ChatGPT first decides whether to search the web at all. If it doesn't, it answers from training and often names no business; if it does, it reads a handful of pages and recommends the businesses those pages mention or belong to. Which names you get depends on the type of question, which pages it pulls for that particular ask, and some chance, so the same question asked twice can return different businesses.
There is no published ranking formula. OpenAI has said that "any website or publisher can choose to appear in ChatGPT search." As eSEOspace points out, it has not disclosed the exact ranking method or which third-party index it uses.
What follows is what can be observed from the outside, from two places: server logs, and our weekly practice of putting the same buyer questions to ChatGPT, Claude, Gemini and Perplexity, three times each.
Does ChatGPT search the web before it recommends anyone?
Not always, and this is the first fork in the road. On 23 September 2026 we tested 20 German business questions. The model answered 10 of them without searching at all, and in those 10 answers it cited nobody. If the assistant answers from memory, your website plays no part in that answer.
When it does search, a different crawler does the work from the one most people worry about. OpenAI documents GPTBot and OAI-SearchBot as separate crawlers that are controlled independently in robots.txt:
- GPTBot crawls for model training.
- OAI-SearchBot surfaces pages in ChatGPT's search results.
You can allow one and block the other. When a real person is waiting for an answer, a live agent such as ChatGPT-User fetches the page at that moment.
We see these live fetches in logs. Over the same fortnight:
- The edtech platform pickaclass.com received 1,918 live retrievals on content pages.
- A client blog of a few dozen articles received 56 retrievals, spread across 24 different articles.
Each of those fetches is a moment when a page could become part of a recommendation.
Where do the search results come from? OpenAI's help documentation states that ChatGPT search "sometimes partners with other search providers," and names Microsoft and Shopify as its disclosed partners, as this comparison of three AI engines quotes. That is one reason Bing indexing matters. We cover that in Does ChatGPT Use Bing? Do I Need Bing Webmaster Tools and IndexNow?
Which sources does it read when it searches?
It reads the pages that best answer the question as asked, and those are often not business websites.
In the 23 September test, wherever the model searched, it cited primary sources on 9 of 10 questions: laws, ministries, the OECD and large accounting firms. Individual firms were named only on "who should I hire" and "how do I choose" questions.
A local example shows the same pattern from the business side. On 20 October 2026 we put 20 patient questions about an orthopaedic practice in Bonn to the four assistants. Out of 80 answers, the practice was named in 12:
- Claude: 5 of 20
- Gemini: 4 of 20
- ChatGPT: 2 of 20
- Perplexity: 1 of 20
The directory jameda.de was cited in 12 answers. The most cited sources overall were competing practices' own websites. In effect, the competitors' pages became the evidence the assistants used to decide whom to recommend.
The type of question decided whether any practice appeared at all:
| Question type | Example | What we observed |
|---|---|---|
| Who to choose | "Which orthopaedic practice in Bonn should I see for knee pain?" | Practices named; jameda.de and competing practices' sites cited |
| Costs, risks, treatments | "What are the risks of a knee operation?" | No practice named in any answer |
| General business or legal questions (23 September test) | "What do I need to register a company?" | Half answered without searching; when searched, primary sources cited on 9 of 10 |
Perplexity behaves in a similar way but relies on its own crawler. LLM Pulse notes that Perplexity documents PerplexityBot, which it uses to surface and link sites, and that it draws from a continuously refreshed index rather than a static training snapshot. So a page that Perplexity can crawl today can show up in its answers soon after.
Why do two asks of the same question name different businesses?
Each answer is generated fresh. The assistant may phrase its search differently, retrieve different pages and sample different wording.
In the 23 September test, two runs of the same question named two different sets of firms. We see this every week: asked the same question three times, an assistant often names different businesses.
This changes how you should read a single result. One screenshot of ChatGPT naming you, or ignoring you, proves very little.
Our own rules for measuring are:
- We count a business as named for a question if any of three asks names it.
- We compare weeks only on the same questions with the same number of asks.
Repeated measurement is cheap. With API prices of October 2026, a report of 20 questions across 4 assistants with 3 asks each costs us about $5.40.
What makes a business more likely to be named?
From our measurements, five things matter more than anything else.
1. Crawlers can reach the page. If robots.txt or a firewall blocks the search crawlers, you cannot be retrieved. We made this mistake ourselves. On itgrows.ai, a rule meant to hide a single redirect page (/b) was written as a prefix. Because robots.txt rules match prefixes, from 11 August to 21 October 2026 it also closed /blog, /business and every business article to all crawlers. Bing Webmaster Tools listed our business page as "discovered but not crawled", and Google had none of those pages indexed. We changed the rule to /b$ on 21 October 2026.
2. You have enough pages that answer real questions. One site had complete markup: server rendering, Article, FAQPage and Organization JSON-LD, and a sitemap. With only 75 URLs, it was still absent from answers on its own article's exact topic. A site with about 190,000 crawlable URLs was crawled constantly. Markup helps, but surface area decides how often you get a chance at all.
3. Assistants can find you, not just Google. On 19 October 2026 we asked assistants for the exact titles of 18 articles, each older than three weeks, on three client sites. ChatGPT found 1 of 18 and Claude found 7. Google had indexed 29 of 30 articles on one of those sites. Being in Google's index did not mean an assistant could find the page.
4. The content is easy to quote. The peer-reviewed GEO paper from Princeton (KDD 2024) built a benchmark of real user queries and found that specific content changes can increase a source's visibility in generative answers by up to 40%.
5. Other sites mention you. In the Bonn example, a directory was cited as often as the practice was named. On 21 October 2026, the sources cited most often for questions about our own field were the blogs of GEO agencies and tool vendors themselves. Assistants recommend whoever the retrieved pages point to.
If competitors keep getting named and you don't, Why Doesn't ChatGPT Mention My Company but Names My Competitors? goes through the usual causes.
Does being recommended actually bring customers?
It can, but don't expect it to show up as website traffic. In about two weeks of logs from three of our course websites, human visitors arriving from AI assistants were:
- Copilot: 1
- ChatGPT, Perplexity, Gemini, Claude, Meta AI and others: 0
Assistants mostly answer in place rather than send a click.
The value lies in being the name in the answer. Our clients report results such as these, in their own words, from October 2026:
- FreightSizer, a B2B SaaS for freight dimensioning, got 2 new B2B customers in one month, both from ChatGPT.
- appslift, an app store optimisation service, gets 2 to 5 leads a month from AI assistants.
These are their reports, not a forecast for anyone else.
What should you do next?
Start with what blocks you, then measure, then build.
- Check crawler access. Look at robots.txt and any bot protection for OAI-SearchBot, ChatGPT-User, PerplexityBot and ClaudeBot. Watch for prefix rules like ours. Is My Website Blocking GPTBot, ClaudeBot and PerplexityBot? walks through it.
- Check indexing in both Bing and Google. Don't assume one implies the other, or that either implies an assistant can find you.
- Ask your buyers' "who should I use" questions yourself, several times each. The free check at https://www.itgrows.ai/ generates five buyer questions from your homepage and shows in about 30 seconds whether ChatGPT and Perplexity name you.
- Publish pages that answer the questions buyers ask, on your own domain. Cover costs, risks and how to choose, not just your services page. Get listed where your sector's directories and comparison pages are.
- Measure weekly on fixed questions. Use the same questions and the same number of asks each time, so you can tell a real change from noise.
This is the work ItGrows does for $499 a month (plus a one-time $499 setup, free with 3 months or a year). It publishes articles on your domain, fixes crawler access and indexing, and counts every week how often the four assistants name you. Nobody can guarantee that an assistant will name a business, and we don't. On 21 October 2026 our own business was named in 0 of 160 answers, which is exactly why we measure.
Frequently Asked Questions
Does ChatGPT always search the web before recommending a business?
No. In our 23 September 2026 test of 20 German business questions, it answered 10 without searching and cited nobody in those answers. Questions that ask whom to hire or how to choose are the ones most likely to produce named businesses.
Why does ChatGPT recommend my competitors instead of me?
Usually because the pages it retrieves for that question are your competitors' pages, or directories that list them. In our Bonn example, competing practices' own websites were the most cited sources. Crawler blocks and missing pages on the exact question are the other common causes.
Is being indexed in Google enough to be recommended by ChatGPT?
No. On 19 October 2026, ChatGPT found only 1 of 18 client articles by exact title, even though Google had indexed 29 of 30 articles on one of those sites. ChatGPT's search draws on partners such as Microsoft, so Bing indexing and crawler access matter as well.
How can I tell whether ChatGPT recommends my business?
Ask the questions your buyers ask, several times each, because answers vary between asks. Count yourself as named if any of the asks names you, and compare only like with like over time. The free check at https://www.itgrows.ai/ does a first pass for five questions on ChatGPT and Perplexity.