A website can offer high-quality content, maintain stable Google rankings, and still never appear among the sources cited by ChatGPT. The reason is not necessarily poor SEO: AI search uses its own mechanisms for finding, selecting, and presenting information. To improve the chances of being cited, you need to check whether search crawlers can access the website, whether its pages are relevant, and whether they contain information genuinely worth using as a source.
However, there is no universal way to force ChatGPT to link to a particular domain. OpenAI does not disclose the full formula used to rank sources, while Google explicitly warns against questionable GEO tactics that promise guaranteed placement in AI-generated answers.
Key Takeaways
- ChatGPT Search can search the web for up-to-date information and include source links, but it does not guarantee citations for any particular website.
- To participate in ChatGPT Search, a website must allow OAI-SearchBot access and avoid blocking it at the hosting or CDN level.
- OAI-SearchBot handles the use of content in search, while GPTBot serves a different purpose related to model training. They can be configured separately.
- A high ranking in Google does not, by itself, guarantee a link in ChatGPT or Google AI Overviews.
- Google does not require a dedicated AI file, special schema markup, or artificially splitting content into tiny sections to appear in AI Overviews.
- The practical priorities are technical accessibility, original data, clear answers, and systematic measurement of AI visibility.
How ChatGPT Finds Sources for Its Answers
It is important to distinguish between an ordinary language-model response and one that uses web search. In the first case, ChatGPT may rely on the model’s existing knowledge without accessing current webpages. In the second, it searches the internet for information and adds source links when appropriate.
According to OpenAI’s documentation, its search system may reformulate a user’s question into several more precise search queries. ChatGPT also sometimes uses search infrastructure partners.
This means the wording people use to find your website on Google may not match the queries ChatGPT performs. The system may look for particular product specifications, current prices, comparisons, or confirmation of a specific fact.
Competition for citations is therefore not about a single keyword. A page should answer a specific question and contain information that makes sense in the context of an AI-generated response.
At the same time, OpenAI does not publish a complete list of the signals that determine source selection. Claims that ChatGPT always cites only Google’s top ten results or favors pages with a certain number of FAQs have no universally applicable official confirmation.
Reason #1: ChatGPT Cannot Properly Access the Website
The first step in an audit is checking technical accessibility. OpenAI uses OAI-SearchBot to discover websites and surface them in ChatGPT’s search features.
If a website owner blocks this crawler through robots.txt or blocks its IP addresses at the CDN or bot-protection level, that may prevent the content from appearing in ChatGPT Search results.
What to Check in robots.txt
Review your robots.txt file and make sure OAI-SearchBot is not prohibited from accessing the relevant sections of your website. If separate rules are defined for different crawlers, the configuration may look like this:
User-agent: OAI-SearchBot
Allow: /This is only an example of permission for OpenAI’s search crawler. The actual configuration depends on other robots.txt rules, the website’s structure, and the owner’s indexing policy.
Also check whether Cloudflare, a WAF, anti-bot services, or server rules are blocking the crawler. Permission in robots.txt will not help if the server returns an access error.
How OAI-SearchBot Differs from GPTBot
OpenAI distinguishes its search crawler, OAI-SearchBot, from GPTBot, which is associated with collecting data to train generative models.
A website owner can allow search crawling while prohibiting the use of content for training through a separate rule for GPTBot.
This matters for publishers, analytics platforms, and businesses that want visibility in AI search but have their own policies on using materials to train models.
According to OpenAI’s documentation, its systems may need approximately 24 hours to recognize changes to robots.txt. This does not mean a site will automatically start appearing in citations the following day.
Reason #2: The Page Repeats Information Already Available Everywhere
If ten websites publish almost identical product descriptions, an AI system does not necessarily need to use all of them. It may choose a source that better answers the question or provides more useful context.
This is not a confirmed universal formula for how ChatGPT selects sources; it is a practical conclusion based on search relevance principles and Google’s recommendations on original content.
In its guide to optimization for generative search, Google specifically emphasizes the value of material that offers firsthand experience, unique data, and information unavailable in dozens of other publications.
For a commercial website, this could mean in-house test results, detailed product specifications, current service terms, comparisons based on actual use, and answers to customer questions.
For publishers, it means primary sources, original expert commentary, verified timelines of events, and a clear distinction between facts and assumptions.
Reason #3: The Content Does Not Answer a Specific Question
A search query may look short but encompass several different intents. For example, someone asking about a CRM for a small business might be looking for a list of tools, prices, messaging-app integrations, or a comparison of plans.
A page with a general description of CRM software will not necessarily be the best source for an answer about specific integrations or restrictions on a free plan.
That is why a content audit should check not only for the presence of keywords but also which specific questions the material actually answers.
It is important to answer without unnecessary introductions, specify terms and dates, and explain exceptions. But there is no basis for claiming that a particular paragraph length or fixed number of lists guarantees a ChatGPT citation.
Reason #4: The Website Has a Weak Technical Structure
Content may be useful but difficult for search engines to access because of complicated JavaScript navigation, incorrect canonicals, indexing restrictions, duplicate URLs, or missing internal links.
In its guidance for AI Overviews and AI Mode, Google explicitly links visibility in generative search to basic technical requirements: a page must be accessible, indexed, and eligible to appear in conventional search with a text snippet.
ChatGPT Search does not follow exactly the same rules as Google. However, accessible HTML, a logical page structure, and the absence of technical blocks remain practical prerequisites for search systems to discover content.
If a website updates articles regularly, it is also important to maintain clear publication and update dates, working links, and correct canonical URLs.
Do schema.org, FAQs, and llms.txt Help?
A separate market of GEO advice has emerged: add dozens of FAQs, create an llms.txt file, split every article into short passages, and use special AI markup.
But Google’s official position is much more restrained. Appearing in AI Overviews and AI Mode does not require separate AI files, special schema.org markup, or artificially fragmented text.
Google explicitly states that it does not use llms.txt as a special signal for its AI search features. This does not mean the file cannot have uses in other systems, but it would be wrong to attribute guaranteed SEO benefits to it.
Structured data remains useful for describing entities and supported content types when it matches the visible page content. FAQs are also appropriate when they answer real user questions rather than existing solely for a search algorithm.
UAGEEK previously explained how to optimize a website for AI search results. Google’s new official guidance helps clarify the main principle: special GEO tricks do not replace technical SEO and useful content.
How to Check Whether ChatGPT Cites Your Website
Because OpenAI does not provide website owners with a universal dashboard listing all citations, an audit must combine several data sources.
- Build a list of queries. Use 20–30 real customer questions covering informational, comparison, and commercial intent.
- Check answers that use web search. Record which websites ChatGPT uses as sources and which specific pages it links to.
- Repeat the checks. Results may vary depending on query wording, source availability, and timing.
- Check server logs. See whether OAI-SearchBot visits your site and whether it encounters access errors.
- Analyze referral traffic. Track visits from ChatGPT in your web analytics, bearing in mind that not all interactions necessarily transmit the same referrer information.
- Compare pages. Identify topics where competitors are cited but your content is missing, then compare the completeness, freshness, and originality of the information.
Such an audit cannot establish the exact percentage of all ChatGPT answers that mention your brand. It is a sample-based observation, not official platform statistics.
It is also important to distinguish three events: a brand mention in the response text, a citation of a particular page, and an actual visit to the website. These are not equivalent and may deliver different business outcomes.
What to Measure Beyond Google Rankings
In traditional SEO, familiar metrics include impressions, clicks, CTR, rankings, and organic conversions. For generative search, it makes sense to add how often a brand appears in a test sample of AI answers, the number of cited URLs, and visits from AI services.
However, these metrics must be labeled accurately. The share of citations in your own query set is neither an official market share nor a universal ranking of your website in ChatGPT.
Analytics capabilities for Google AI Overviews and AI Mode are developing separately. UAGEEK has already covered analytics for AI features in Google Search Console. Google’s data should not be applied automatically to ChatGPT: these are different search products.
GEO Audit Plan: What to Do in the First Week
Instead of rewriting an entire website for AI, it makes more sense to start with checks that can reveal specific problems.
- Day 1. Check robots.txt, OAI-SearchBot access, server responses, WAF, and CDN.
- Day 2. Identify priority commercial and informational queries for which you want your website to appear in answers.
- Day 3. Collect examples of competitor citations and identify which pages are being used as sources.
- Day 4. Audit your own content for freshness, facts, primary sources, examples, and relevance to specific questions.
- Day 5. Fix technical issues and update the most promising pages without creating artificial duplicates.
- Day 6. Set up referral traffic monitoring and a log of AI-answer checks.
- Day 7. Record baseline metrics for comparison with future periods.
Do not expect every change to affect citations within seven days. This is an initial audit plan, not a guaranteed timeframe for results.
SEO or GEO: What Does a Business Actually Need?
GEO—Generative Engine Optimization—generally refers to efforts to improve brand and content visibility in generative answers. But treating it as the opposite of SEO is misleading.
Google explicitly confirms that its generative search features rely on core search and ranking systems. For ChatGPT, technical accessibility, relevance, and the ability to find the necessary information also matter, although its source-selection mechanism differs.
The most rational strategy, therefore, is not to abandon SEO but to expand it: focus on original information, brand visibility, technical access for AI search, and measurement of new traffic sources.
For publishers, this means more original reporting and primary sources. For SaaS and service companies, accurate documentation, comparisons, and up-to-date product pages. For e-commerce, comprehensive specifications, availability, prices, and useful product-selection content.
UAGEEK has also examined how AI answers are already affecting traditional organic traffic in its article on how AI Overviews are changing SEO.
Conclusion
If ChatGPT cites competitors, that is not evidence your website has been placed under a hidden filter. The reason may be technical, related to content, or linked to how the system formulated a particular search query.
The starting point should not be buying GEO services that guarantee citations, but checking crawler access, analyzing real answers, and creating information that contributes something new to the topic.
The goal is to make a website a reliable, useful source for users and search systems. Whether it is cited in any particular response is up to the AI system itself.
Sources for Verification
Sources for Verification
OpenAI — Searching the web with ChatGPT. Official description of ChatGPT’s web search, its use of sources and search partners, and the conditions for websites to appear in results.
How ChatGPT Search Works
OpenAI — Overview of OpenAI Crawlers. The purposes of OAI-SearchBot and GPTBot, access rules in robots.txt, and how changes are recognized.
OpenAI Crawler Documentation
OpenAI — Publishers and Developers FAQ. Recommendations for publishers on content accessibility, citations, and configuring search crawling.
OpenAI’s Guidance for Website Owners
Google Search Central — Optimizing for generative AI features. Official guide to generative search, the role of basic SEO, unique content, and the limits of GEO tactics.
Google’s Guide to AI Search Optimization
Google Search Central — AI features and your website. Technical requirements for pages to appear in AI Overviews and AI Mode, including indexing, accessibility, and structured data.
How Google Uses Websites in AI-Generated Answers
