How to Get Your Website Cited by ChatGPT and Other AI Search Engines

We tested real questions from potential buyers in ChatGPT, recorded every source it cited, and checked our own server logs to learn what helps a website get quoted by AI search tools.

To get ChatGPT to cite your website, let OpenAI's search bot read your pages. Then publish content that answers one specific question per page, gives the answer first and uses data from your own work. The same approach can help with Perplexity, Claude and other AI search tools. None of these AI companies has published the exact rules it uses to choose sources, so nobody can promise that your page will be cited. You can control whether the AI can access your page and whether your content is worth quoting.

Rather than repeating what SEO companies say, we tested this on our own website. Some results were encouraging, while others were not. Most importantly, our website was never cited in the test.

What is AI search optimisation, and what are AEO and GEO?

AI search optimisation means making your website easy for AI tools to find, read and quote when someone asks a question. You may also see it called answer engine optimisation (AEO) or generative engine optimisation (GEO). The names are different, but the goal is the same. When someone asks ChatGPT, Perplexity or Claude a question and the tool searches the live web, you want it to find your page, understand it and trust it enough to cite it.

This work is related to regular search engine optimisation (SEO), because many AI tools search the web before answering questions. But the two are not exactly the same. The main difference is how the information is used. A search engine ranks your page in its results. An AI tool reads the page, takes the information it needs and may or may not show where that information came from.

Which AI crawlers visit your website, and what does each one do?

The major AI companies use several bots, and each bot has a different job. Confusing them is an easy way to block your website from AI search by mistake. According to the companies' own documentation, their bots generally fall into three groups: bots that collect information for training AI models, bots that build a search index, and bots that fetch a page when someone asks a question.

Two things stand out. First, bots that collect training data visited our site very little. Bots that fetch pages to answer live questions made up most of the activity. Second, the user-request bots are not regular crawlers, according to the companies themselves. OpenAI says robots.txt rules may not apply to ChatGPT-User, and Perplexity says the same about Perplexity-User. This means you can block bots that collect training data and still allow your site to appear in search. However, bot names can be faked, and we did not check the IP address ranges to confirm who made the requests. So these numbers count traffic that identified itself using those names.

How do you check that ChatGPT can crawl your website?

Start by checking your robots.txt file. If it blocks all AI bots or contains a rule you copied without understanding it, you may be blocking more than you intended. OpenAI's crawler documentation says changes to robots.txt can take about 24 hours to take effect. It also explains that you can allow search bots while blocking bots used for training.

Next, check that people and bots outside your network can load robots.txt. We discovered that our content delivery network (CDN) was returning a saved redirect that pointed back to the same address. This created a loop, so every crawler that asked for the file got stuck and could not read it. The server behind the CDN was serving the file correctly, which is why everything looked fine from inside our system. Test the file from a network outside your own, both with and without a query string added to the URL.

Finally, check your server logs. They can tell you more than any plugin. On 25 and 26 October, the ChatGPT-User bot made about 126,800 requests to 6,421 different URLs on our website. Of those requests, 99.7% returned a normal page. Almost all the visits were to timely news articles. On 26 October, 877 of its 65,080 requests, or 1.3%, went to our longer guides. These guides make up about 5% of everything we publish.

But being accessible does not mean a page will be selected. A bot can fetch a page many times without ever citing it as a source. Your logs show that the door is open, not whether anyone is coming inside.

What does ChatGPT actually cite when people ask buyer questions?

We wrote 48 questions that a founder or small business owner might realistically ask. We entered the first 13 into a logged-in ChatGPT account, one at a time, and recorded every source it cited.

• Seven of the 13 answers cited at least one source. Six cited nothing and answered from information already available to the model.
• Across all 13 answers, ChatGPT cited 14 links. None pointed to our website.
• In a separate test, we added the words "cite sources" to the question. One answer then included eight cited links among 20 links in total. Asking for sources changed the result.

We then opened the 14 cited pages. Twelve loaded correctly. Apart from one report that was 54,000 words long, the pages ranged from 528 words to about 5,600 words. Three used FAQ schema, three included a table, and six showed a visible date when they had been updated, including pages updated that month. There was no single format that all the pages followed. What they had in common was that each was an original source for the question it answered. For example, an organisation explained its own product, a company documented its own tool, or a data team published its own quarterly figures.

This was a small test using one account on one day, and ChatGPT's answers can change each time you ask the same question. So treat these results as clues, not firm rules. We stopped after 13 questions because the results were predictable: we had not created our content with citations in mind, and testing the other 35 questions would probably have shown the same thing.

How do you get cited by ChatGPT, Perplexity and Claude?

No one outside the AI companies knows exactly how these tools choose sources. SEO companies suggest that AI tools prefer passages that are easy to copy and that match the smaller questions an AI breaks a larger query into. This is an observation, not a rule confirmed by any AI company. With that in mind, the following ideas are possible explanations based on what the cited pages had in common.

• One question per page. Build each page around one clear, specific question instead of trying to cover a broad topic. The pages cited in our test focused on narrow subjects.
• Give the answer in the first 100 words. State your main answer first, then explain and support it.
• Include something only you have. Share your own numbers, test results, decisions and what happened afterward. A general how-to article competes with similar content from large organisations and may struggle to stand out.
• Name your sources and link to them. Link to the original source for every claim that does not come from your own work.
• Show the date and give readers a reason to trust it. Display the date the page was updated, and actually update the content when needed.
• Use a simple structure. Add headings written as questions, a table when comparing things, and a short FAQ with schema. These are easy changes to make, but our test suggests they are not enough on their own to earn a citation.

How do you track traffic and citations from AI search?

There are three useful signals, but none gives you the full picture. First, check your server logs for visits from ChatGPT-User, OAI-SearchBot and PerplexityBot. Second, look for visits from AI search tools in your website analytics. Links that ChatGPT shows to users include the parameter utm_source=chatgpt.com, which we found on the cited links in our test. Third, create a fixed list of questions, test them on a regular schedule and record the results. This is the only way to see how your results change over time.

Do not rely on a single test. Independent research into ChatGPT search has found many citations that were inaccurate. This means a citation, or the absence of one, is not always a reliable measure of what is happening.

What we are doing about our own zero

We are rebuilding our longer guides using the ideas listed above. We are starting with specific questions for which we have first-hand data, and we will test the same questions again afterward. If the results improve, we will report it here. If they do not improve, we will report that too.

Frequently asked questions

Does blocking GPTBot hide my website from ChatGPT?

No, not by itself. OpenAI says GPTBot is used for training, while OAI-SearchBot controls whether pages can appear in search. You can block the training bot while allowing the search bot.

Can I pay to be cited by an AI search tool?

We found no evidence of a paid option for ordinary websites to get citations. The documentation we reviewed describes access to crawlers, rather than payment, as the requirement.

How long until a new page can be cited?

OpenAI says changes to robots.txt take about 24 hours to be picked up. No AI company has published how quickly a new page gets added to its search index, so be cautious about anyone promising a specific time.

Updated 27 October 2026. The test described here was run on 27 October 2026 using a single ChatGPT account. The server log figures cover 25 and 26 October.