ChatGPT SEO: how to rank in ChatGPT answers and recommendations

ChatGPT SEO is the work of getting ChatGPT to find your site when it searches the web, cite it as a source and mention your business when it makes recommendations. According to OpenAI, the published requirement is to let its crawler, OAI-SearchBot, through, including at your firewall. Everything else depends on factors it does not detail, and placement is not guaranteed.

To check it on your own site: AI visibility audit

This article goes deep on a single engine; the general framework is in the guide to AI search optimization. A declaration of interest: this blog belongs to SmoothSeen, a website audit tool that measures visibility in search engines and AI assistants.

How does ChatGPT decide which sites to cite?

ChatGPT cites websites when it searches the web, and it chooses among what its searches return using criteria OpenAI does not publish in detail. When it answers without searching, it draws on what it learnt in training and does not link to sources.

OpenAI's help centre describes the process in three steps1:

1. It decides whether to search. ChatGPT may search the web automatically when a question would benefit from current information. Users can also force a search from the tools menu. Web search is available on every plan, including Free, and to people who are not signed in.
2. It rewrites the question. When it partners with other search providers, ChatGPT typically turns the question into one or more targeted queries and, after reviewing the first results, may send more specific ones. For a local question, it may add the approximate location it infers from the user's IP address.
3. It ranks and cites. OpenAI puts it this way: ChatGPT ranks results using multiple factors intended to help users find relevant, reliable information, and "placement is not guaranteed". Answers that use search may include citations and a Sources panel.

In ChatGPT you are competing for very few slots. SISTRIX found that ChatGPT cites 3 to 4 domains per answer on average, against 14 to 16 in Google AI Mode2. And it rotates them heavily. SISTRIX's headline figure for ChatGPT is 74% new domains every week, which is the figure for Germany, the market it covers in most detail; in the UK it was 60% and in France 42%. For the median prompt, not a single domain appeared in all 17 weeks of the study. The sample: 82,619 prompts and 1,548,213 snapshots in six countries, including the UK and the US, from 17 December 2025 to 8 April 2026.

One more finding matters if you compete in a language other than English. For German-language queries, 68% of the sources in ChatGPT's core (the ones that recur week after week) were in English2. A business selling in French, German or Spanish may find that its rival for the citation is an English-language site, not just the local competition.

Which OpenAI crawler needs access to your site?

The one that decides whether you appear in search answers is OAI-SearchBot. OpenAI documents four user agents and states that each setting is independent of the others3:

OpenAI adds that a robots.txt change can take around 24 hours to reach its search systems, and that if you allow both OAI-SearchBot and GPTBot it may use a single crawl for both purposes3. The guide to AI crawlers and robots.txt applies the same logic to Anthropic, Perplexity and Google.

An annotated robots.txt for ChatGPT

This example lets ChatGPT search in and opts out of training. It uses example.com, which is reserved for examples; adapt the private paths to your own site:

# ChatGPT search: required to appear and be cited in its answers.
# A crawler with its own group ignores the "*" group, so the
# private paths are repeated here.
User-agent: OAI-SearchBot
Allow: /
Disallow: /cart/
Disallow: /my-account/

# Training of OpenAI's models: a separate editorial decision.
# Blocking it does not remove you from ChatGPT search.
User-agent: GPTBot
Disallow: /

# All other crawlers.
User-agent: *
Disallow: /cart/
Disallow: /my-account/

Sitemap: https://www.example.com/sitemap.xml

The private paths are repeated because, under the robots.txt standard (RFC 9309), a crawler that finds a group naming it obeys that group and only falls back to the * group if none matches4. A group for OAI-SearchBot containing nothing but Allow: / also opens up the basket. Blocking GPTBot, on the other hand, is optional: it is a decision about training, not about search.

Two failures robots.txt will not show you

The first is on the server. OpenAI asks that your host or CDN allow traffic from the IP addresses it publishes for its search crawler1. A firewall or bot-protection service that blocks those addresses keeps you out even with a perfect robots.txt. To check, search your access logs for OAI-SearchBot and match the IPs against the official list, openai.com/searchbot.json. If there have been no visits for weeks, something is stopping it.

The second is confusing Disallow with noindex. OpenAI explains that if it obtains the URL of a disallowed page from a third-party search provider or by crawling other pages, it may show just the link and page title in ChatGPT Atlas, its browser5. To keep a page out entirely, use the noindex meta tag, and let the crawler in so it can read it.

What makes ChatGPT recommend a business?

When ChatGPT recommends businesses, it combines two sources: what it learnt in training and what it finds when it searches. In both, what others say about you carries weight, not only what your own site says.

OpenAI explains that its models are trained on information publicly available on the internet, information accessed through third-party partnerships, and information provided by users, human trainers and researchers. It also warns that the same question can produce different answers, because generation has an inherent element of randomness6. A ChatGPT recommendation is not a fixed list you can "occupy".

The most useful independent data here comes from Ahrefs. In a study of 75,000 brands published on 12 December 2025, branded web mentions correlated with appearing in ChatGPT (0.664) far more than Domain Rating did (0.266)7. YouTube mentions were the most strongly correlated factor across the three platforms studied, at around 0.737. The number of pages on a site had almost no relationship with AI visibility (around 0.194). Ahrefs used Spearman correlation on brands with a DR above 40, and warns that correlation is not causation: pushing those metrics up does not guarantee more visibility.

The practical reading: it helps to show up where your sector is discussed, in comparisons, trade press, directories, forums and reviews.

If you sell products

Product listings follow a different route. ChatGPT shows product options when it detects shopping intent, and OpenAI states that those results are selected independently: they are not ads and are not influenced by OpenAI partnerships, and ads are kept separate8. According to the same page, updated in early October 2026:

• ChatGPT considers structured metadata from first-party and third-party providers, such as price and product description, as well as what the model itself answers before it looks at new search results.
• Review summaries are generated by the model from reviews on public websites.
• When several merchants sell the same item, they are ranked by availability, price, quality and whether they are the maker or primary seller.
• Shops on Shopify are already integrated through Shopify Catalog. Other merchants can apply to OpenAI to send a direct product feed.
• For some eligible products and merchants, ChatGPT offers Instant Checkout, which lets the buyer pay without leaving ChatGPT.

If you run a local business

ChatGPT uses an approximate location for "near me" questions and rewrites the query with the user's town or city1. For restaurants, it can show availability and bookings from OpenTable, Resy or Yelp: OpenAI lists OpenTable as available globally, Resy in the United States, and Yelp in the United States and Canada1. The guide to local SEO for AI search explains what to do with your site and your business profile so that search finds you.

What to do to rank in ChatGPT

This table sums up what is backed by a source you can check. It runs from the essential to the helpful.

The fifth row has its own guide: how to write content that AI will cite, with before-and-after examples.

Two things are left out of the table on purpose. In the GEO study, keyword stuffing performed 10% worse than doing nothing on Perplexity9. And OpenAI does not document that its search crawler reads an llms.txt file.

How do you know whether ChatGPT cites you?

There are three ways to find out, and none is enough on its own:

• Traffic. ChatGPT adds utm_source=chatgpt.com to the links in its answers, so those visits can be filtered in your analytics tool5. How to see them in GA4 is covered in the guide to tracking AI traffic in GA4.
• Repeated queries. Ask ChatGPT, with search switched on, the questions a customer would ask, and note whether it cites you, names you, and whom it cites instead. Repeat the same questions every week. With the churn SISTRIX measured, a single snapshot tells you nothing.
• Server logs. Visits from OAI-SearchBot tell you whether the crawler gets in. Visits from ChatGPT-User are fetches made at the request of a ChatGPT user.

SmoothSeen covers both halves. The causes: its AI visibility analysis checks whether your robots.txt lets AI search crawlers in and whether the page has an extractable answer, sourced data and authorship, and turns the gaps into an action plan. The outcome, from the Professional plan upwards: you write the questions a customer would ask, and every week SmoothSeen puts them to ChatGPT and Gemini with web search on, to see whether they cite you, only name you, or show someone else instead.

The difference between what a model learns in training and what it looks up when it answers, across every engine, is explained in LLM SEO.

Frequently asked questions

How do I get ChatGPT to recommend my business?

There is no direct or guaranteed way. What the evidence supports is letting OAI-SearchBot read your site, explaining clearly what you do and where, and getting other sites in your sector to mention you. In the Ahrefs study of 75,000 brands, mentions on the web and on YouTube were the signals most correlated with appearing in ChatGPT. Even so, the same question can produce a different answer each time.

Does blocking GPTBot remove me from ChatGPT?

No. GPTBot is the crawler that collects content to train OpenAI's models, and OpenAI says each setting is independent. What removes you from search answers is blocking OAI-SearchBot. You can allow OAI-SearchBot and block GPTBot at the same time: you appear in search, and your content is not used for training.

Does ChatGPT search Google or Bing?

OpenAI does not say which provider it uses for a given query. It says ChatGPT sometimes partners with other search providers, to which it sends rewritten versions of the question, and its help page links to the privacy policies of Microsoft and Shopify. It also runs its own crawler, OAI-SearchBot. The sensible approach is to be well indexed in the major search engines and not to block OAI-SearchBot.

Can I ask OpenAI to include my website?

For websites there is no submission form: OpenAI says any public website can appear in ChatGPT search as long as it does not block OAI-SearchBot. For products there is a route: merchants can apply to send a direct product feed, and Shopify shops are already integrated through Shopify Catalog. Neither route guarantees an appearance in any particular answer.

What to do next

Open your robots.txt and check that OAI-SearchBot is not blocked, either by name or by a blanket Disallow: /; then look for its visits in your server logs. With access sorted, work through the remaining blocks in the guide to AI search optimization.