What the public docs say about ChatGPT and Perplexity sources

Introducing ChatGPT search is dated 31 October 2024. It says ChatGPT can choose to search the web based on the question, and that answers include links to web sources. A person can also turn search on. The post describes the search model of that time as a fine-tuned GPT-4o, and says ChatGPT search uses third-party search providers plus content provided directly by partners. The post does not name those third-party search providers, and it does not list Saudi Arabia or the UAE.

The update dated 5 February 2025 says that, in regions where ChatGPT is available, search is available to everyone without signup. That is a statement about regional availability, not a country list. The post also says any website or publisher can choose to appear in ChatGPT search. The crawler control for that choice is in a separate document. The post is not a claim that publishing a page produces a citation.

The web search guide describes the API, not the chatgpt.com screen. It says web search lets the model use current information from the internet and answer with sourced citations. By default the answer includes inline citations for URLs in the search results. A url_citation carries the URL, title, and location. The sources field lists URLs the model consulted, and the guide says that list is often longer than the inline citations.

The guide says search_context_size does not set an exact token count and does not guarantee a specific number of sources or citations. user_location.country is a two-letter ISO country code. The guide’s examples are US and GB. The field is not a brand-visibility rank.

The crawler overview separates the jobs. OAI-SearchBot is for search. The page says a site that opts out of OAI-SearchBot will not be shown in ChatGPT search answers, though it can still appear as a navigational link. GPTBot is for crawling that may be used to train foundation models, and that switch is independent of search. ChatGPT-User fetches pages when a person acts in ChatGPT or a custom GPT. The page says robots.txt rules may not apply, and that ChatGPT-User is not used to decide whether content may appear in Search.

The same page says a robots.txt change for search can take about 24 hours to be picked up. Blocking GPTBot is not the same as leaving ChatGPT search. Allowing GPTBot is not a citation.

The Perplexity crawler page says PerplexityBot is used to surface and link sites in Perplexity search results, and is not used to crawl content for AI foundation models. Perplexity-User may visit a page when a person asks a question, and may include a link in the answer. The page says that, because a person requested the fetch, this fetcher generally ignores robots.txt. It is not used for automatic crawling or for training foundation models. The page says a robots.txt change may take up to about 24 hours to show up.

The Sonar API schema describes citations as the URLs of sources used to generate the response, and the field may be null. search_results are the results used as context. search_language_filter uses ISO 639-1 codes, with the examples en, fr, and de. language_preference is a separate ISO 639-1 code for the response language. user_location.country is documented as an ISO 3166-1 alpha-2 code for search personalization. When disable_search is true, the schema says the model answers from training data and does not search the web.

None of the documents above publishes a complete, reproducible ranking formula, or an official metric named for a brand’s visibility in the UAE. What they do specify are citation fields, location or language parameters, and which crawler is tied to search or training. A measurement record has to name the surface: the product, OpenAI’s web_search tool, or Perplexity’s Sonar API. Those are not one record.