Guide · ChatGPT ranking
To rank in ChatGPT your site has to be readable by OAI-SearchBot, the crawler behind ChatGPT search, answer the question in its first lines and be named by the sites ChatGPT reads. There is no submission form: ChatGPT cites the pages it finds when it searches the web, and names the brands it learned from its training data. This guide covers both routes, the six steps that get your pages found and how to check whether ChatGPT already cites your site.
Ranking in ChatGPT takes two things. First, ChatGPT search must be able to read and cite your site: OAI-SearchBot and ChatGPT-User allowed in robots.txt and in the firewall, text present in the HTML without JavaScript, the answer in the first lines and structured data that matches the text. Second, the brand must be named by third-party sources such as news sites, comparison articles, forums and reviews: that is where the model learns it and where search finds it. There is no submission form, and citations cannot be bought: ChatGPT ads are slots separate from the answer.
A ChatGPT answer can draw on two places. Both need work, but different levers move each one.
The text the model was trained on, up to a cutoff date. A brand gets in if, before that date, it was already named on public, widely read sources: Wikipedia, news sites, comparison articles, forums, podcast and video transcripts. After the cutoff it no longer changes: what you do today counts for the next model.
When ChatGPT searches the web it reads pages in real time with OAI-SearchBot and ChatGPT-User and lists the links among the sources under the answer. This is the part you can act on now: crawler access, text readable without JavaScript, the answer up front, structured data.
In order of impact. The first is the minimum condition: without crawler access, nothing else helps. The second and third work on the brand, the last three on pages and measurement.
In robots.txt allow OAI-SearchBot (the ChatGPT search index) and ChatGPT-User (pages opened on a user's request) on every public page, and keep private areas such as /app or /admin disallowed. GPTBot only collects training data: blocking it costs no citations. Then check the firewall: Cloudflare and other CDNs have rules that block AI bots even when robots.txt allows them.
ChatGPT answers 'what is [brand]' in one sentence. Write that sentence yourself and use it everywhere: at the top of the homepage, on the About page and in the Organization structured data. A Wikidata entry, and Wikipedia if the brand qualifies, reinforce it.
When it searches, ChatGPT reads comparison articles, best-of lists, forums such as Reddit, reviews and trade press. One page on your own site is not enough: you need several independent sites that name you next to your competitors. Start from the pages ChatGPT already cites for your sector's questions and ask to be included.
One page, one question. Open with a 40-80 word paragraph that answers on its own, then headings that are real questions, lists and tables. Add FAQPage, Article and BreadcrumbList that match the visible text: these are the pieces ChatGPT search lifts cleanly into an answer.
The questions that bring customers look like 'best X for Y' and '[competitor] alternatives'. Publish a comparison page for every competitor your customers name, and a buyer's guide to your category with stated criteria, in which you appear too.
Ask ChatGPT your customers' questions with search on, at regular intervals, and note who gets named and which sites are among the sources. Answers change on every run, so rely on several repetitions rather than a single test. GeoSuite does this across ChatGPT, Gemini, Claude and Perplexity and shows the questions where competitors appear and you do not.
When ChatGPT searches the web, it picks the pages to cite by looking at a few things.
Server-rendered HTML, a valid sitemap.xml, text that does not depend on JavaScript. An answer that only appears after rendering in the browser often goes unread.
Organization, Article, FAQPage, HowTo and BreadcrumbList that match the visible text. They help tell who the brand is and where the answer sits, without parsing prose.
How many independent sites name the brand and link to its pages. ChatGPT cross-checks several sources, so a name that recurs across many sites weighs more than a single link.
A short paragraph at the top that answers the title's question. Pages that hide the answer under the introduction lose to pages that give it right away.