Stop counting clicks. Start measuring whether you're the answer.

Somewhere today, your best prospect asked an AI what to buy — and got a shortlist. You weren't in the room. You may not even know whether you were on the list.

That conversation — invisible, unmeasured — is where more and more purchase decisions now form. For twenty years, marketing measured visibility the same way: clicks, rankings, sessions, traffic. Those metrics worked because discovery happened on a page you could see and count. That page is disappearing. Buyers increasingly ask ChatGPT, Perplexity, Gemini and Google's AI answers, then act on a synthesised recommendation without ever clicking through. The decision forms inside the answer — before the click your analytics was built to count.

This shift has a name, or two. Answer Engine Optimization (AEO) — being cited in AI answers. Generative Engine Optimization (GEO) — the broader discipline of being the brand these systems retrieve, trust and recommend. Together they're where visibility is now won and lost, and they've quietly broken the metrics that ran marketing for two decades. So the real question isn't "how do we do AEO?" It's sharper: what do we measure now — and what is the one number that matters?

Why the old metrics went blind

Three shifts, one consequence.

• Zero-click. The answer is the destination now, not a step toward your site. Traffic undercounts you, because you can shape a decision without ever earning a visit.
• No rankings. There is no page one — just a synthesised paragraph that names a few brands and omits the rest. You're in the answer or you're invisible. "Position seven" no longer exists.
• Fragmented attribution. Pipeline gets influenced before the prospect touches a trackable channel. By the time they arrive, the AI has already made your case — or a competitor's.

The consequence: the dashboards that told you marketing was working are now blind to the exact place buying decisions form.

The north star has moved

For two decades the implicit north star was traffic — visibility you could count in clicks. In the AI era that's the wrong star, because the most valuable visibility now produces no click at all. The new north star is simpler to state and harder to earn: are you the answer?

Concretely: of all the moments your buyers ask an AI a question where you could be the recommendation, how often are you — named, cited, and accurately represented? Practitioners are converging on a term for the competitive version of this: Share of Model — your share of the AI's recommendations against your rivals. That is the number to organise around. Everything else is a supporting metric that explains it.

The old question was "how much traffic did we get?" The new one is "when our buyer asked, were we the answer?"

The trap: presence is not preference

Here is where most brands will go wrong — the same mistake they made with social. They'll count the easy thing. The easy thing is mentions: how often the AI says your name. It feels like progress and it's largely a vanity metric. Being mentioned is not being recommended. A brand can appear in every answer as the expensive option, the legacy one, the "you could also consider." Presence without preference is noise.

So the north star has to hold three things at once:

• Presence — are you in the answer at all? The baseline.
• Preference — when you're there, are you the pick or the afterthought? This is Share of Model.
• Portrayal — is what the AI says about you accurate and favourable? The one almost everyone forgets.

Portrayal is the sleeper. These systems synthesise your brand from everything they've ingested, and they will confidently describe you as something you stopped being three years ago. If the model's account of you is wrong, citation frequency won't save you — you're simply being accurately recommended as the wrong thing.

The core metrics, in order

Under that north star, five metrics actually matter — and the order is the point:

• AI citation / presence rate — how often you appear in answers to your buyers' real questions. The baseline.
• Share of Model — your share of those answers versus competitors. The competitive truth.
• Sentiment and accuracy — how the AI characterises you. The portrayal check.
• Prompt coverage — the breadth of buyer questions you show up for. The depth of your presence.
• AI-influenced pipeline — whether any of it becomes revenue. The metric your board will ask for, and the one that keeps the other four honest.

The failure mode is inverting this — opening a dashboard of forty indicators with no hierarchy. Pick the north star, then read these five as its explanation. A metric that doesn't ladder up to "are we the answer?" is a distraction.

How you actually measure it

The method is more prosaic than the mystique suggests. You don't wait for a magic dashboard — you run the test. Define the set of questions your buyers actually ask AI during their decision. Run those queries across the engines that matter — ChatGPT, Perplexity, Gemini, Google's AI answers — on a regular cadence. Read each result for three things: are you present, are you the pick, is the account of you accurate. That's your baseline and your scoreboard. Tools are emerging to automate it, but the discipline — a defined query set, a steady cadence, an honest read — is what produces signal. Not the software.

The uncomfortable part

One correction, because the hype gets it wrong: GEO does not replace SEO, and treating it as a fresh technical trick will fail. Google's AI answers still draw on its search index — if you don't rank, you won't be cited. And the deeper truth is that being the answer is mostly a positioning and authority problem, not a markup one. AI systems recommend the brand they understand most clearly and trust most — the one with a sharp, consistent identity and genuine standing in its category. You cannot schema-hack your way to being the recommendation. You earn it the way you earn everything in go-to-market: by being unmistakably clear about who you're for, and credible enough that even a machine repeats it.

The north star for the AI era isn't a new kind of traffic. It's a new kind of question — asked on your buyer's behalf, thousands of times a day, by a machine you don't control: when they ask, are you the answer? Most brands can't answer that today; they're still counting clicks on a page fewer people visit. The ones that will win are already measuring the thing that replaced it.