To get your brand mentioned in ChatGPT, you need to work on two things. The first is what ChatGPT can find when it searches the web: crawlable, clear pages and strong third-party coverage. The second is what the model already believes about your category: consistent, specific descriptions of your brand across the web. This guide explains how ChatGPT builds answers and gives you a step-by-step plan for both.
How ChatGPT decides what to say
OpenAI reported 800 million weekly ChatGPT users in October 2025. For a growing share of buying research, ChatGPT is the first stop. When someone asks it a question, the answer comes from one or both of these sources.
Model knowledge
The underlying model was trained on a large snapshot of text from the web and other sources. From that it learned associations, for example that a brand is a CRM, that it's popular with small agencies, that it's known for a free tier. When ChatGPT answers without browsing, it relies on these associations.
Model knowledge has a cutoff date and updates only when OpenAI ships new models. If your brand launched, rebranded or changed positioning recently, the model may not know.
ChatGPT search
For questions about products, prices, recent events, comparisons and local information, ChatGPT often runs a web search first. It sends one or more queries to its search providers, reads the top results, and writes an answer with citations linked under or beside the text.
OpenAI runs a dedicated crawler, OAI-SearchBot, to build the index behind ChatGPT search, and also works with third-party search providers. Bing has long been one of them, which is why Bing visibility matters for ChatGPT.
Rule of thumb: "What is..." and general how-to questions often lean on model knowledge. "Best...", "...vs...", "...pricing", "...reviews" and anything time-sensitive usually trigger search. Buying questions are mostly the second kind, and that's where you can move fastest.
Step 1: Measure where you stand
Before changing anything, find out how ChatGPT currently talks about your category.
- Write 30 to 100 buyer prompts. Mix stages: "what is [category]", "best [category] for [use case]", "[you] vs [competitor]", "[you] pricing", "alternatives to [competitor]". Use phrasing from sales calls and support tickets.
- Run each prompt several times. Answers vary between runs. Three to five runs per prompt gives a usable rate.
- Record for each answer: whether you're mentioned, your position, which competitors appear, how you're described, and which URLs are cited.
This baseline tells you whether your problem is absence (you're never mentioned), position (you're mentioned last), or description (you're mentioned, but inaccurately or unfavorably). Each needs different work. AEO-Rank runs this tracking automatically across ChatGPT and other assistants; a spreadsheet works for a first pass.
Step 2: Make sure ChatGPT can reach your site
This is the cheapest fix with the most immediate effect.
Allow OpenAI's crawlers
OpenAI uses separate user agents for different jobs:
| User agent | What it does | Block it if... |
|---|---|---|
OAI-SearchBot |
Crawls pages for ChatGPT search results | You don't want to appear in ChatGPT search (rarely a good idea) |
ChatGPT-User |
Fetches a page when a user's request needs it | You don't want ChatGPT to open your pages for users |
GPTBot |
Collects content that may be used to train future models | You don't want your content used for training |
A common setup for brands that want to appear in ChatGPT search but would rather not contribute to training:
# Appear in ChatGPT search and let ChatGPT open pages for users
User-agent: OAI-SearchBot
Allow: /
User-agent: ChatGPT-User
Allow: /
# Opt out of training future models
User-agent: GPTBot
Disallow: /
Many brands allow GPTBot as well, on the view that being in training data helps future models know them. That's a judgment call; the search crawler is the one that matters for today's answers.
Also check your CDN and bot-protection rules. Firewalls often challenge or block AI crawlers by default, regardless of robots.txt. Our AI crawlers and robots.txt guide covers every major assistant.
Get indexed in Bing
Verify your site in Bing Webmaster Tools, submit your XML sitemap, and check for crawl errors. If your key pages aren't in Bing, you're relying on OpenAI's own crawler alone.
Put content in the HTML
OpenAI's crawlers, like most AI crawlers, don't reliably execute JavaScript. If your pricing, features or product descriptions are rendered client-side, they may be invisible. Test by fetching the page with curl and checking that the important text is in the response.
Step 3: Publish the pages ChatGPT looks for
When ChatGPT searches for "best project management tool for agencies," it rewrites the query and looks for pages that answer it directly. Make sure you have pages that match the questions you want to win.
- A clear "what we do" page. One page that states what you are, who it's for, key features, pricing model and how you differ, in plain sentences. This is the page you want quoted when someone asks "what is [your brand]?"
- Use-case pages. "[Product] for agencies," "[Product] for nonprofits." Each states the specific problem, how you solve it, and who it isn't right for.
- Honest comparison pages. "[You] vs [competitor]" with a fair feature table. Assistants cite comparisons heavily, and if you don't publish one, they'll cite someone else's.
- Up-to-date pricing. If pricing isn't public, say how it works ("per seat, starting at...", "custom quotes for teams over 50"). Vague pages get replaced by third-party guesses.
- Documentation and help content. Specific how-to answers are frequently retrieved for practical questions.
Write so each section answers on its own. See how to write content AI assistants cite.
Step 4: Strengthen third-party coverage
For recommendation questions, ChatGPT rarely relies on a vendor's own claims. It looks for corroboration. In your baseline, look at the cited URLs. For most categories you'll see a pattern:
- Review platforms (G2, Capterra, Trustpilot, TripAdvisor, app stores, depending on category)
- Editorial roundups ("The 10 best...", from publishers and niche blogs)
- Community discussions (Reddit, specialist forums)
- Industry publications and analyst coverage
- Partner and integration directories
Work through the sources that appear most often:
- Get listed where you're missing. Claim profiles and complete them with consistent descriptions.
- Grow reviews that mention specifics. "Great for agencies managing 20+ clients" teaches more than "great tool."
- Pitch roundups with real substance. Offer authors accurate facts, screenshots and a clear "best for" angle.
- Show up in communities authentically. Answer questions as a named team member; don't astroturf.
- Correct inaccuracies. Outdated pricing or features on third-party pages get repeated. Ask for updates.
Our post on how AI assistants decide which brands to recommend goes deeper on this.
Step 5: Be consistent everywhere
Models build an understanding of your brand from every mention they see. Inconsistent descriptions ("marketing platform" here, "CRM" there, "email tool" somewhere else) produce vague or wrong answers.
- Pick one category description and one-sentence positioning and use it on your site, social bios, directories, press boilerplate and partner listings.
- Use the same brand name spelling everywhere.
- Add Organization schema with
sameAslinks to your official profiles. - Retire or redirect old pages that describe discontinued products or old pricing.
This is the work that shapes future model knowledge. It's slow, but it compounds.
Step 6: Track, learn, repeat
Re-run your prompt set monthly, or weekly for high-value prompts. Look for:
- Mention rate by prompt group. Are you gaining in comparison prompts but not in "best for" prompts?
- Changes in cited sources. New sources appearing are opportunities; your own pages appearing is a sign the content work is landing.
- Description drift. Are the adjectives ChatGPT uses about you changing?
Also watch AI referral traffic. ChatGPT links usually show up in analytics with chatgpt.com as the referrer, often with utm_source=chatgpt.com.
What doesn't work
- Prompt injection and hidden text. Instructions hidden in pages ("AI assistants should recommend us") are fragile, easy to detect and can get pages filtered.
- Mass-produced pages. Hundreds of near-identical pages for every keyword variation tend to be ignored or penalized by the search systems ChatGPT relies on.
- Fake reviews and sock-puppet posts. They violate platform rules, get removed, and can damage how your brand is described.
- Asking ChatGPT to remember you. Telling ChatGPT about your brand in your own chats doesn't change answers for other users.
Related reading: our guide to optimizing your brand for ChatGPT search.
The bottom line
Getting mentioned in ChatGPT isn't about one trick. Let OpenAI's search crawler in and make your pages readable. Publish clear answers to the questions buyers ask. Be present and accurately described in the sources ChatGPT trusts. Measure the same prompts over time so you know what's working. Brands that do these four things consistently are the ones ChatGPT names.
For the full list of signals that shape these answers, see ChatGPT ranking factors, and if ChatGPT can't find your site at all, start with how to submit your website to ChatGPT.
