You get found in AI chatbots by writing citable content, keeping your site readable for AI crawlers, and measuring the result across multiple models, because ChatGPT, Gemini, Claude and Perplexity each fetch their answers differently. Type your own brand name into ChatGPT and you get a tidy answer, but when a customer asks "which provider would you recommend?", the chatbot names your competitor. With more than 800 million weekly ChatGPT users, as Sam Altman said in October 2025, that second case is what matters. Below: how the chatbots fetch sources, why they cite different ones, and what to do about it.
How do AI chatbots fetch their answers?
AI chatbots each fetch their answers in a different way, and that difference shapes how you get found. Some models draw on their own search index, others connect live to an external search engine. Treat them as one block and you miss exactly where your brand does or does not surface.
The common thread: almost every chatbot fetches current web sources at the moment of the question and cites them. That means classic accessibility matters, because a page the crawler cannot reach cannot be cited. How ChatGPT does that specifically, with the OAI-SearchBot crawler, is covered in ChatGPT SEO. The other models work on the same principle with their own choice of sources; Google's own search answers are covered in Google AI Overviews.
Does every chatbot cite the same sources?
No, and that is the single biggest reason to measure your visibility broadly. Platforms differ sharply in which sites they trust. An analysis by Profound of 680 million citations (August 2024 to June 2025) found that Wikipedia accounted for nearly 48% of ChatGPT's top-10 sources, while Reddit made up around 47% of Perplexity's top-10 sources. Those are shares within the most cited sources, not overall totals, but the pattern is clear: every model has its own preferences. This is international research, indicative for any market rather than exact.
The practical takeaway: a single content strategy does not automatically win across all AI chatbots at the same time. Be well represented in the sources ChatGPT trusts, and you may still be absent from what Perplexity cites. That is why you measure your AI visibility across all four of the big chatbots.
That difference also changes how you deploy content. For a model that leans heavily on encyclopedic sources, factual, well-structured explanation weighs most. For a model that cites forums and reviews, it matters whether people talk about you elsewhere. A separate site per model is unnecessary. You do need to know which model you are missing where. Veesie measures the AI visibility of your brand for marketers and agencies and shows where you stand per source: ChatGPT, Perplexity and Google AI Overviews on every plan, plus Claude and Gemini from Plus.
Which AI chatbot should you focus on?
Focus on the models where your audience is, but measure broadly, because the field shifts fast. The audience is already there: a study by Eight Oh Two (January 2026) found that 37% of the 500 AI users it surveyed start a search in an AI tool instead of Google. ChatGPT is the largest with more than 800 million weekly users (October 2025), but Google is pushing AI hard into its search engine: in July 2025 Google AI Overviews reached 2 billion monthly users across more than 200 countries. In practice that makes ChatGPT and Gemini the first two to watch.
The trap is fixating on one model because you happen to score well there. Because each platform cites different sources, your position in one is no predictor for another. How AI visibility fits into your broader marketing mix is covered in what GEO marketing is. The practical principle stays the same: measure all four, and prioritise your work where your audience actually searches.
What works to get found in AI chatbots?
The strongest lever is citable content: factual, structured text with numbers, sources and clear phrasing. The GEO study from Princeton (KDD 2024) scored quotes, statistics and sources 30 to 40% higher, the best about 41%, on a relative visibility measure within five pre-selected documents, while keyword repetition backfired. That is a proxy and no expected effect on your visibility: a survey of 45 GEO studies (July 2026) found no technique with a stable, longitudinal, cross-platform causal effect on organic discoverability. But visibility in AI chatbots reaches further than your own site. Three things count together:
- Citable content on your own site. Open every section with a direct, factual answer, backed by a number and a source. That is literally easier to cite than a wall of running text.
- A site that AI crawlers are allowed to read. Block the search crawler of an AI model by accident, and you disappear from its answers. The technical groundwork is covered in the AI-Readiness score, the full check in the GEO audit checklist.
- Mentions on authoritative external sources. AI models cite heavily from Wikipedia, trade media, comparison sites and forums. Being named where your audience already looks counts. But beware of forced tactics: more persistent misconceptions are debunked in 8 GEO myths.
Do not overestimate the complexity. Amanda Natividad of SparkToro sums it up plainly:
...a lot of AI optimization advice feels like using a pressure washer to rinse a spoon.
Amanda Natividad, SparkToro, in AI optimization is mostly just good marketing
In other words: good, authoritative content and a readable site get you furthest. The full step-by-step plan is in how to do GEO, the SEO side of it in AI SEO and the concrete actions in 8 tips to improve your GEO Score.
Why you cannot rely on a single measurement
One check in an AI chatbot says little, because AI answers are non-deterministic. Even with identical settings and the same prompt, a language model does not produce an identical answer. Research into non-determinism in LLMs (2024, revised 2025) measured differences of up to 15% in accuracy between repeated runs with exactly the same input. Ask once today whether ChatGPT names you, and the answer can be different tomorrow, without you changing anything on your site.
So your AI visibility is not a fixed position but a probability distribution. You only measure it reliably by asking the same questions repeatedly, across multiple models, and averaging out the noise into a trend. That is also why the AI answer can replace the visit to your site: click-through fell from 15% to 8% when an AI summary appeared, in US data from the Pew Research Center (July 2025), as what GEO is explains. Whoever sits in that answer stays in view. Veesie does the repeating for you. Every plan measures weekly, Claude and Gemini are asked up to three times per question with the majority answer counting, and the trend over several weeks smooths out the rest of the noise. Start the free 7-day trial to see it for your own brand.
How do you measure your visibility in AI chatbots?
You measure your visibility by asking relevant questions to multiple AI models repeatedly and counting how often and how your brand is named. Three numbers together tell the story: how often you get named, how you compare to your competitors, and whether that mention lands positively or negatively. A high mention rate with negative sentiment is no win, and you only see that when you measure all three.
Because each chatbot cites different sources, you measure across all the big chatbots at once, not on one. Veesie tracks weekly whether a brand is cited. The GEO Score (0 to 100) is an index over ChatGPT, Perplexity and Google AI Overviews; from the Plus plan, Claude and Gemini are measured as well and shown per source next to it. How to set that up yourself is covered in the KPI guide: how to measure AI visibility, and sector averages for Belgium and the Netherlands are in the GEO benchmark, a snapshot. The difference from classic SEO is covered in GEO vs SEO.
Conclusion: measure broadly, optimise precisely
Getting found in AI chatbots is not a matter of one trick for one model. ChatGPT, Gemini, Claude and Perplexity each fetch their answers differently and cite different sources, so your visibility differs per platform. The foundation is the same everywhere: citable content, a site AI can read, and authority that reaches beyond your own domain.
So start by measuring broadly and optimising precisely. First know where you stand in each model, then adjust on the weak spots. Veesie measures that AI visibility for marketers and agencies, per source and per question, so you steer on measurements. Compare the pricing and features to see which sources each plan covers.