location, such as country, state or city, from the IP address.
  • Device location sharing is optional and off by default. When shared, it allows more specific answers about nearby businesses.
  • A VPN or network location may affect the inferred location.
  • On iOS and Android, relevant results may include a map.
  • ChatGPT may share location with third-party providers that supply local information such as business listings.

For restaurants, ChatGPT can show reservation availability from OpenTable, Resy or Yelp. OpenAI states OpenTable reservations are available globally, Resy in the United States, and Yelp in the United States and Canada.

The practical point, which is our judgment, is simple. State your city, service area and specialty in plain text on the page, so a rewritten query that names them can match. Keep your listings on the platforms that feed local results accurate. For a wider view of options on a small budget, see ChatGPT Ads for small business.

How do products appear in ChatGPT shopping results?

Products follow a separate path, built on structured product data. OpenAI states that when deciding which products to surface, ChatGPT considers structured metadata from first-party and third-party providers, such as price and product description, along with other third-party content.

There are three routes in, according to OpenAI:

RouteWhat OpenAI states
ShopifyProduct data is already integrated through Shopify Catalog. No additional work is required from individual merchants.
Direct product feedMerchants can apply for direct feed access. Onboarding is currently available to approved partners.
Public product pagesShopping research reads product pages on publicly available retail sites directly.

Merchant order is documented too. When several merchants sell the same product, they are ranked on factors like availability, price, quality, and whether the merchant is the maker or primary seller.

Shopping research, introduced on November 24, 2025, builds a buyer's guide over a few minutes. OpenAI describes its results as organic, and notes that some retailers block automated access. When that happens, shopping research skips those sources or relies on other sites with similar products. A blocked store hands the mention to someone else.

Instant Checkout, announced on September 29, 2025, lets users buy inside the chat from eligible merchants. OpenAI states that Instant Checkout items are not preferred in product results.

If you send a direct feed, three documented limits matter:

  • Full snapshots are recommended at least daily.
  • A product missing from a snapshot is not deleted at once. OpenAI retains its most recent record for up to 14 days.
  • For a first test, OpenAI suggests a small sample of around 100 items.

For feed content, OpenAI's best practices ask for concise, factual descriptions. Our guide to the ChatGPT product feed covers the fields.

What do citations look like, and who sees them?

Responses that use web search may include citations. Users can select a citation to open the source, and select Sources to see cited sources and other relevant links.

OpenAI tells users that results and citations can be incomplete, outdated or incorrect. It advises them to open the source, check that it supports the answer, and review when it was published or updated. A visible date on your page helps a careful reader trust it.

The audience is wider than the audience for ads:

SurfaceWho can see it
Web search and citationsFree, Go, Plus, Pro, Business, Enterprise and Edu, plus people who are not signed in
AdsFree and Go plans only

A citation can therefore reach paying subscribers, who never see an ad.

How should you write pages so ChatGPT can use them?

OpenAI publishes no content rules for search, so everything in this section is our judgment, reasoned from the documented process.

  • One page, one question. Rewritten queries are narrow. A page with a single clear subject matches a narrow query better than a page that covers everything.
  • Answer first. Put the direct answer in the opening lines under a heading phrased the way people ask.
  • Use the specific terms. Follow-up queries name products, places and dates. Write those names in plain text.
  • Keep facts as text. Prices, service areas and opening hours inside an image cannot be read as words.
  • Do not hide key content behind a login, a region block or a script challenge. Those are the same barriers OpenAI lists for crawlers.
  • Date your pages and keep them current. Search is triggered by questions that need current information.

Two popular recommendations deserve a caveat. The OpenAI pages in our sources do not mention an llms.txt file or any particular structured-data markup as a requirement for search. They may do no harm, but OpenAI does not list them as a way in.

How can you check whether ChatGPT cites you?

OpenAI offers no publisher dashboard for citations in the pages we read, so checking is manual. This method is our judgment:

  1. Write a fixed list of the questions your customers really ask.
  2. Run them on a schedule, in a signed-out session. OpenAI confirms that people who are not signed in can use web search.
  3. Record for each question whether you were cited, named without a link, or absent.
  4. Note the location you tested from. Approximate location changes local results.
  5. Check your server logs for OAI-SearchBot to confirm the crawler still gets through.

Expect variation. Answers depend on the question, the location and the user's memory, so one test proves little. Look at the pattern over many runs.

A working checklist

The first four items are documented. The rest are our judgment.

  • OAI-SearchBot is allowed in robots.txt.
  • Your CDN, firewall and bot protection let it through.
  • Key pages load without a login, a CAPTCHA or a region block.
  • You waited about 24 hours after a robots.txt change before testing.
  • Each page answers one specific question, in its opening lines.
  • Facts such as prices, areas served and opening hours appear as text.
  • Pages carry a visible date and are kept current.
  • The site is indexed by conventional search engines.
  • If you sell products, your product data is accurate wherever it is published.

Frequently asked questions

Does blocking GPTBot remove my site from ChatGPT search?

No. OpenAI states that each crawler setting is independent. GPTBot relates to model training, and OAI-SearchBot controls search visibility.

How long does a robots.txt change take to work?

OpenAI says it can take about 24 hours from a robots.txt update for its systems to adjust for search results.

Can a blocked site still show up at all?

Partly. OpenAI states that sites opted out of OAI-SearchBot will not be shown in search answers, though they can still appear as navigational links.

Does ChatGPT guarantee placement if I allow the crawler?

No. Allowing the crawler makes a site eligible. OpenAI states that placement is not guaranteed.

Does running ads improve my organic visibility?

No. OpenAI states that ads do not influence answers. For advertisers it recommends allowing OAI-SearchBot as well as OAI-AdsBot, but it frames that as helping it understand public web content, not as a ranking benefit.

Does ChatGPT share the user's identity with search providers?

No. OpenAI states that ChatGPT does not share the IP address itself, or any ChatGPT account information, with third-party search providers to run a search.

Do Shopify merchants need to do anything to appear in product results?

OpenAI states that for merchants on Shopify, product data is already integrated through Shopify Catalog and no additional work is required from individual merchants.

How long does it take to be cited?

OpenAI publishes no timeline. The only documented interval is the roughly 24 hours for a robots.txt change. Be careful with anyone who promises a position or a date.

Where can I look up the terms used here?

Our ChatGPT Ads glossary defines the advertising terms. For help with the paid side, see our services or contact us.