AI visibility isn't a retail problem — it's a readability problem

The first wave of writing about AI shopping was all about catalogues — barcodes, feeds, product markup. That was a real problem, but treating it as the whole problem gave everyone else the wrong idea: that this is something happening to retailers. It isn't. The same request goes out when somebody asks for a conveyancing solicitor, an emergency electrician or a help desk tool under $200 a month, and the same rule decides the answer — the assistant names whoever it can confidently understand.

The mechanism is identical; the evidence isn't

What an assistant is really doing

It reads pages, extracts what it can, and then has to decide whether it's confident enough to put a name in front of a person who asked. Confidence comes from things it can confirm. For a product that's a barcode and a set of attributes. For a local business it's an address, opening hours and the area you cover. For a professional firm it's registrations, named people and evidence of outcomes. Different evidence, same job.

Which is why a scorecard built for shops is worse than useless

Score a law firm against a catalogue rubric and you get zero out of a hundred and a list of instructions it can never follow. That's not a harsh grade — it's the wrong question, and it hides the three or four things that would genuinely move the firm's visibility.

What actually goes wrong outside retail

Services described only in prose

Your practice areas or trades are page headings and paragraphs. A person reads that fine. A machine gets no structured statement of what you do, so when it needs to name someone who does exactly that thing, it picks a directory instead — because the directory published it as data.

A location no machine can find

Address and hours in a footer image, coverage described as "and surrounding areas". Every "near me" question in your town goes past you, and nothing on your site looks broken.

Proof that isn't on the page

In categories decided on trust rather than specification — law, health, trades, advice — credentials, registrations, named practitioners and real outcomes are the equivalent of a barcode. Missing, they leave an assistant nothing to justify recommending you with.

Pricing behind a form

Especially in software. An assistant won't state what it can't confirm, so a tool whose plans are visible gets compared and a tool whose plans require a demo request gets left out of the comparison entirely.

How to check

Run any website through a free audit. It works out what the business actually is first — what it offers, where it serves, how someone becomes a customer, what decides the choice in that category — and grades it against that. Then it puts real buyer questions to the assistants and tells you whether you were named, somebody else was, or nobody has claimed the answer yet.