Published 4 September 2026 · Methodology
A transparent account of what AEO v2 measures, how evidence is scored and where the automated audit deliberately stops.
The aim is simple. A useful score should show what the tool saw. It should also show what the tool could not see. A crawl can read a page. It cannot prove that another source trusts the site. It also cannot prove that an AI tool will cite it. Those checks need their own test. This page sets out the rules so anyone can judge the result.
Our methodology follows six steps. They are discovery, understanding, extraction, evidence, citation-worthiness and recommendation readiness. Each measured layer receives a score out of 100. The audit reads public pages and crawl files. It awards no points for claims it cannot check.
The quick assessment reads 1 homepage and up to 5 content pages from the XML sitemap. It favours guides, research, case studies, FAQs and explanatory pages. Account, checkout, administration and policy routes are excluded. Our data for the automated score consists only of the text and technical signals retrieved in that run.
We measured evidence through named project details, dates, prices, quantities and documented methods. A live client link or official record can help a reader check the context. An outbound link is not treated as an endorsement. Before and after comparisons are only called results when both states were measured.
Our benchmark uses a 100-point scale for every measured layer. Unknown evidence never becomes a pass. External corroboration and live citation testing keep a blank score marked “Not tested”. Technical SEO also stays separate from content opportunities. We correct parser errors, but we do not hide a missing title, description, canonical URL or crawl failure.
No. A high on-site score means a site provides useful material for discovery, understanding and extraction. It does not prove independent authority, ranking or citation. A live test needs representative prompts in current AI and search products. The test must record appearances, competitors and cited sources at that time.
On 4 September 2026, we tested the revised engine against the KHDS public-site fixture. We also used regression cases for metadata order, structured data, decorative images, genuine findings and unverified layers. We measured 100/100 for verified technical SEO and 90/100 for on-site AEO before this page was published. This is an internal verification record. It is not an independent endorsement or a promise of search performance.
See the method in practice
The free snapshot shows the measured layers and labels anything the crawl could not verify.