Measuring AEO without making things up

AEO measurement is the stage where most enterprise programmes quietly die. The dashboard looks busy, monthly numbers swing 10–15% in either direction for reasons nobody can explain, the CMO can't tell a CFO what's actually moving, and by the third quarterly review the budget conversation has already turned — finance has decided the line item isn't defensible, and the work goes with it, regardless of how good the work was.

Most enterprises don't have one. Most consultancies don't sell one. Most teams running AEO programmes are producing data and calling it measurement.

The Proof Layer is what stands between that quiet death and a programme that survives. It's the framework, the cadence, and the attribution model that turn five inherently noisy metrics into a story finance can sign off on at a budget meeting — and discover the difference at exactly the worst possible moment if you don't have one in place.

Why this stage exists at all

AEO measurement is harder than SEO measurement for a structural reason. There's no SERP. There's no shared, public leaderboard every team can point to and argue about. AI responses vary across users, sessions, regions, and platforms — the same prompt asked twice in a row can return two different answers. SISTRIX's 2026 citation drift study found 56–74% of sources cited by AI rotate week-over-week. The volatility isn't a bug in the measurement; it's a property of the medium. A measurement framework that doesn't account for it produces noise so loud it looks like failure every other month.

The default response — the response we see in almost every measurement review we run — is to buy a tool, log in to its dashboard, and call that measurement. It isn't. A tool dashboard is just data. Measurement is the framework around the data: which prompts are being tracked and why, what “good” looks like for this specific business, what cadence absorbs the noise without losing the signal, which tools are being triangulated across, what the rolling averages actually say once the drift is stripped out, and — frankly the most important part — how the report reads to three different audiences at once: a CMO who wants the trend, a CFO who wants the line, and a board who wants the story.

The Proof Layer is also where AEO becomes defensible at the line-item level. The work in Foundation, Authority, and Distribution is the substance — but Measurement is what proves the substance is moving. Without it, AEO sits next to “brand campaigns” in the marketing budget as another initiative the team can't quite justify when pressed. With it, AEO sits next to “infrastructure” — and infrastructure is a different conversation entirely.

Brand campaigns or infrastructure. The Proof Layer decides which conversation finance has about you.

The playbook

Five moves, in order. The first four set up the system. The fifth is the work that decides whether the engagement renews at the twelve-month mark.

The framework is the document everything else runs against. It defines what's being measured, why, how often, against what targets, and what the report looks like when it lands. It gets built once at the start of the engagement, signed off in writing by the client's leadership, and refreshed annually. The five-metric set has now standardised across the industry — HubSpot's vocabulary, since the AEO tool launched in April 2026, is becoming the de facto default. The five metrics:

Setting targets is the work most teams skip — and the work that protects the engagement when the inevitable bad month arrives. Targets get written down explicitly, with three horizons: 90-day (modest, mostly noise), 6-month (where the trend becomes interpretable), 12-month (where the case for next year's budget actually gets made). Sample target language: “Move citation rate from 12% to 25% across priority-1 prompts over six months. Move share of voice from 18% to 30% relative to top three competitors over nine months.” Without targets, every report is a Rorschach test. With them, the report has a structure the board reads in two minutes.

Every AI visibility tool has blind spots and incentives. Some mark too many mentions as positive. Some miss specific platforms entirely. Some have AI modules bolted onto legacy SEO products that are still maturing. The honest framing is that no single tool is reliable enough to make decisions from — which is why the rule is always two, sometimes three, in parallel.

The minimum stack:

The discipline that matters: when two tools disagree, that's signal, not failure. The reconciliation work — figuring out why Tool A shows the brand cited in 32% of priority prompts while Tool B shows 24% — is where the actual judgement lives. The consultancy that gives the client a single number from a single dashboard is, frankly, selling the appearance of measurement rather than measurement itself. One operational warning: don't over-tool. Three is the ceiling; even at three, data reconciliation eats hours that should be going into interpretation. A single thoughtful tool plus the Bing dashboard plus GA4 is a stronger framework than five tools running in parallel that nobody has time to read.

Citation drift is the central methodological challenge of the entire Proof Layer. It is not a problem to solve. It is a property of the medium to absorb.

The SISTRIX 2026 study established the baseline numbers: 56–74% of sources cited by AI rotate week-over-week, depending on platform. Google AI Overviews is the most stable — over half of tracked prompts show zero source change ever, which is its own kind of weirdly informative finding. Google AI Mode shows a core-plus-carousel pattern, where a small set of citations stay locked in and the surrounding references rotate. ChatGPT is the most volatile, with near-total weekly fluctuation in many categories.

Three platform architectures, three measurement behaviours, one principle: read trends, not snapshots. The operational implications, in order:

The team that absorbs drift without flinching renews the budget at year-end. The team that doesn't, doesn't.

Monthly absorbs the noise. Quarterly tells the story. Annual makes the case. Each has a different job; running them as the same job is one of the most common failure modes in the entire stage.

Three tiers, three jobs:

Three cadences, three different muscles. The team that runs all three keeps the engagement. The team that flattens them into a single monthly grind — because monthly is what the client asks for — loses the strategic case for the work at the budget review.

The hardest measurement question every client eventually asks: does this work drive business outcomes? Most consultancies don't have a good answer. The honest answer is that AEO attribution is structurally partial — and naming the limits up front is what makes the rest of the framework credible. Three things are true at once: AI referral traffic is genuinely small in absolute terms; the gap between AI influence and AI attribution is wide (someone reads a ChatGPT recommendation, opens a new tab, types the brand into Google, lands via “organic” — and analytics credits the wrong channel); and B2B buying cycles compound the gap further, with AI surfaces invisible by the time the deal closes months later. The Proof Layer doesn't solve this. It names it, and builds the partial-attribution model that holds up anyway. Three components:

The under-claim is what makes the rest of the number credible. The consultancy that claims definitive AI attribution is selling a model that doesn't exist.

What “good” looks like

The framework holds up at a board meeting if five things are in place. If they are, the CFO can read the report in three minutes. If they're not, the work survives until the next budget review and then doesn't.

Common failure modes

Five patterns, present in almost every measurement review we run, all the difference between a programme finance renews and one finance quietly cuts.