Gabriel Espinheira
A founder you respect just told you how she picked her new web partner. She opened ChatGPT, typed a version of "who should I hire to fix a site that gets traffic but no leads," and it gave her three names. She called the first one. You do that exact work, for exactly her kind of business, and your name never came up.
That's the new shape of a lost deal, and most owner-operated European businesses have no idea it's happening to them. Getting recommended by AI now decides shortlists you never see. It isn't that an assistant rejected you; it couldn't confirm you existed, so it named the ones it could. This post is why your name is missing, and what puts it back.
TL;DR: Buyers increasingly ask AI assistants who to hire, and those tools name only businesses they can verify from consistent, corroborated facts across the web. To get recommended by AI, make your identity, expertise and track record checkable everywhere you appear, then confirm it by asking the assistants what your buyers ask.
The shortlist now forms before anyone clicks
The vendor a buyer already favours before contacting anyone wins the deal about 80% of the time, according to 6sense's 2025 research, and 95% of the time the eventual winner was on the buyer's shortlist before a single sales conversation. That shortlist is increasingly built inside an AI chat. G2's April 2026 study of 1,076 software buyers, "The Answer Economy", found 51% now start research with an AI chatbot more often than with Google, up from 29% a year earlier.
The decisive moment has moved upstream of your website. By the time a buyer reaches your homepage, they are usually confirming a name they already trust, or they never arrive at all. The same G2 study found 69% of buyers chose a different vendor than they had originally planned on the strength of what a chatbot told them. The room where you win or lose the enquiry is now a conversation, and for most founders it happens without them in it.
This is not ranking, which is why your Google wins don't carry over
Ranking on Google and being recommended by an AI assistant are separate games, and winning the first does not win the second. SOCi's 2026 Local Visibility Index found that appearing in ChatGPT's local recommendations is roughly 30 times harder than ranking in Google's local results, and that fewer than half of the businesses leading Google's local pack surface in AI recommendations at all. A LocalFalcon analysis of nearly 190,000 ChatGPT results was blunter: 83% of restaurants were invisible in ChatGPT, against 14% missing from Google.
Ranking earns you a link a human might click. A recommendation gets you named as the answer, before a click exists. Google indexes individual pages; an assistant assembles a reply from what it can corroborate about a business across many sources. And buyers have moved with the tools. BrightLocal's 2026 Local Consumer Review Survey found 45% of consumers used ChatGPT, Gemini or Perplexity for local business recommendations in the past year, now the third most popular source for finding a business, rising to 64% among 30-to-44-year-olds.
None of this is confined to software buyers, or to the US. The hard numbers here come mostly from American and global samples, but the tools, and the habit, are the same for a buyer weighing up options in Lisbon, Berlin or Dublin.
Your name doesn't come up because AI can't verify you, not because it rejected you
Absence from an AI recommendation is almost never a verdict on your quality. It's a gap in what the model can confirm about you. An assistant builds a recommendation by cross-checking what independent sources say about a business (your own site, directories, review profiles, and the places other people mention you), and it leans on the businesses whose story is consistent and checkable. When your details conflict across the web, or barely appear, the model does the safe thing: it names the competitor whose facts line up.
There is a structural tilt on top of that. A February 2026 paper from Jin and colleagues (arXiv) showed that the order in which sources are retrieved strongly shapes which businesses a generative engine surfaces. Position, not merit alone, decides prominence, which quietly disadvantages smaller and newer players. So two things have to be true at once. You have to be verifiable, and you have to be retrievable. A founder controls the first far more than the second, which is where the work should start.
"On a first call, I'll paste a founder's business into ChatGPT and ask it the question their buyers ask. Half the time it recites a competitor's strengths back to us. The other half it gets their own service list or location wrong. Neither founder knew until we looked together." — Gabriel Espinheira, founder, SharpHaw
The leak that never shows up on a dashboard
You can't A/B test a conversation you were never part of. That is what makes this the quietest leak in your marketing. When a buyer asks an assistant for options and yours isn't among them, there is no bounce to analyse, no form abandonment to review, no line in Search Console. The enquiry never arrives, and nothing on any report tells you it was ever in play.
So the founder who senses something is off does the thing that produces numbers: buys more ads. Ads report back. This reports silence. The only signal you tend to get is anecdotal: a peer mentioning "I asked ChatGPT and it suggested someone else," and by then the shortlist closed weeks ago. Measuring the loss is close to impossible, which is precisely why it goes unfixed while easier, louder problems get the attention.
How to get recommended by AI: start by checking whether you already are
The first move to get recommended by AI costs nothing and takes ten minutes: ask the assistants what your buyers ask. Open ChatGPT, Gemini and Perplexity, type the question a good-fit customer would use ("who should I hire to…", "best [what you do] for [your kind of business] in [your market]"), and read what comes back. Note who gets named, and what the tools say, or get wrong, about you.
Then fix the plumbing, roughly in this order:
- Make your core facts identical everywhere. Name, what you do, who it's for, where you operate, how to reach you — the same on your site, your directory listings, and every profile. Conflicting facts are the most common reason a model skips a business it can't pin down.
- Put your expertise in plain, structured language on your own site. Answer the buyer's real questions directly, in text a machine can extract. Checkable claims beat marketing adjectives, which a model has no reason to trust.
- Earn corroboration you don't control. Third-party mentions, an active and genuine review profile, being cited by others. These are the signals a model weighs more heavily than your own copy.
- Keep it current. Assistants favour recently confirmed information. A site untouched for two years reads as stale to a model, the same way it does to a buyer.
None of this is a growth hack, and that's the point. You can't buy your way in — there is no ad slot for an organic recommendation, which is the reverse of how honest human pricing works, where the number is already on the Plans page. You earn the recommendation the way you earn a sceptical buyer's trust: by being consistently, checkably who you say you are.
Why this rewards the boring and punishes the clever
This is plumbing, not magic. The businesses AI recommends are rarely the ones running the cleverest campaign. They're the ones whose facts have been consistent and corroborated long enough that a model can trust them without checking twice. That work is unglamorous, and it compounds. Every consistent mention and every honest review adds to a record that gets easier to verify over time, until a model can name you without hesitating.
Which is why this favours operators who keep shipping. A content record that grows every week, and stays consistent, becomes the thing an assistant reaches for. It's the same reason we run our own content in public and keep the work visible inside one shared workspace. Visibility isn't the slogan here; a consistent, checkable track record is the asset that gets you named. The clever tactic fades in a quarter. The boring record keeps paying out. Digital work that compounds.
Frequently asked questions
How do I check whether AI recommends my business?
Ask the assistants directly. Open ChatGPT, Gemini and Perplexity and type the question a good-fit buyer would use to find someone like you. Note whether you're named, which competitors are, and whether the facts about you are correct. Repeat it monthly, because the answers shift as the web underneath them changes.
Does ranking on Google mean ChatGPT will recommend me?
No. They are separate systems. SOCi's 2026 data found fewer than half of the businesses leading Google's local results appear in AI recommendations. Google ranks individual pages; assistants assemble a reply from consistent, corroborated facts about your business across many sources.
Can I pay to get my business recommended by AI?
Not directly. There is no ad slot that buys you a place in an organic recommendation. Sponsored AI placements are starting to appear separately, but the recommendation itself is earned through verifiable, consistent information and third-party corroboration, not payment.
How long does it take to get recommended by AI?
Longer than a campaign, shorter than old-school SEO. Because it depends on consistent facts and corroboration accumulating across the web, it builds over weeks and months rather than switching on. The businesses named today were being consistent well before anyone asked about them.
What to do next
The shortlist your buyers build now forms in a conversation you will never watch, and you get onto it by being verifiably, consistently who you say you are — everywhere, over time. That's slower than an ad and quieter than a ranking. It's also far harder for a competitor to copy.
Do the ten-minute check this week: ask the three assistants what your buyers ask, and read what they say about you. If the answer stings, that's useful information. When you want a senior partner to build and maintain that checkable track record week after week, book a 30-minute call — bring the exact question your buyers ask, and we'll run it live.
