How people find businesses now
When someone asks AI for what you do, does it say your name?
Assistants like ChatGPT and Gemini answer with a shortlist of a few businesses — not ten pages of links. Only the ones they can read make that list. Shop, law firm, SaaS, plumber, studio: paste any address and find out. Free check, 60 seconds, no account.
Why now
This is already how your customers look.
Not a forecast. Every month more of them open an assistant instead of a search box, describe what they need — a product, a contractor, a tool, an advisor — and get back a handful of names. That handful is the entire shortlist, and it's decided before anyone reaches your site.
In one year, visits to U.S. retail sites from AI assistants grew more than twelvefold — and they were still climbing at the end of 2025. Retail is simply where it was measured first.
Half of U.S. shoppers used AI to help them buy something last year. About two in three say they will this year. The same habit is what now picks contractors, agencies and software.
Roughly that many open ChatGPT in a given week — over 800 million people. That's the size of the room where these recommendations get made.
Sources: Adobe Analytics (2025) · PartnerCentric consumer survey (Dec 2025, n=1,004) · OpenAI (2025).
You're in the answer, or you're nowhere
An assistant names a handful of businesses — often just one. If yours isn't among them, the person never scrolls down to find you, because there's nothing to scroll. No second page, no listings below, no second chance at that customer.
The businesses a machine can read get named again and again
Assistants lean on the sources they can parse and confirm. The big platforms, marketplaces and directories were wired in from the start; everyone else was left out by default — not for being worse, but for a site a machine can't make sense of.
Every month the gap gets harder to close
More people ask an assistant first, and assistants keep returning to the names they already trust. Visibility you don't claim now is visibility a competitor gets by default — and you never see the customer who didn't arrive.
And it's high-intent traffic: people who arrive from an AI assistant browse 12% more pages and bounce 23% less than other visitors. — Adobe Analytics, 2025
Any website
There's no dropdown for what kind of business you are.
We used to only read shops, and everything that wasn't one scored zero and got handed a list of demands it could never meet — feeds, barcodes, product markup. That was a category error, and we fixed it. There is now one pipeline, and it starts by working out what you actually are.
Products, services, software, content or advice
Most businesses are a mix, so we don't force a choice. A hobby shop that also runs classes is both, and gets graded as both — no special case, no wrong template.
One street, one region, or everywhere
A local business is measured on the questions people actually ask locally — including "near me" — and on whether a machine can find your premises and hours. A global one never is.
On spec and price, or on trust and proof
Nobody picks a solicitor the way they pick a cable. Where the category turns on credentials, reviews and named expertise, that's what we grade — not barcodes you'll never have.
And when a question doesn't apply to you, it comes back not applicable — never as a failing mark. A software company has no premises; that's a fact about the business, not a gap in it. Where we genuinely couldn't tell, we say that too, instead of guessing and grading you on the guess.
What we measure
Two questions, answered with real output.
Can the assistants read your business at all — and when someone asks for what you do, do they name you, or somebody else? We answer both, and show you the actual answers we got.
A plain-English score of how well the assistants can read you, and exactly what's in the way.
For real questions your customers ask about what you do: are you named, is a competitor named, or is it still open?
When you're not named, we tell you who was — so you know who you're actually up against.
The 360 view
We read the whole business, then follow it all the way to the sale.
A score on its own is a number. What makes it useful is knowing what you are, what your buyers ask, where the answer breaks, and what happened after somebody clicked.
Understand
Before we grade anything, we read your site and answer the same fixed set of questions about every business: what you offer and how it's organised, where you serve, how someone becomes a customer, who your buyers are and what they worry about, what proof you carry, what an assistant can actually retrieve, and who else is in your category's answers.
Ask
Those answers turn into the questions your real buyers ask — in their words, not yours, and localised only when we know where you serve. We put them to each assistant and record who got named, in which position, and what the model said about them.
Follow through
Being named is only the first stage. Connect your analytics and we join it to the rest: named → clicked → engaged → and, where you have it, converted. That's the difference between "we're not visible" and "we're visible, and the page is losing them."
Not named. The work is in what an assistant can read about you.
Named, but nobody clicks. How you're described is the problem, not whether you're there.
Named, clicked, then lost. The recommendation is landing; the page isn't closing.
Working end to end. Do more of it, in the clusters next door.
One recommendation at a time, and always with the working shown. If we can't show you where a number came from, we don't show you the number.
How it works
Check → Fix → Keep watch
Not another dashboard to check. Get in front of the assistants, and stay there as they change.
Audit
Paste your URL. We read your site the way an assistant does and tell you what's keeping you out of its answers — pages it's told to stay away from, services described only in prose, products with no barcode, a location no machine can find, listings too thin to match on.
Fix
We fill in what's missing and publish it where machines read it: the identifiers and details assistants match on, your offering marked up as the thing it actually is, titles and descriptions in the words your buyers use. We never invent a fact about your business.
Keep watch
We re-check on a schedule: where you're named, who took the answers you lost, and which questions opened up — because every assistant keeps changing how it reads sites.
The flywheel
Every check makes the next recommendation better.
Anyone can ask ChatGPT a question. What can't be bought is the record of what happened next — which changes actually moved a business like yours, on which assistant, at which stage.
A record nobody can backfill
Every question, every surface, every run: who was named, in what position, in what words. You can buy the breadth of that on any given day. You cannot buy last year's.
What your category is actually asked
Connect your own search and analytics data and the picture of what buyers ask sharpens — not just for you, for everyone in your category. Each connection makes the map better.
What worked, and what didn't
We're in the seat where the change happens, so every fix is timestamped against the numbers it was meant to move. Over time that becomes an answer to the only question you care about: if I do this, will it work?
Three tests before we call something a win
Enough runs before and after to mean anything. A check that the rest of your untouched pages didn't drift the same way. And a comparison against similar businesses that made no change over the same weeks. Results come back as worked, no measurable effect, or not enough evidence — and a "didn't work" is a finding we'll tell you, because it saves you the next month.
No benchmark built on three businesses
Category comparisons and trend calls stay switched off until a category has enough businesses in it to be honest, and no single figure is ever published that could expose one competitor's numbers. When something is suppressed, the report says so rather than quietly showing you noise.
We never claim to know how many people ask an assistant a given question — nobody publishes that, and anyone quoting you a monthly volume is guessing. What we can tell you is what's rising, how far ahead of the pack you'd be, and whether you're named yet.
The assistants
Each one reads your site differently — and keeps changing.
We check each of them on its own terms, and keep checking as they change.
OpenAI
Answers from a live read of the web. It favours sites it can parse and claims it can confirm. How it reads →
Answers from search, feeds and the business listings Google already holds. How well you're categorised and described decides who gets named. How it reads →
Microsoft
Doesn't yet answer shopping questions we can measure directly, so we check you're ready for when it does. How it reads →
What waiting costs
Nothing breaks. That's the problem.
There's no error message when an assistant skips you. No bounced order, no missed call, no angry email — just someone who asked, got three other names, and went with one of them. You find out by not finding out.
Someone else becomes the answer
Assistants keep going back to the names they've already read and trusted. Every week you're unreadable is a week a competitor is being learned as the answer for what you do — and being the established answer is a lot easier than becoming it.
Fixes take effect on their schedule, not yours
Publishing better data doesn't flip a switch. The assistants have to come back and read your pages again before anything changes in their answers. The clock starts when you fix it, not when you decide to.
A fix that works today can quietly stop working
Each assistant keeps changing what it reads and how it weighs it. Businesses that fixed things once and stopped looking slide back without noticing — which is the whole reason we keep checking instead of handing you a one-off report.
Pricing
Open pricing. The audit is free.
Run the check with no account. Subscribe when you want the fixes made for you and the re-checks running on their own.
- Full check + the fixes done for you
- Monthly check across every assistant
- 1 business
- See what changed after each fix
- Everything in Starter
- Checks every two weeks
- Connect your analytics — the full funnel
- Track who keeps winning your answers
- Everything in Growth
- Weekly checks
- Multiple sites or locations
- What-worked playbook for your category · developer API
See whether the assistants can find you.
No install, no login, nothing to set up. Paste your address and read the report.
Run your free audit See an example