Perplexity and ChatGPT can both answer “what is the best tool for this” with a list of names and a set of links, which makes them look alike from the outside. Underneath, they differ in when they search, which crawler feeds them, how they present sources, and how much control a site has. Those differences change what you should measure and what you should fix. This comparison sticks to what the two companies document, and says so where the documentation is silent.
When each one searches
Perplexity is built around retrieval. Its crawler documentation describes PerplexityBot as designed to surface and link websites in search results on Perplexity, and the product’s point is an answer with sources.
ChatGPT is a general assistant with search as one of its tools. OpenAI’s help center says ChatGPT may search the web automatically when a question would benefit from current information, and that people can start a search by hand. When it does not search, the answer comes from the model’s memory and carries no citations.
The practical result is that ChatGPT can name your brand with no link at all, because a memory-only answer has no sources to link. Perplexity is designed to attach sources to its answers. That is why ChatGPT is worth judging on mentions as well as citations.
Where the results come from
For ChatGPT, OpenAI says search uses its own crawler, OAI-SearchBot, and sometimes partners with other search providers. When it does, ChatGPT typically rewrites the question into one or more targeted queries, and after reviewing the first results it may send more specific ones. ChatGPT may also use your IP address to estimate a general location and share that with search providers, so local questions can produce different sources for different people.
For Perplexity, PerplexityBot builds the index behind its results, and Perplexity says it may take up to 24 hours for robots.txt changes to take effect. Both companies tell sites that want to appear to allow the search crawler and the vendor’s published IP ranges.
How citations appear
In ChatGPT, responses that use web search may include citations. Selecting one opens the source, and a Sources view lists the cited pages and other relevant links when it is available. Through the API, OpenAI returns a url_citation annotation with the URL, title and location of each source. OpenAI’s help page also warns that search results and citations can be incomplete, outdated or incorrect, and recommends opening the source to check it.
Perplexity attaches linked sources to its answers. In both products, the thing that is cited is a specific URL. It matters which of your pages that is: a pricing page, a comparison page and a blog post are cited for very different questions.
One question, three possible outcomes
Suppose a buyer asks both products for the best invoicing tool for freelancers. Perplexity returns an answer with linked sources. ChatGPT may return an answer with no sources at all, naming brands it learned about in training, or it may search first and cite pages. The same brand can be missing from one answer, named without a link in the second and cited in the third, all for the same question. This is why a single prompt asked once tells you very little, and why a scan repeats a prompt set across engines.
Crawlers and robots.txt side by side
Both companies separate the crawler that feeds search from the fetcher that acts on a person’s request, and OpenAI also has a separate crawler for training. That split is what lets a site opt out of training without leaving search answers.
| Topic | Perplexity | ChatGPT |
|---|---|---|
| When it searches | Built around search: the product is an answer with linked sources. | Searches automatically when a question would benefit from current information, or when the user turns Search on. Otherwise it answers from memory. |
| Crawler for search | PerplexityBot, which surfaces and links sites in Perplexity search results. | OAI-SearchBot, which surfaces sites in ChatGPT search. ChatGPT also sometimes partners with other search providers. |
| Training crawler | Perplexity lists none. It says PerplexityBot and Perplexity-User are not used to train foundation models. | GPTBot, which crawls content that may be used to train OpenAI’s models. |
| Fetch for a user request | Perplexity-User, which Perplexity says generally ignores robots.txt. | ChatGPT-User, where OpenAI says robots.txt rules may not apply. |
| How sources appear | Linked sources attached to the answer. | Inline citations, plus a Sources list when available. |
Two details deserve attention. Blocking OAI-SearchBot means the site is not shown in ChatGPT search answers, according to OpenAI, while blocking GPTBot does not. And because user-triggered fetchers may ignore robots.txt, a Disallow line does not stop a page being opened when someone asks about it directly. You can see how your robots.txt treats all of these with the AI crawler check.
What this means for tactics
- For both: allow the search crawler, make sure your CDN or firewall lets it through, and write a page that answers each buyer sub-question clearly. The guide to how engines choose sources goes through the steps.
- For Perplexity: its answers are assembled from retrieved pages, so the pages you want cited must be crawlable and clear about the exact question. Watch which URLs it cites for your category.
- For ChatGPT: you are in two games at once. Being retrievable matters when it searches, and being well known across the web matters when it answers from memory. Mentions without citations are normal, so judge it on both signals.
Measure them separately
Do not merge the two into one number. For each engine, track the mention rate and the citation rate on their own. A brand that is cited on Perplexity and only mentioned on ChatGPT has a different problem from the reverse. askvisibly reports every engine in its own row. The free scan asks Perplexity and Claude, and the paid plans add ChatGPT, Gemini and Google AI Overviews across 25 or 40 prompts per scan. See the plans for the details, and how to measure AI visibility for the method.