Chapter 3

The visibility trap

Every era of marketing has had its visibility number. The channels changed. The habit didn't.

The first row needs a distinction. Recognising a name from a list, prompted awareness, is the stand-in. Coming to mind unprompted in a buying situation is the real thing. Professor Jenni Romaniuk and her Ehrenberg-Bass colleagues call that second kind mental availability (Ehrenberg-Bass).

AI visibility is the newest row, not a new idea. A brand can be named in an answer and still not be the one the answer recommends. None of these numbers was foolish. Each was the best view available of a choice nobody could watch.

A stand-in becomes dangerous the moment someone is paid to raise it. Economists know this as Goodhart's law: a measure that becomes a target stops being a good measure. Marketing built an industry on top of it.

How channels became departments

I've worked in most corners of marketing. Digital marketing. Digital experience. Sales process design. Customer experience. A full-service creative agency.

Each had its own scoreboard, and each sensibly chose numbers it could move. There's no point being held to account for something you can't control.

On websites, that meant engagement, time on site and enquiries. What happened after the enquiry wasn't ours. There were times, I'm not proud to say, when we pointed straight at the sales team. Enquiries were up. We'd done our job.

Every other discipline worked the same way. Each reported its own slice fairly, and even with all the slices on the table you couldn't see the whole picture.

The conversation that really mattered was one nobody wanted to have. Should the client be paying for our specialty at all, right now? Not how to improve it. Whether to do it. I can't remember anyone asking it in a meeting I was in.

Most brands did a bit of everything, and quite possibly nothing wonderfully. Nobody was asking whether any of it got the client chosen. And nobody would open that kettle of fish, in case theirs turned out to be the service the client didn't need.

Each new channel arrived with its own number, then a specialist, a team, a budget and an agency, all judged on that number.

Now hand every team in that table a machine that makes more of whatever it's measured on. That's what AI has done: multiplied the output and left the aim untouched.

The number was real. The cause was not.

eBay ran one of the clearest published tests of this. Its reports showed paid search ads driving sales, because clicks on ads were followed by purchases. Its economists switched the ads off across 68 US regions for 60 days and compared sales with the regions left on (Berkeley Haas). Sales showed no measurable difference. Separate tests of ads on eBay's own name found no measurable short-term benefit either. People typing "eBay" were already on their way there.

Ads on other search terms did win some new and infrequent shoppers, but most of the money went on frequent customers who would have bought anyway, so the average return was negative (Blake, Nosko and Tadelis).

The lesson isn't that paid search fails; for an unknown brand it may be the best money spent. eBay was paying to be seen by people who had already chosen it, and its reports couldn't tell the difference.

That study is a decade old. The trap isn't. An AI visibility score counts how often you're named, which isn't how often you're selected.

Why the trap pulls money the wrong way

Marketing's real work is done before the buyer shows up, where it's hardest to count. The activity nearest the sale is the easiest to measure, and budgets follow what can be measured. So money drifts toward the end of the journey, where the choice has mostly formed, and away from the beginning, where it's won. The better the dashboard, the stronger the drift.

What it looks like from the inside

The monthly review where every slide is green and revenue is flat. The standing argument between marketing and sales about lead quality. The agency report that proves its own channel worked. Each is the same problem: the organisation is shaped like its channels, and the buyer's decision isn't. Every team owns a number, and nobody owns the choice.

The people who fund marketing have noticed. In a 2024 Gartner survey, only 52% of senior marketing leaders said they could prove marketing's value and get credit for it (Gartner). Chief financial officers and chief executives were the most sceptical.

Visibility can't answer them, because a number that counts who saw you will never show who picked you. It's an early sign of Selectability, never the result.

AI is sorting the selected from the seen

AI is now making the difference between seen and selected visible. Two travel companies show it, with a caution: Tripadvisor is an intermediary people pass through, and Booking is a destination where they book. This shows what AI does to each kind, not a fair test of one strategy against another.

Tripadvisor built its hotel business on being found: travellers searched, landed on a review page, then clicked through to book elsewhere. In February 2026 its chief executive named the rise of AI overviews as one reason for falling visits (Skift).

By the second quarter, its hotel revenue was down 23% on a year earlier, though the company blamed a fluctuating economy (Tripadvisor), and its troubles with search are older than AI. So AI isn't the whole cause.

Booking Holdings is the name travellers go to on purpose. About two-thirds of its consumer bookings come direct. Its chief executive said its brands were "showing up an awful lot" in AI assistants. Yet its finance chief put bookings from them at "less than 1% of total room nights" (PhocusWire).

Both companies were seen. One was selected.

AI didn't take Booking's sale. It took Tripadvisor's place as the page people passed through. In the few cases so far, AI is sorting the brands buyers pick on purpose from the ones they only pass through.

The way out is to measure Selectability itself. That starts with a question marketing has rarely asked: what is a choice actually made of?

Whose evidence this is. Gartner sells research and advice to marketing leaders. Skift and PhocusWire are travel trade press, and Tripadvisor's figures come from its own filing. Blake, Nosko and Tadelis used eBay's own data, in an NBER working paper.