No form gets your business into ChatGPT's answers. To get recommended, make four things true: AI crawlers can read your site, you describe your business the same way everywhere, your pages answer buyer questions in their first lines, and other sites mention you. I audited my own site against that list. The crawlers got in fine; the weak spot was my own description of myself.
Who is Lucy Liu? The problem I started with
Say "Lucy Liu" and most people picture the actress. I'm the other one, an operator in Vancouver who builds the systems behind early-stage fintech companies, and hard to find by name.
I typed the question a prospect might ask into ChatGPT and Perplexity: "Who is Lucy Liu, the fractional operator in Vancouver?" The answer called me a fractional operator for regulated fintech founders, "not the actress of the same name," and cited lucyliu.ca for almost every line. It also warned that other Vancouver professionals share my name, realtors among them, and told the reader to sort us out by niche and website.
On a second read, I saw that each phrase came from my own homepage, down to my years at each company and a line naming my current employer. I've decided this site won't name my employer, so that line comes out in this release, and AI answers will keep repeating the old wording until the crawlers visit again.
My question also did half the work: it carried my niche and my city. An AI assistant doesn't know you. It builds an answer from what it can find, and if that material is thin or inconsistent, someone with your name, or a competitor with a sharper description, gets the sentence instead.
How does ChatGPT decide what to recommend?
It draws on two places: what the model absorbed in training, and live search. For live search, OpenAI documents a crawler called OAI-SearchBot and says sites that opt out of it won't be shown in ChatGPT search answers. The same page says robots.txt changes take about 24 hours to reach its systems.
GPTBot collects content for training, so blocking it is a training decision, separate from search. ChatGPT-User fetches a page when a person asks ChatGPT to look something up.
Yoast says ChatGPT search uses Bing's index along with other sources, and tells site owners to check their indexing in Bing Webmaster Tools. OpenAI's crawler page doesn't mention Bing at all, so I treat the strong version ("not in Bing, never in ChatGPT") as unproven. Checking costs nothing.
What I checked on my own site
I started with access. I requested my homepage, a blog post and my llms.txt file under nine crawler names: GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot, Bingbot, Googlebot and CCBot. All nine got a 200 response, and my robots.txt allows everything.
Three problems were mine to fix.
- My llms.txt file, which I updated by hand, listed 11 posts. The blog has 14 live. The three newest were missing.
- My Person schema, the machine-readable description of me on the homepage, linked to one other page: LinkedIn. Apart from a job title and a city, nothing in it told a machine which Lucy Liu I am.
- My post titles were written to be clever. I like "Verification Is the Job" as a title. Nobody types it into ChatGPT.
All three are my doing. I build by directing AI tools, and I check that every page loads before it ships. I never checked how the site described me to a machine. I had run one test and skipped the other.
The fixes ship with this post. The build that publishes the blog now writes llms.txt from the same list of posts, so the file can't fall behind again. The schema now carries a one-sentence description of me, in close to the words the AI used back to me. Profile links are still open: the schema points to LinkedIn and nothing else. New posts, starting with this one, open with a direct answer and close with a short FAQ.
What should you do this week?
Run these in order. Most of it fits in an afternoon; step 4 takes longer.
- Test the crawlers. Request your homepage under the user-agent names above and confirm each gets a 200. Then read your robots.txt, your firewall rules and any bot-blocking switch on your CDN. A rule someone added to stop spam can shut out the crawlers you want.
- Write one description of your business and use it everywhere. Same name, same one-sentence description, same category on your site, LinkedIn, directory listings and guest posts. If you share a name with someone famous, put your niche and city in that sentence.
- Add that description and your profile links to your schema. Google's guidance on AI features says you need no special files and no special schema to appear in them, and that your structured data should match the visible text.
- Rewrite your most-visited pages answer first. Pick the question a buyer would type. Put the answer in the first two sentences, in plain text, with your business name in the same paragraph as the fact. Do three pages before you do thirty.
- Get mentioned somewhere you don't control. Directories, partner pages, a guest post, a podcast. ClickForest cites an Ahrefs study of 75,000 brands in which mentions predicted visibility in Google's AI Overviews better than backlinks did. That is a correlation, and I leave out the vendor numbers I couldn't check.
- Verify Bing and Google. Add your site to Bing Webmaster Tools, submit your sitemap, and do the same in Google Search Console.
- Ask the questions yourself, monthly. Type what your buyers would type into ChatGPT and Perplexity. Write down what comes back and which sources it names. That log is the closest thing to a measurement you will have.
What to skip
Don't pay for an "AEO package" that promises a ranking. No agency controls the answer, so no agency can promise one.
Don't treat llms.txt as a ranking trick either. Vendor log studies found most llms.txt files get no requests at all, and Google's John Mueller called the format "purely speculative for now" in June. I keep mine because it takes about an hour and helps a person or an agent find my best pages. I don't expect it to change a single answer.
Two guides I read are worth your time. ClickForest's covers reviews and outside mentions, and the agency tracks how often ChatGPT names it. Superblog's covers keeping your business details identical across the web. Neither covers sharing a name with someone famous, or with a competitor.
Takeaway
Before you buy anything, spend an afternoon on the crawler test and start the monthly question log. You will learn whether your problem is access, description or reputation. Mine is description, and most of my fix list is plain text.
My checking routine for AI-built work is in Verification Is the Job, and the site itself came from How I Shipped Our Website Without Writing Code.
If you're a founder who wants your company to show up when buyers ask an AI, and you'd rather have someone audit the site and fix it with you, book a call. I run the same checks on your site first, so we start from what the crawlers see.
FAQ
Can I pay to get my business recommended by ChatGPT?
Not through any channel OpenAI documents that I could find. Agencies sell "AEO" packages, but none can guarantee a mention. Spend the money on clear pages, accurate listings and outside coverage first.
Should I add an llms.txt file?
It is cheap and harmless, so add one if you like. Don't expect it to change what ChatGPT says about you. Google says its AI features need no special files, and the log studies I read show few crawlers asking for it.
How long until ChatGPT mentions my business?
I can't give you a number. OpenAI says robots.txt changes take about 24 hours to reach its search systems; after that, timing depends on how often crawlers visit your pages and who else mentions you. So far, no one has told me they found me through an AI tool.