Flagship report

The State of AEO

In progress — no numbers published yet

The State of AEO will report how machine-readable the average website really is, measured from real Audaeo audits. It aggregates anonymized CLEAR-score distributions across many sites to answer one question: when an AI engine reads the typical B2B site, how much can it actually use? The report is being built. This page describes what it will contain and how it is measured, and it does not publish a single number until the dataset is real.

[PLACEHOLDER — the State of AEO report needs a real dataset: run N audits with N set to a defensible threshold, aggregate anonymized CLEAR distributions, and publish the methodology before any number ships. Do not invent statistics. Every figure below is a placeholder until it comes from an instrumented product run.]

What will the report measure?

The report measures the machine-readability of real sites: how well their content survives the trip from HTML into an AI engine's answer. Each headline number comes straight from the CLEAR scores Audaeo already produces, aggregated across the sample and stripped of anything that identifies a site. These are the figures it will lead with.

Average machine-readability score of a B2B site
Credibility
[PLACEHOLDER]
Share of sites shipping no schema.org markup
Credibility
[PLACEHOLDER]
Share of sites blocking one or more AI crawlers
Alignment
[PLACEHOLDER]
Average query fan-out coverage (sub-queries answered)
Relevance
[PLACEHOLDER]
Average gap between the human score and the AI score
All five
[PLACEHOLDER]

How is it measured?

The method is the product, run at scale and anonymized. We publish the methodology in full before any number, because a report that cannot show its work is not evidence. Here is the approach the first edition will follow.

  1. 1
    Draw a defensible sample

    Audit a set of real sites large enough to report on, with the segment and sample size stated up front. [PLACEHOLDER — set N and the segment definition with the data owner.]

  2. 2
    Score every site with CLEAR

    Run the same CLEAR audit on each site, scoring every page 0 to 100 across the five dimensions from both the human and the AI-agent view. The method is identical to a customer audit, which is what makes the aggregate credible.

  3. 3
    Anonymize and aggregate

    Strip anything that identifies a site, then aggregate the distributions: averages, medians, and the share of sites failing each machine-readability check. No single site is named or reconstructable from the report.

  4. 4
    Publish the methodology, then the numbers

    Ship the sample definition, the scoring method, and the date range first. Only then do the figures go live, each one traceable to the run behind it, and refreshed on a stated cadence.

When does it publish, and how often?

The plan is to publish once the sample is large enough to defend, then refresh on a regular cadence so the report tracks the category over time, a State of AEO 2026, then 2027. The date and cadence are set with the data owner, not guessed.

[PLACEHOLDER — set the publish date and refresh cadence (proposed: quarterly) once the dataset threshold is met and instrumented.]

Why will these numbers be trustworthy?

Because they come from the product, not from a survey or a guess. Every figure is a real CLEAR score from a real audit, measured the same way for every site, with the methodology in the open. That is harder to fake than a headline stat, and it is the reason the report is worth waiting for rather than filling with numbers we cannot stand behind.

Want your site in the picture?

Run a free audit and get your own CLEAR scores now, long before the aggregate publishes.