Academy · Updated 2026-08-23 · 6 min read
AEO vs. SEO reporting: what to show clients about AI visibility
Rankings and CTR don't exist inside an AI answer. Here is what replaces them, why your SEO stack can't report it, and what an honest client-facing AI-visibility report contains.
An SEO report answers "where do we rank and what did it earn?" An AEO report answers a different question: "when a buyer asks an AI assistant, do we exist in the answer — and if not, who does, and why?" The metrics, the data source, and the failure modes are all different, which is why bolting a chatbot column onto a rank tracker produces numbers nobody can defend to a client.
The metrics that replace rankings
Position and click-through rate are unobservable inside an answer engine. What is observable, from stored answers: mention rate (how often the brand is named at all), share of AI voice (the brand's slice of all tracked-brand mentions), average rank when mentioned (named first, or named fourth), recommendation rate (mentioned is not endorsed — the extraction marks which mentions are actual recommendations), and citation share (how often the engines cite the brand's own domain as a source).
One split does more work than any single metric: unprompted versus branded questions. A question that hands the engine your brand name will score near-perfectly and proves nothing. Unprompted questions — phrased the way a cold buyer asks — are the real scoreboard. Any report that blends the two is inflating itself; ours scores them separately by design.
Why a separate layer, not another rank-tracker column
Three structural reasons. Answers are probabilistic: the same question re-asked can produce a different brand list, so a defensible number needs repeated sampling and stored evidence, not a daily spot-check. Grounding matters: an answer built from live web search reflects the evidence pool today, while an ungrounded answer reflects training data months old — a report that doesn't label which is which is measuring the wrong thing (the grounding rule). And the diagnosis is different in kind: losing a question is explained by a content, authority, freshness, format, or schema gap — a classification, with the winning competitor's cited page as the evidence, not a position delta.
What the client-facing report contains
The AEOSearch report an agency shows a client has five parts: a scorecard (the metrics above, per engine, with the unprompted/branded split explicit); the question set with one stored answer shown verbatim per question; a source map — every domain the engines cited, ranked, which is where "why competitors win" stops being a theory; a prioritized fix plan tied verbatim to lost questions; and the methodology, on the page, so the client can check the sampling instead of trusting a slide.
For a brand starting from zero, the report adds a readiness checklist run against the brand's own domain — AI-crawler access, schema.org, llms.txt, citation presence — because when unprompted visibility is 0%, prerequisites explain more than content does. The number to watch from there is time to first unprompted mention.
We publish our own numbers because we ask you to trust ours. On 2026-08-23 we ran the full AEOSearch audit on AEOSearch itself — 300 stored, grounded answers across ChatGPT, Claude, Gemini, Grok and Perplexity. Our unprompted mention rate was 0%. Our citation share was 0%. That is the honest starting line for a young brand, and every method on this page is what we are running on ourselves, in public, to move it.
Questions, answered straight
What metrics belong in an AI-visibility report?
Mention rate, share of AI voice against tracked competitors, average rank when mentioned, recommendation rate, and citation share — each computed from stored, grounded answers with repeated sampling, and each split between unprompted questions and questions that name a brand. Blending that split inflates every number in the report.
Do I need a separate platform for AEO reporting?
You need a separate measurement layer: answer engines are probabilistic and grounded, so defensible numbers require repeated sampling, stored answers, and grounding labels — things a rank tracker was never built to do. Whether that layer is a separate platform or a module in your stack matters less than whether the evidence is stored and auditable.
How often should AI visibility be reported?
Monthly is the honest cadence for the headline numbers: answer engines are volatile enough that daily deltas are mostly sampling noise, and content or authority fixes take weeks to enter the evidence pool. Continuous monitoring is for regressions — a dropped citation or a blocked crawler — not for re-litigating the scoreboard every morning.
Can a brand rank well in Google and still be invisible in AI answers?
Yes, and it is common: ranking is about the index, while being named in an answer is about the model's evidence pool and its learned associations. A page can hold position two and never be cited by an engine. That gap — visible in search, absent in answers — is exactly what AEO reporting exists to expose.
— The AEOSearch team