site-audit
Skill1 starsChanged 3 days agoMITView on GitHub
Audit a whole website for AI search: whether AI apps and search engines can read and index each main page, whether the site states the facts that buyers ask about, and whether each page can be quoted. Returns an ordered list of exact changes. Use for "audit my site for AI search", an AEO or AI SEO site audit, or a site check before content work. Needs no Frontmost AI account.
10 other projects use a version of this.
Adopted by
Most recently taken up by
- Shash-Builds/Shash.buildstheir own version, first seen Nov 16, 2026
- zaynelovecraft/agent-ostheir own version, first seen Nov 8, 2026
- advisorppc-org/advisorseo-claude-plugintheir own version, first seen Nov 1, 2026
- denchik2245/Skillstheir own version, first seen Nov 1, 2026
- jbobbink/seo-gsc-wizardtheir own version, first seen Oct 23, 2026
- khadinakbarlabs/vibeseo-plugintheir own version, first seen Oct 13, 2026
Other versions10 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 10
- mattbutlerengineering/mattbutlerengineering★ 3same name
- DarynaRodriguez/seo-skills★ 1same name
- denchik2245/Skills★ 1same name
- Shash-Builds/Shash.builds★ 1same name
- zaynelovecraft/agent-os★ 1same name
- advisorppc-org/advisorseo-claude-plugin★ 0same name
- ai798-Lab/UPUP★ 0same name
- jbobbink/seo-gsc-wizard★ 0same name
- khadinakbarlabs/vibeseo-plugin★ 0same name
- statusdone/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository18 other files it gives its agents.
- ai-readinessSkill
- ai-search-strategySkill
- answer-accuracySkill
- answer-pagesSkill
- community-repliesSkill
- customer-surveySkill
- directory-listingsSkill
- listicle-outreachSkill
- mention-checkSkill
- name-checkSkill
- page-draftingSkill
- placement-trackerSkill
- question-draftingSkill
- question-gapsSkill
- rival-gapsSkill
- scan-reviewSkill
- source-mapSkill
- tactic-checkSkill
The file
We split this skill into the rules it gives an agent
12 rules. 0 are common, 10 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
10
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 17
Add each page that the person names.
- Distinctive: line 34
Check that the sitemap lists each picked page.
- Very specific: line 35
Check that the sitemap holds no address that redirects, returns an error, or has noindex.
- Distinctive: line 36
Check that each page has a title and a meta description that say what the page is about.
- Distinctive: line 37
Ask the person to read the index status of the main pages in Google Search Console and in Bing Webmaster Tools.
- Very specific: line 38
If the person cannot open those tools, a site: search is a rough check.
- Distinctive: line 41
If the site sends new addresses with IndexNow, a 200 response means "received".
- Distinctive: line 88
Use evidence that you saw.
- Distinctive: line 89
Not verified means that a check did not run.
- Distinctive: line 91
Never write a fact for the business.
- Distinctive: line 92
If you work in the site's code and the person asks, make the changes on a branch for the person to review.
- Distinctive: line 93
Do not say that a change will raise citations or mentions.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Site audit
The ai-readiness skill checks whether AI apps can read the home page. This skill checks the main pages of the site, and what those pages say.
An AI app can quote only what a page states in text. So the audit has four parts: access, indexing, facts, and pages that can be quoted.
1. Pick the pages
- Fetch
https://<site>/robots.txt. Read itsSitemap:lines. If there is none, tryhttps://<site>/sitemap.xml. - Fetch the home page, with its response headers.
- Pick up to 12 pages: the home page, the pricing page, each main product or service page, the About page, the contact page, the comparison pages, and the guides that the home page links.
- Add each page that the person names.
- Say which pages you read. The audit covers those pages only.
2. Access
For each page, do the five checks of the ai-readiness skill for that page's path. Then check three more things:
| Check | Passes when |
| Status | The address returns the page itself. It does not return an error or a redirect to a different page |
| Canonical | The page's canonical link is the page's own address, or the address that the owner intends |
| No leftover block | No noindex remains from a test version of the site |
3. Indexing
Google's AI features read Google's index. Copilot reads Bing's index. A page that is in neither index cannot appear in those answers.
- Check that the sitemap lists each picked page.
- Check that the sitemap holds no address that redirects, returns an error, or has
noindex. - Check that each page has a title and a meta description that say what the page is about.
- Ask the person to read the index status of the main pages in Google Search Console and in Bing Webmaster Tools. Those two tools are the only sources of the real status.
- If the person cannot open those tools, a
site:search is a rough check. Say that it is rough. - Ask the person to read two reports, if their account shows them. Google Search Console has a generative AI performance report. Bing Webmaster Tools has an AI Performance report. On November 13, 2026, the Google report showed impressions only, and a small site can have no report. Record "not visible" when a report is absent.
If the site sends new addresses with IndexNow, a 200 response means "received". It does not mean "indexed".
4. Facts
A buyer asks an AI app about eight facts. For each fact, record the page and the exact sentence that states it, or write "not stated".
| Fact | The site states it when |
| What the business sells | One sentence gives the product or the service and its category |
| Who it is for | A sentence gives the buyer: a job, a business type, or a need |
| Where it works | A sentence gives the countries, the cities, or the service area |
| Price | A page gives the price, or the things that set the price |
| Limits | A page says what the product does not do, or who it does not fit |
| How it differs | A page compares the business with alternatives that it names |
| Proof | A page gives customers, reviews, or numbers, each with its source |
| Who is behind it | A page gives the people or the company, and a way to contact them |
Then check two things for each fact that the site states:
- The fact is in the text. A fact that is only in an image, a video, a PDF file, or a part of the page that needs JavaScript does not count.
- The fact is the same on each page. Compare the name, the one-line description, and each price across the pages. Report each conflict with both quotes and both addresses.
5. Pages that can be quoted
For each picked page, read the HTML that the server sends. Check:
| Check | Passes when |
| Direct answer | The first paragraph says what the page is about in two or three sentences that make sense alone |
| Clear headings | Each heading gives the subject of its section |
| Sections that stand alone | A section gives the name of the product or the business. It does not rely on "it" or "our tool" from an earlier section |
| Specific facts | The page gives numbers, names, and dates. It uses "best" or "leading" only with a source |
| One subject | The page covers one use, one product, or one question |
| A date | A page with facts that change, such as prices, shows when it was last checked |
To test a section, read it without the rest of the page. If you cannot tell which business and which product it describes, the check fails.
Order of the changes
- Access: a blocked crawler, a
noindex, or text that needs JavaScript. A page that AI apps cannot read gains nothing from the other changes. - Indexing: the sitemap, the canonical link, and the missing titles and descriptions.
- Facts that conflict, then facts that the site does not state.
- Pages that cannot be quoted.
- Pages to add. Use the
page-draftingskill for each one.
Rules
- Use evidence that you saw. Quote the line, the tag, or the sentence, with the page address.
- Not verified means that a check did not run. It is not a failure. Give no change for it.
- Make each change exact: the page, the text there now, and the new text.
- Never write a fact for the business. If a fact is missing, ask the person for it.
- If you work in the site's code and the person asks, make the changes on a branch for the person to review.
- Do not say that a change will raise citations or mentions.
Never suggest
llms.txtfiles or "AI schema" as a fix. They are unproven. Google states that its own AI features need no new files and no special markup.- Text that only AI apps can see, or hidden text of any kind.
- A keyword repeated to fill a page.
Output
- A table of the pages that you read: the page, access, indexing, and the checks of part 5 that fail.
- The eight facts, each with its sentence and page, or "not stated".
- The conflicts, each with both quotes.
- The changes, in the order above. For each one: the page, the evidence, and the exact change.
- What you could not check.
End with one line: the audit shows what the site says. It does not show whether AI apps mention the business. For that, use the mention-check skill.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

