How does ChatGPT work and where do its answers come from?
Training data and live web search explained: how ChatGPT chooses sources and why a top position in Google isn't required.
Lees dit in het Nederlands →$ trace --system=chatgpt --direction=source
How does ChatGPT work and where do its answers come from?
ChatGPT works along two routes: it builds answers from the knowledge that went into the model during training, and from live web search results when the search function is switched on. In both cases it chooses sources based on content and structure, not on Google's ranking order. This piece explains how it works in plain language; there's no formula involved and nothing is being sold.
Anyone who wants to know how to end up in ChatGPT answers must first know where those answers come from. That's less mysterious than it seems. There are two routes, and they work fundamentally differently.
01 / route oneKnowledge from training
The language model behind ChatGPT is trained on a large volume of text from the open web, up to a cut-off date. Everything the model read about your field, your market and possibly your company during that period is embedded in the model itself.
If you ask ChatGPT about your market without the search function, it answers entirely from this memory. As a result, the answer can lag years behind reality, and the model rarely mentions how old its knowledge is.
This route is slow and incomplete. If you're not in it, you can't fix that quickly: new content only reaches the model at the next training round, and nobody outside OpenAI decides what goes into it. What does help: texts that have been online for years, are widely read and frequently cited, carry the most weight here. It's the long term of AI visibility.
02 / route twoLive web search
If the search function is switched on, ChatGPT searches the web live at the moment of the question, reads a number of pages and builds an answer from them with source citations. This is the route you can act on in the short term, because what matters here is what's currently on the web and whether your pages are readable and can be found at that moment.
The selection works in two steps. First, a search query determines which pages are retrieved. Then the model decides which passages from those pages make it into the answer. That second step is where structure makes the difference: a page that literally answers the question, in a passage that reads independently, is more likely to be cited than a page where the answer is spread across twelve paragraphs.
Note a detail that's often missed: the question the user asks is rarely the search query the system actually runs. ChatGPT reformulates the question, sometimes splits it into multiple search queries and combines the results. So you don't need to cover every conceivable phrasing; you need to treat the topic fully and clearly, so your page surfaces across several of those derived search queries.
The two routes side by side:
| Route one: training | Route two: live web search | |
|---|---|---|
| Source of the answer | Knowledge in the model itself, up to the cut-off date | Pages read at the moment of the question |
| Timeliness | Can lag years behind | What's currently on the web |
| Short-term influence | Almost none: waits for the next training round | Direct: today's readability and visibility count |
| Source citation | Rare | Standard, with reference to the pages used |
In practice, this means that any improvement you make today can already appear in the next answer via route two, while the same improvement only filters into the model itself much later. Anyone working on their visibility therefore almost always works on route two first; route one follows naturally as texts stay online longer and get cited more often.
03 / broader than googleWhy a top position isn't required
ChatGPT looks more broadly than the top results of a search engine. Where Google's AI Overviews lean heavily on its own rankings, ChatGPT draws its sources from a wider selection and relies less on position than a search engine does.
For an AI answer, you're not competing with ten blue links, but with every page that answers the question better than yours does.
That has two sides. The favourable one: a well-structured page from a small company can be cited without that page ranking top in Google. The unfavourable one: a top position in Google is no guarantee of a place in the answer. The selection criteria overlap, but they aren't identical. How those criteria differ per platform is explained further in how Perplexity works and how you get into the sources.
04 / third partiesWikipedia, Reddit and trade media count too
ChatGPT strikingly often cites sources that aren't the companies themselves. According to Profound's analysis covering August 2024 to June 2025, Wikipedia accounted for 7.8 per cent of all ChatGPT citations and Reddit for 1.8 per cent, with review sites and trade media adding to that. An answer about your market can therefore be built from what others write about that market, without a single company site being cited.
For your visibility, this means your own site is only half the story. What's said in third-party places about your market and your company also determines whether and how you appear in answers. What AI visibility means in full is covered in what is AI visibility; what we practically do about it is on datadriftdigital.nl.
Who doesn't need to worry about this: companies without customers searching online. For everyone with inflow from search traffic, these two routes together determine where that inflow will come from in the years ahead.
Where does ChatGPT get its information from?+
Do I need to rank top in Google to appear in ChatGPT?+
Can I influence what ChatGPT says about my company from training data?+
Why does ChatGPT cite Wikipedia and Reddit so often?+
What's the difference between ChatGPT and Google AI Overviews as a source of answers?+
Curious which sources the answers about your market are built from?
The scan
This is the baseline measurement: fixed search queries on ChatGPT, Gemini, Claude, Perplexity and Google. Current prices are on the pricing page. Even without a follow-up, you keep the report.
This text was produced with AI assistance and checked and approved by a human before publication.