How does ChatGPT decide which businesses to recommend?
When ChatGPT searches the web to answer a question, it cites and links the pages it used. OpenAI says "any public website can appear in ChatGPT search." It doesn't publish how it picks which pages to cite, so the reliable levers are the ones OpenAI documents: letting its search crawler in, and publishing content worth citing.
ChatGPT finds pages in two ways:
- Its own crawler. OpenAI uses OAI-SearchBot "to surface websites in search results in ChatGPT's search features."
- Other sources. OpenAI says it may also get a page's URL "from a third-party search provider or by crawling other pages."
Which OpenAI crawlers matter, and what do they do?
OpenAI runs three user agents, and each one does a different job:
| User agent | What it does | Effect of blocking it |
|---|---|---|
| OAI-SearchBot | Surfaces websites in ChatGPT's search features | Sites opted out "will not be shown in ChatGPT search answers, though can still appear as navigational links" |
| GPTBot | Crawls content that may be used to train OpenAI's models | Signals that the site's content should not be used in training |
| ChatGPT-User | Visits a page when a user's request needs it; not used for automatic crawling | Because these visits are user-initiated, OpenAI says "robots.txt rules may not apply" |
OpenAI says each setting is independent: "a webmaster can allow OAI-SearchBot in order to appear in search results while disallowing GPTBot."
How do I let ChatGPT search reach my site?
- Check robots.txt. Make sure it doesn't disallow
OAI-SearchBot. - Check your firewall and CDN. OpenAI recommends "allowing requests from our published IP ranges." Google gives similar advice for its own crawler, listing crawling allowed "by any CDN or hosting infrastructure" alongside robots.txt.
- Decide on training separately. Allow or disallow
GPTBotbased on whether you want your content used in model training. It doesn't affect ChatGPT search. - Use noindex for pages you want kept out. OpenAI says that in ChatGPT Atlas, a page blocked in robots.txt can still appear as a link and page title if OpenAI finds the URL elsewhere. The noindex meta tag prevents this, as long as the crawler is allowed to read the page.
What content does ChatGPT cite?
OpenAI doesn't publish citation criteria. Its guidance focuses on access: "make sure you aren't blocking OAI-SearchBot." Beyond access, the practical approach is the same as for any search engine: pages that answer a customer's question directly, in text, with accurate facts about the business.
For a moving company, that means pages that state plainly what the company does, where it works, and how its process and pricing work, plus accurate details on its Google Business Profile and other listings. Google's guidance for its own AI features says the same fundamentals apply, with "no additional requirements." See what is answer engine optimization.
Who does ChatGPT cite for moving-company questions?
In a Mover Marketing AI test on September 27, 2026, ChatGPT (gpt-4o with web search) and Perplexity (sonar) were each asked 12 questions moving company owners ask about marketing and running their business, through DataForSEO's LLM response API. The 24 answers cited 189 different domains. 153 of those domains were cited only once.
The most-cited domains were smartmoving.com and support.google.com (7 answers each), then reddit.com and elromco.com (6 each). No single marketing agency appeared in more than 4 answers.
This is a small sample from one day, and answers change between runs. It does show that for these questions, citations were spread across many sites rather than concentrated in a few.
How do I know if ChatGPT is sending visitors?
ChatGPT adds the parameter utm_source=chatgpt.com to referral links. In Google Analytics or a similar tool, visits from ChatGPT search show up under that source.
What should a business be wary of?
Claims that a vendor can guarantee a recommendation in ChatGPT's answers. OpenAI's publisher guidance for search answers describes crawler access, not a way to buy a place in them. Tactics that try to plant instructions or biased content for AI systems to repeat are a separate risk, covered in our take on AI recommendation poisoning.