How AEOCheck works: audit scope, scoring and limitations
A score is useful only when people can understand what produced it. This page documents what the current scanner observes, how it prioritizes findings and what it cannot tell you.
What a scan does
- Starts from the public URL you submit and resolves its destination.
- Discovers eligible same-origin pages using crawlable links and sitemap information.
- Inspects up to 40 pages in one scan, with bounded request sizes, redirects and timeouts.
- Uses up to six bounded page workers and caps the saved findings set.
- Checks site-level robots and sitemap information, then evaluates page HTML and cross-page patterns.
- Shows evidence, affected URLs, severity and suggested fixes in the report.
What the 46 checks cover
Bot & AI access 11 checks
Can search and AI bots reach your pages?
Titles & page basics 10 checks
Does every page say what it is about?
Content clarity 7 checks
Is your text easy to read and quote?
Structured data 3 checks
Can machines tell what your page is?
Links & redirects 8 checks
Do your links lead somewhere?
Mobile, speed & safety 5 checks
Is the page fast, safe and phone-ready?
Share previews 2 checks
Do shared links look right?
For every rule name and its plain-English explanation, visit the complete audit check catalogue.
How the diagnostic score is calculated
The scanner starts with 100 points after at least one page has been analyzed. It groups findings by rule and category so identical warnings are not blindly subtracted once for every occurrence. Each issue receives a severity penalty, adjusted by a category weight and by how broadly it affects the scanned site. The final value is clamped between 0 and 100 and rounded to one decimal place.
| Severity | Base penalty | How to read it |
|---|---|---|
| Critical | 25 | Major access, safety or availability problem |
| High | 12 | Important issue that can affect multiple pages or key functions |
| Medium | 5 | Meaningful technical or content-quality issue |
| Low | 1.5 | Lower-impact improvement opportunity |
| Info | 0 | Observation that does not reduce the score |
Category multipliers and repeated-page coverage affect the total penalty, so two reports with the same number of findings can receive different scores. A site-wide problem is treated differently from one isolated page issue. The formula is designed to prioritize repeated, higher-severity problems; it is not calibrated against Google's ranking data.
What AEOCheck does not measure
- Google rankings or keyword positions: no connection to Google Search Console is made by the scanner.
- Live AI mentions or citations: it does not submit prompts to ChatGPT, Claude, Gemini, Perplexity or Google AI Mode.
- Backlinks or domain authority: it does not run a backlink index or competitor-link comparison.
- Real-user Core Web Vitals: server-observed HTML request timing is not LCP, INP or CLS field data.
- JavaScript rendering: the crawler reads delivered HTML and does not run the site in a real browser.
- Guaranteed indexing: sitemap discovery and crawl access do not guarantee that a search engine will index or rank a URL.
How to use the report responsibly
Fix access and indexability problems first, then review high-severity findings and verify every suggested change on the live page. Use Search Console for actual Google queries, impressions, clicks and indexing information. To assess AI visibility, build a fixed set of relevant questions and periodically record the answers and cited URLs across the engines you care about.
For general best practices, see Google's generative AI search guidance. Google states that its normal Search fundamentals apply to AI features and that eligibility does not guarantee crawling, indexing or display.
Run the current audit
Inspect up to 40 public pages and review findings with the documented limitations in mind.
Run AEOCheck