Free software. Open to inspection.

Your page,
under review.

An open-source audit for AI search. Find crawl blockers, check whether your content answers the query, and see the evidence behind each recommendation.

Runs on your machine. No account or AI API key.

backup-retention.html
Sample page / recorded audit
Target query

“database backup retention”

60/100Partial audit (HTML)

Three checks to explore

HTML only. Crawler access and live response checks were not performed.

Needs attentionconditional evidence

The title leaves out the subject.

This page explains database backup retention. Its title says only “Fieldnotes”, so that useful context is missing from the page title.

<title>

Fieldnotes

0% of the body's key terms appear in the title

Compare two saved runs.
Rule and sourcestructural-alignment.title-entity-alignmentsageo-arena-kdd-2026

Try it on a page
you know.

Start with a URL. Add a question your audience asks. Read the findings before changing the page.

Installation and first audit
npx aiseo-audit https://yoursite.com

Replace the example URL with your own. Requires Node.js 20.19+ and npm 10+.

Check a specific questionnpx aiseo-audit https://yoursite.com --query "your audience’s question"

A score should
have receipts.

You should be able to question a recommendation. Each finding names the check, its evidence, and the limits of what it can tell you.

Read the research

Show the evidence

Checks distinguish direct research support from conditional findings, practical heuristics, and unscored observations.

How evidence is labeled

Change the tool when the evidence changes

Version 2 removed points for simply having lists and tables. Counting formatting didn’t justify a better score.

Decisions and corrections

Be clear about the limits

The audit checks the page you provide. It cannot see a search engine’s private index or predict whether your page will be cited.

What the audit can’t measure

Help improve the next audit.

Maintained by Jeff Patterson at Agency Enterprise. The source, scoring rules, and research reviews are public and MIT licensed. Found a misleading result? Bring the page and the finding.