Mentions vs money: what AI visibility actually tells you

The AI visibility category answers one question well: does AI mention you? Tools in that space track how often engines name you, which prompts you appear in, and when your position moves. That is genuinely useful work, and NetBack does a free version of it as its front door.

It stops one question short of the one that decides budgets: did any of that traffic sign up, and did it pay?

Mentions are a leading indicator, not the outcome

Being on the shortlist an AI recommends is upstream of revenue, the same way an impression is upstream of a sale. It is worth measuring. It is not the thing you report.

The gap shows up the moment someone senior asks a follow-up question. Of the people who arrived from ChatGPT last quarter, how many converted? Which engine sent the ones who stayed? Is the AI channel worth another hire? A visibility score cannot answer any of those, because a mention and a customer are different objects, and nothing in a visibility tool connects them.

That is not a criticism of visibility tracking. It is a description of where it ends.

Why the AI channel hides in your analytics

Two things make AI traffic hard to see, and both are structural rather than a reporting bug.

Referrers get flattened. A visit that began in an AI answer often lands in your analytics as Direct or Referral, so the fastest-growing channel is the one you cannot name. Published measurements put AI referral share in the low fractions of a percent of visits, rising sharply month over month — which means most teams are reading a number that understates a channel that is changing shape while they read it.

The dark funnel. A buyer asks an engine which tool to use, reads the answer, then types your name into a search bar an hour later. The AI conversation caused the visit; nothing in the click carries that fact. No referrer-based method recovers it on its own — you have to ask the visitor, and keep that self-reported answer separate from what you detected.

Neither problem is exotic. Both are simply what happens when a channel arrives faster than the measurement built for the last one, and the reported conversion figures for AI traffic vary from roughly 1.3x to 23x depending on segment precisely because everyone is measuring a slightly different thing. We walked through that spread in does AI traffic actually convert.

What it takes to answer the revenue question

Three pieces, in order:

1. Detect the AI-referred visit. A lightweight first-party pixel that recognises engine referrers and records the visit, including the ones analytics buries.
2. Match the visit to the outcome. Connect Stripe, and optionally GA4, so a visit becomes a signup and a signup becomes a dollar, matched on a first-party identifier rather than modelled from aggregates.
3. Report it honestly. Deterministically detected and self-reported figures kept apart, never blended into one flattering number, and reconciled against the revenue total you already report — with the share of conversions you could actually match stated up front.

That last point matters more than it sounds. The first question anyone senior asks about an attribution number is how it ties back to the number they already trust. An AI revenue figure that cannot be reconciled does not survive a board deck, however good it looks.

The practical split

If your job this quarter is to get mentioned more, visibility tracking is the loop you want, and you should keep doing it.

If your job is to prove the channel makes money — to defend the budget, justify the hire, or decide whether AEO work is worth continuing — you need the layer underneath: the visit, the conversion, the dollar, and the honest coverage rate.

What to do with that this quarter

If you have no measurement in place at all, the order is boring and it works. Get a baseline for the visits you can already attribute, even knowing it undercounts. Add the self-reported question so the dark funnel stops being invisible, and keep that answer in its own column. Then, before you report anything upward, decide what your match rate is — the share of conversions you could actually tie to a visit — and say it out loud alongside the number. A figure with a stated match rate survives scrutiny. A figure without one gets picked apart the first time someone asks how it reconciles.

The teams that get this wrong are rarely the ones who measured nothing. They are the ones who reported a confident number they could not defend, got challenged once, and lost the channel's budget for a year.

Mentions are the leading indicator. Revenue is the point. You can see where you currently stand on the first one in about a minute, and the second one is the work that follows.

Run NetBack's free scan → — it takes about a minute and tells you where you stand today.