A friend who runs a new legal firm called me last month, half annoyed and half spooked. She’d typed “best legal consultant for startups in Singapore” into ChatGPT, expecting to at least see her name somewhere. She got three firms. None of them were hers. Two of them, she told me, were noticeably worse than her.
It isn’t losing a pitch — it’s finding out that you were never in the room..
More clients are starting the practice now. Not with a search engine and ten blue links, but with a question and an answer that comes back in an AI assistant. If you’re not in that answer, you’re not in the game at all. So it’s worth understanding what’s actually going on when an AI assistant ignores you.
It usually isn’t that you’re worse
AI assistants don’t hand out recommendations on values, or at least not values as you’d define it. They can’t try your product. They only “read” about it. What they’re really doing is assembling an answer out of whatever the internet has already said, and that leans heavily toward whoever has the longest paper trail — the roundups, the reviews, the comparison posts, the years of being mentioned.
Which means a two-year-old firm with a thin footprint loses to a fifteen-year-old firm with a thick one, even when the new firm is way better. It’s an incumbency effect, not a quality judgment. Worth remembering before you conclude something is wrong with the business.
The website tells human what you do. It might not tell a machine.
Most sites are written for a human who’s already curious when they land at your page. The homepage says something like “Professional legal services for startups,” and a person scrolls and explores the information from context. A model reading that page gets almost nothing it can use. What legal services? Professional? Doing what? And for what types of startups?
The stuff that AI assistants need is boring and specific. What you do, who it’s for, where you operate, how much it costs, all types of “natural” questions that a customer may ask, all in plain words. A lot of businesses have never written any of that down in one place, because everyone internally already knows it.
The other half of this is structure. Schema markup, clean headings, an FAQ that answers real questions in complete sentences. All of these make a page easier to parse and quote. If your best explanation of what you do lives inside a PDF or an image, treat it as invisible.
You’re missing from the places that get quoted
This is the fact that surprises people. What matters isn’t only what’s on your site — it’s what exists about you elsewhere.
AI assistants lean on third-party sources, because independent source is worth more than self-description. Industry directories, review platforms, local listings, forum threads where somebody asked for a recommendation – a model builds its picture of a market from all of it.
If your competitors turn up across a dozen of those and you turn up in two, the gap has nothing to do with your website. It’s a citation gap, which takes time to close because you can’t do it unilaterally. Your work turns out to be have people who don’t work for you talk about you, appreciate your service.
An interesting change worth to note: plain unlinked mentions count for more here than they did in classic SEO. A model learns from language, not just link graphs. Someone naming your company in a forum thread is a real signal even with no hyperlink attached.
Nobody can tell whether you’re still open
Freshness matters more than it should. A model reading a site with a copyright notice from two years ago, a blog that stopped update in mid-2024. AI may have a natural question: is this business still operating? When it’s unsure, it hedges by recommending something it’s confident about.
Dates, updated pages, current pricing, recent activity. These are small things but clearly signal that someone is still at home.
What I’d actually do about it
If I were in my friend’s position, I’d start narrow rather than trying to fix everything.
Ask AI assistants the questions your customers would ask — the real phrasings, with the location and the qualifier attached — and write down who is recommended. That list is your actual competitive set. Then look at what those firms have that you don’t. Usually it’s obvious within an hour: they’re in three directories you’ve never joined, they’ve got a comparison page you don’t, someone wrote them up.
First fix your own pages so a machine can describe you without guessing. Then go earn mentions in the places that clearly get read.
None of that is fast. That’s the honest catch — organic AI visibility compounds, and compounding takes time, which is precisely why it favors whoever started earlier. Some businesses can wait it out. Others need to be in the answer for a specific query this quarter, and for them paid placement inside AI assistant is the shortcut — as long as it’s clearly labeled as sponsored and priced on results rather than impressions.
Either way, the first move is the same, and my friend has already made it: find out what AI says when someone asks for a business like yours. Next rebuild your website, work on mentions and reviews by authorities, or find paid services.
FAQ
Why is ChatGPT recommending businesses that seem like a worse fit than mine?
ChatGPT may have more public evidence for those businesses, even if they are not objectively better. Long-standing competitors often have more reviews, directory profiles, comparison pages, press mentions, and customer discussion that AI systems can retrieve and use. Being left out is often an evidence gap, not proof that your business is less capable.
How do I find out what ChatGPT is getting wrong or missing about my business?
The fastest way to find out is to test the same detailed prompts your customers would use and compare the answer with your actual offer. Look for facts ChatGPT misses, gets wrong, or cannot explain: who you serve, where you operate, what you provide, how you are different, and what a customer can do next. Those gaps show what your public information is failing to make clear.
Is ChatGPT missing information about my business, or does it lack reasons to trust it?
It may be either, and the fix depends on which evidence is missing. If an AI assistant cannot accurately describe what you do from your own pages, you have a content and clarity problem. If it understands your business but keeps recommending competitors with stronger reviews, directories, comparison coverage, or industry mentions, you have a credibility and corroboration problem.
Could outdated information make my business look inactive or unreliable to ChatGPT?
Yes. Old pricing, outdated service details, an abandoned blog, and no recent signs of maintenance can make a business appear less reliable or even inactive. Keep the information customers rely on current, visible, and consistent across your website and public profiles.
I need to show up for an important query, but organic visibility seems very slow. What should I do?
Treat organic visibility and short-term demand as two separate workstreams. Keep improving your site and earning credible third-party evidence, because organic AI visibility compounds over time rather than changing immediately. If you need visibility for a high-value query sooner, evaluate clearly labelled, outcome-focused paid placement while continuing to build the organic foundation.
