How to rank in ChatGPT: what it reads, what it cites, and how to check

How do I rank in ChatGPT?

You ask ChatGPT to suggest a business like yours in your area, and it names two competitors and a directory. This guide covers what decides which businesses it names, and what you can change.

ChatGPT can answer in two ways, and they work differently:

• From training: the model answers from text it learned before it was released. OpenAI’s GPTBot crawler collects content that may be used for that training. You can’t see what a model learned about you, and it changes only as new models arrive.
• From search: when ChatGPT searches the web, it draws on sites that OpenAI’s OAI-SearchBot crawler has surfaced, writes one answer and links some of the pages it used. This is the part you can work on.

“Rank” is a borrowed word here. A ChatGPT answer has no page two: you’re named or you aren’t, and the same question asked twice can come back with different sources. So the useful measure is how often you appear across a fixed set of buyer questions, which a later section sets up.

Where does ChatGPT get the sites it names?

OpenAI publishes a page on its crawlers (read November 1, 2026). Three of the agents it lists matter here, and OpenAI treats each one separately:

Three things follow from that page:

• Training and search are separate choices: OpenAI says each setting is independent of the others, so you can keep your content out of training and still let OAI-SearchBot surface your site.
• Robots.txt isn’t the only gate: OpenAI also recommends allowing requests from its published IP ranges. A firewall or CDN rule that turns away unfamiliar bots can block OAI-SearchBot even when robots.txt lets it in.
• A ChatGPT-User visit isn’t a ranking signal: seeing that agent in your server logs means someone’s question led ChatGPT to your page. OpenAI says the agent isn’t used to decide whether content appears in search.

How do I rank in ChatGPT search?

Once the crawler can reach your site, the page itself decides whether it’s worth citing. Take a plumbing company that wants to be named when someone asks whether a tankless water heater is worth it in a house with hard water. A page ChatGPT can cite for that question:

• Answers it at the top: the first two sentences say yes, no or it depends, and on what, before any background.
• Says it in text: the answer sits in the page’s text, not only in an image, a PDF or a video, so a crawler can read it.
• Carries specifics nobody else has: how often the company flushes units on local water, which models it fits and what happens when the flushing is skipped. A model has already read the generic version many times.
• Is dated: facts that change, such as prices, rebates and code rules, carry the month they were checked.
• Matches everything else about the business: the name, services and service area on the page are the same as on its Business Profile, directory listings and social profiles.

Write one page for each question buyers actually ask, rather than one long page that touches on twenty. A page about one question is easier to quote, and easier to update when the answer changes.

Why doesn’t my brand show up in ChatGPT?

When ChatGPT leaves you out, five causes are worth checking. Work down them in order: the first is a setting you can check today, and the rest take longer.

A quick test shows where you stand. Ask ChatGPT about your business by name, then ask three questions a buyer would ask without naming you, and note the sources it links each time. A wrong fact in an answer often traces back to one of those pages, and correcting it there is the fix that lasts.

Which other sites does ChatGPT cite about your field?

ChatGPT’s answers about a kind of business draw on what other sites say about it. In Ahrefs’ US data for October 30, 2026, ChatGPT answers cited reddit.com 151,313 times and forbes.com 17,065 times, across every topic. Each field also has directories of its own. Take web design agencies, where three agency directories turn up:

These are Ahrefs’ counts of citations in US AI answers, October 30, 2026. Ahrefs tracks a sample of questions, not every answer, so read them as a direction. For a dentist or a roofer the directories differ, and the same checks apply:

• Be listed, correctly: claim your profile on the directories your field uses, and keep the name, services and contact details identical to your site’s.
• Earn reviews the honest way: ask real clients to review you where buyers look. Never buy reviews or write your own.
• Answer questions where they’re asked: a useful reply in a forum thread about your field, from a named person and without a sales pitch, is the kind of page these counts show answers drawing on.
• Tie your profiles to your site: Google’s documentation for Organization markup lists a sameAs property, the URL of a page on another site with more information about your organization (September 2026). Listing your directory and social profiles there connects them to your site in a form machines can read.

How does a small website rank in ChatGPT?

A small site is unlikely to beat Reddit or Forbes on broad questions, and it doesn’t need to. It can answer narrow ones in full:

• Pick questions with a place or a specialty in them: “is a heat pump worth it in Pasadena” has fewer good answers than “is a heat pump worth it”, and a local installer knows the local answer.
• Say what the big sites leave out: the permit your city asks for, the rebate that ended, the mistake you keep finding on site visits. That’s content only you can write.
• Use one exact form of your name: if another business shares it, pair it with your town or trade on your site and profiles, so an answer can tell you apart.
• Earn real mentions: a local news story, a supplier’s dealer page or a trade association’s member list each puts your name on a site with readers of its own.
• Skip pages written only to catch phrasings: ten near-copies of one page give an answer nothing new to cite.

None of this needs a large site. It needs pages that answer a narrow question fully, with the owner’s own experience behind them.

How do you show up in Perplexity answers?

Perplexity rests on the same base: it searches the web, writes an answer and links the pages it used. Its crawlers are set separately from OpenAI’s, and its documentation (updated October 29, 2026) describes two:

• PerplexityBot: surfaces and links websites in Perplexity’s search results. Perplexity says it isn’t used to crawl content for AI foundation models, and recommends allowing it in robots.txt and letting its published IP ranges through.
• Perplexity-User: visits a page when a person’s question calls for it, and may link that page in the answer. Because a person asked for the visit, Perplexity says this fetcher generally ignores robots.txt.

If your site sits behind a web application firewall, Perplexity says you may need to allow its bots explicitly, and it gives the steps for Cloudflare and AWS. As with OpenAI, each setting works on its own.

In the Ahrefs counts above (October 30, 2026), Perplexity answers cited each of the three agency directories three to four times as often as ChatGPT answers did. Part of that gap may come from how many answers Ahrefs samples on each tool and how many links each answer carries, so treat it as a direction too. Run your question set on both tools, and expect different sources for the same question.

How can you tell whether ChatGPT mentions you?

You can’t see the answers given to other people, so measure with a fixed sample:

1. Write 20 to 30 questions buyers ask: in their words, not as keywords. Include some that name your town or specialty, and keep the list the same each month so the months compare.
2. Ask each one in a new chat with no history: past conversations can steer an answer. Then run the same list in Perplexity.
3. Record four things each time: whether you’re named, whether your site is linked, which competitors appear and which sources are cited.
4. Work out your appearance rate: the share of questions where you’re named. Track it monthly and read it as a trend, since answers vary from one day to the next.

Your analytics add a floor. Visits referred from chatgpt.com and perplexity.ai show people who clicked through, but some visits from AI tools arrive without a referrer and are counted as direct, so the real number can be higher than the report shows.

What can you fix this week?

In order, from settings to content:

1. Check robots.txt: confirm that OAI-SearchBot and PerplexityBot aren’t disallowed, and treat GPTBot as a separate, deliberate choice.
2. Check your firewall or CDN: its bot rules shouldn’t turn away the crawlers robots.txt allows.
3. Ask about yourself: run five buyer questions in ChatGPT and in Perplexity, and note the sources each answer links.
4. Correct your facts: fix the name, services and contact details on the listings those answers cite.
5. Rewrite one page: take the question buyers ask most and put the plain answer in its first two sentences.
6. Start the list: write your 20 to 30 questions and run them once, so next month has something to compare with.

Google’s AI features follow settings of their own, set out in AI Overview optimization, and how this work relates to SEO is in AEO vs SEO vs GEO.

Our answer engine optimization work makes a business easier for AI answers to find, read and cite. It runs a fixed question set each month and reports the appearance rate as a measure, with no promise attached. If you’ve started your list of buyer questions, bring it to a 15-minute call.