2026 Technical SEO & AEO Audit Checklist with Jev

If a site looks complete in your browser but returns little usable HTML to a crawler, the first problem is not an AI citation rate. It is that the page may have no accessible fact to cite. An AEO audit checklist starts there, before it asks whether ChatGPT or another engine mentions you.

Start with one domain, not a spreadsheet: run SearchD’s free technical SEO + AEO audit. SearchD measures what its crawler can see in public HTML, including fact tables, and produces a checklist with evidence and next actions. Jev, when available, assesses short text excerpts for clearly named subjects and specific detail — questions HTML structure alone cannot answer. The 18 questions below also include items that need a live engine test or a full link graph; those stay unknown or unverified, not silently passed.

Can AI crawlers retrieve your pages?

If your server or firewall refuses a request, better writing and schema will not fix what that request could not retrieve. Start your technical SEO and AEO audit here.

1. Can AI-search user agents reach the homepage? Check the HTTP responses to OAI-SearchBot, ClaudeBot and PerplexityBot requests. SearchD shows the status or failed request for each root probe; a user-agent string alone cannot prove an official crawler IP gets through.
2. Do pages return meaningful text in their raw HTML? Compare View Source with what the browser displays after JavaScript runs. SearchD reports the average extracted-text length of the HTML it fetched, not what a browser assembled later.
3. Does robots.txt allow the access you intended? Check root rules for those user agents and the wildcard group. SearchD reports explicit root blocks and leaves failed probes unknown; its check does not evaluate every path-specific rule.
4. Does each page identify its own canonical URL? Compare the page URL with its canonical link. SearchD counts missing or non-self-referencing canonicals in fetched pages; it does not fetch the declared target to confirm that target is live.

Can an answer engine extract a useful passage?

Retrieval is only the first gate. A page also needs a clear heading and a fact that makes sense when lifted out of context.

5. Do H1–H3 headings describe the page and its sections? Read the outline without the body copy. SearchD checks for one H1 and local heading-level jumps on retained pages; passing that structural check does not prove the headings say anything useful.
6. Are specifications and comparisons real HTML tables? A labelled table can keep a value attached to its product or attribute. SearchD detects candidate fact tables in fetched HTML; it cannot confirm every cell is correct.
7. Can a sentence stand alone with its subject named? “SearchD checks public HTML” survives extraction better than “It checks the page.” Jev can assess a bounded excerpt for named subjects when available; the site-wide standalone-sentence row remains unverified.
8. Are important pages substantial enough to answer their question? Inspect pages flagged as thin instead of padding every short page. SearchD counts fetched pages below 500 extracted text characters as review candidates, not automatic SEO failures.

Do links make the right pages discoverable?

A useful answer on the wrong or unreachable page will not help someone find it. The current audit gives link evidence, not a complete map of the site.

9. Does one canonical page own each important search intent? Compare pages that answer the same buyer question and resolve contradictory claims. SearchD does not cluster intents or certify the absence of cannibalization; this row is unknown.
10. Can navigation or relevant pages lead to every priority URL? Check the full indexable URL set against internal paths. SearchD cannot prove that no pages are orphaned from its limited discovery crawl; this row is unknown.
11. Are links placed next to the facts they explain? Use descriptive anchors in relevant paragraphs or lists, not only in navigation. SearchD counts fetched pages with links in those contexts, but that count is not a complete internal-link graph.

Can machines identify your organization and its facts?

Structured data is useful when it describes the same real thing the visitor sees. It is not a substitute for visible, accurate content.

12. Is your JSON-LD parseable on the pages that need it? SearchD counts fetched pages with parseable JSON-LD and lists the @type values it finds; syntax alone cannot validate every entity relationship.
13. Do sameAs links point to profiles you actually control? SearchD can find valid-looking profile URLs in JSON-LD, but cannot verify ownership. Confirm each profile yourself; the ownership row stays unverified.
14. Do visible claims match structured-data claims? Compare names, prices, availability and dates in the page with its markup. SearchD does not perform a field-by-field parity check, so this row stays unverified.

Is your site actually appearing in AI answers?

Technical readiness is a prerequisite to investigate, not proof of a citation. Finish the checklist with tests the free technical audit does not pretend to run.

15. Does ChatGPT name or cite you for real buyer questions? Save the prompts and the answers, then check the cited URLs. SearchD’s technical audit does not run those questions; this row is unverified.
16. Can Claude or Cursor extract your documentation accurately? Test specific tasks and compare the output with the underlying page. The free audit does not execute this extraction test.
17. Do Gemini or Google AI Overviews include you? Measure the same target questions with a recorded locale and date; the free audit does not compute an AI Overview win rate.
18. Are your search-engine reporting gateways configured? Verify sitemap submission and indexation in Google Search Console and Bing Webmaster Tools. SearchD cannot log in to those accounts for you.

How does Jev turn this checklist into an audit you can use?

SearchD collects public HTML from your domain and reports the evidence it actually found, up to its crawl deadline. The resulting 18-row checklist shows an observed status and next step for each item when at least one readable page is retained. If it cannot retain a readable page, it does not invent a score.

Jev helps where a structural check stops: SearchD supplies whole-page HTML table counts, while Jev assesses a bounded text excerpt for clearly named subjects and specific factual detail. The table counts come from SearchD’s HTML checks, not from reading that excerpt. Jev’s typed answers are supplemental, not the source of the technical score; if a call is unavailable, SearchD does not make one up.

The video above shows a completed searchd.ai audit replay, not a live crawl. That recorded run retained 53 readable pages and completed 53 Jev calls; your domain will have its own page count, findings and run time. Our original Jev demo post shows why we built it, but the useful next step is to inspect your site’s evidence.

What should you do after the audit?

Start with a result you can confirm: a blocked root request, missing raw-HTML text, an absent canonical or a page with no clear subject. Fix the page or access rule you control, then rerun the technical audit. For a weak Jev excerpt, review the exact text it assessed before changing the whole page; the guide to quotable passages shows how to make a fact stand alone.

For any row marked unknown or unverified, arrange the separate check instead of treating the technical score as a citation forecast. Ready to see what your site exposes to AI search today? Enter your domain and get the checklist with the underlying evidence and next actions.

Run the free AEO + SEO audit on your site