How to get your company cited by ChatGPT

Someone asks ChatGPT for the best company to build their product, and three or four get named. Being one of them is worth more than a page of search results, because the shortlist has already been made for them.

Almost every guide to this starts with how to write your content. That is genuinely the third most important factor. Two things come first, and if either is wrong the writing advice is wasted.

First: can the crawler reach you at all?

Retrieval crawlers fetch your page at the moment the question is asked. If they get a 403, you cannot be cited, no matter what the page says. Check with one line per crawler:

curl -s -o /dev/null -w '%{http_code}' -A "Mozilla/5.0 (compatible; OAI-SearchBot/1.0)" https://yourdomain.com/

Run it for OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot and GPTBot. Anything other than 200 is a block, and it is usually a CDN default nobody chose — we found exactly that on this site, with a permissive robots.txt and a firewall returning 403 to every one of them.

The second reachability question is whether your content is in the HTML. Many crawlers run little or no JavaScript, so fetch your own page with curl and read what comes back. If the text only exists after hydration, assume it does not exist.

Second: does anyone else mention you?

This is the uncomfortable one. When a model answers "best X companies", it is drawing on pages about the category — listicles, directories, forum threads, comparison posts. Your own site claiming you are excellent carries almost no weight, because every site claims that.

Search your target question and look at what actually ranks. For "best AI software development companies" the entire first page is agency blogs, each listing itself near the top. That is the corpus the model reads. A company absent from all of it does not appear in the answer.

What to do about it, in rough order of effort:

1. Claim free directory listings — Clutch, G2, GoodFirms, DesignRush. Unglamorous, and they are cited constantly.
2. Publish your own comparison page for the category. Everyone ranking for these queries does it. Disclose that you are on your own list — ours does.
3. Answer questions substantively where your buyers ask them: Reddit, Hacker News, Stack Overflow, Quora. An unlinked mention still counts, because models ingest the text either way.
4. Ask to be added to existing listicles. Many authors accept submissions; some will trade a mention.
5. Publish something with data nobody else has. First-hand numbers get quoted and re-quoted, which is how a single page becomes many mentions.

Third: write in passages a model can lift

Now the writing advice, which is real but only matters once the first two are true. A model quotes paragraphs, not pages, so the unit of optimisation is the passage.

• Lead with the answer, then justify it. First sentences carry disproportionate weight.
• Repeat the subject instead of using pronouns. "Remotepuzzle builds…" survives extraction; "We build…" does not.
• Use headings that are the question someone types, not clever labels.
• Put specifics in the text — numbers, ranges, named technologies. Vagueness is never quoted.
• Make each section answer one question completely, so it survives being read alone.

Mark up your FAQs with FAQPage schema. A structured question-and-answer pair is the most directly reusable format there is, and it is cheap. An llms.txt helps a model that has already found you describe you correctly, though it does nothing to help it find you.

Be a consistent entity

Models answer about things they can identify. If your name, one-line description and contact details are identical everywhere — your site, your schema, every directory — you are far easier to recall as a single entity than a company whose details vary by listing.

Pick one canonical description and use it verbatim. Put Organization schema on every page, not just the home page, so any single page a crawler reads states which entity it belongs to.

How to tell whether it is working

Citations usually produce no referrer, so the honest method is to ask directly and keep a record.

1. Write down your ten highest-intent buyer questions.
2. Run them through ChatGPT, Claude and Perplexity on the first of every month. Record whether you are named and which sources are cited.
3. Track the sources as much as your own appearance — they are the pages you need to be on.
4. Watch for referrals from chatgpt.com and perplexity.ai in analytics. Small numbers, but they convert unusually well.
5. Track branded search volume in Search Console. People who meet you in an answer often search your name afterwards.

Expect months, not weeks. The model has to encounter you repeatedly, across sources it did not get from you. That is the whole mechanism, and it is also why it is defensible once it works.

This is what our AI-search SEO work covers, and GEO versus SEO explains why the two need different tactics.

Frequently asked questions

References

1. OpenAI bots and user agents — OpenAI
2. FAQPage structured data — Google Search Central
3. The /llms.txt proposal — llmstxt.org