Why isn't my brand showing up in ChatGPT?

Start with the crawler, it's the quickest to rule out

Before anything subtle, find out whether ChatGPT can physically read your site. When it answers a live question it fetches pages with named crawlers, mainly GPTBotfor training and OAI-SearchBot for search. A page it can’t fetch is a page it can’t quote.

Open yoursite.com/robots.txt and read it. Look for User-agent: OAI-SearchBot or User-agent: GPTBot paired with Disallow: /. Block OAI-SearchBot and ChatGPT’s live search can’t reach you at all. Block GPTBot and you also keep yourself out of the next training run. Plenty of sites added those lines back in 2023 to stay out of AI training, forgot about it, and now wonder why they’ve gone quiet. If you find a block you didn’t intend, remove it and wait for the next crawl.

The model has too little to work with

Say the crawler reads you fine and you’re still absent. The next suspect is volume of evidence. Models tend to name what they’ve seen described, reviewed, and discussed many times over. A brand that launched last quarter, or one with a light footprint, hands the model almost nothing. Nothing is penalizing you for being small. The model just hasn’t seen enough to put you forward with any confidence.

Stale training data stacks on top of that. The version answering you today was built on a snapshot of the web from months ago, sometimes longer. If you rebranded, shipped a new product, or only found traction this year, the model’s built-in knowledge may predate all of it. Live search patches some of the gap, which is the whole reason the robots.txt check above comes first.

The model isn't sure what you are

Now suppose the crawler reads you and you’ve been around a while. You can still get skipped when your own story doesn’t line up. Your homepage calls you a customer data platform. Your G2 profile says marketing analytics. An old directory listing still has the founder’s last company in the name field. To a person, those are close enough. To a model weighing whether you’re the right answer to a narrow question, the mismatch reads as doubt, and doubt loses to a competitor who describes themselves the same way in every place.

Settle on one plain sentence for what you are and who it’s for. Then make every source the model reads say it: your site, your review profiles, your social bios, your directory listings. Consistency is boring to maintain and it does more than most page edits.

You're missing from the sources it trusts

The part founders tend to resist: a lot of what a model “knows” about your category never came from your site. It came from G2 and Capterra, from Reddit and Hacker News threads, from the occasional writeup in an established publication. Those sources cleared the training filter, and its live search leans on them too. If competitors are all over them and you’re nowhere, you can lose the recommendation with the stronger product and the cleaner homepage.

You can’t write these into existence on your own domain. You earn them. Get listed and reviewed where your buyers already compare tools, show up in the threads where your category gets argued about, give a writer a concrete reason to name you. Watching how those brand mentions accumulate is slower than shipping a landing page, and it’s usually what changes the answer.

Your answer is on the page, just buried

Sometimes the content is right there and the model still can’t use it. Models lift short, self-contained passages, not whole pages. If the answer to “what’s the cheapest option for a two-person team” is scattered across a pricing table, a feature grid, and a paragraph that starts three scrolls down, there’s nothing clean to pull.

Write the answer as a sentence a model could copy with no surrounding context, and put it near the top of the page. Back it with a number or a source someone could check. That’s most of what makes a passage worth an AI citation instead of getting skimmed past.

Fix it in the order that pays off

Roughly, work top to bottom:

• Unblock the crawler.Read robots.txt and drop any OAI-SearchBot or GPTBot disallow you didn’t mean. This is the one that can hide everything else.
• Make your entity info consistent. One description of what you are, repeated wherever the model reads about you.
• Publish extractable answers. Plain, self-contained answers to the questions buyers actually ask a chatbot, placed near the top of the page.
• Earn third-party presence.Reviews, community threads, press. The slow, durable part, and the one competitors can’t easily copy.
• Measure across engines, and watch the trend. A single answer wobbles between attempts. The pattern over a few weeks is the signal.

If you want to go further on the technical side, some sites now publish an llms.txtfile, a plain-text map pointing models at their key pages. It’s an emerging proposal, not a requirement, but it’s cheap to add once the basics above are in place.

Then keep watching

Whatever you change, don’t grade it off one reply. Ask the same buyer questions a handful of times, across ChatGPT, Gemini and Perplexity, and note who gets named and how often. A free AI visibility audit runs that first sweep for you and shows where you land today. After that you’re tracking a number that moves, which beats reloading ChatGPT and hoping.