• Playbook
  • AI Search
  • 5 min read

The Nine-Check Audit in One Week

Technical checks first because they are fast, trust checks because they are serious, coverage and baseline because they become the plan.

Author
ftes.ai Editorial
ftes.ai Team

The short answer

The nine-check AI search audit runs in one working week: Monday, content and schema without JavaScript; Tuesday, llms.txt and entity consistency; Wednesday, answer-first openings and chunk self-containment on key pages; Thursday, verifiability of sampled claims; Friday, fan-out coverage and the sampled measurement baseline. Each check has a pass condition, and the fails in order are the content plan.

The role this piece is aboutSearch Visibility FTE · AI Search Visibility Engineer

Key takeaways

  • One week, one check-group per day, each with a pass condition you verify yourself.
  • Technical fails are fixed as they are found; they are fast.
  • The verifiability check is the serious one; invented claims are fixed before anything else publishes.
  • The measurement baseline on Friday is what makes every later change judgeable.
  • The fails, in order, are the next quarter's plan.

Monday: crawlability and schema (checks 1–2)

Fetch the key pages as a crawler that executes no JavaScript and confirm the main content, headings, and JSON-LD are in the initial HTML; on ftes.ai this is a build requirement, and on any site it is the first thing to check. Then validate the schema: Organization on the root, Article or BlogPosting with author and dates on content, FAQPage only where a visible FAQ exists, BreadcrumbList, Product or Service where relevant, and nothing describing content that is not on the page.

Fix fails as they are found. These are the fast ones, and a page invisible to a scriptless crawler is invisible to most AI engines.

Tuesday: llms.txt and entity consistency (checks 3 and 6)

Check llms.txt exists at the root, is current, and reads as an honest map: site name, a plain summary, grouped links described by what each page answers. Then run the entity check: list the brand, products, and key concepts, record every name variant across the site and the listings you control, and find the canonical definition for each. Two names for one thing, a definition that differs between pages, a price that differs: each is a finding.

Unify the names now where the fix is a find-and-replace; schedule the rest.

Wednesday: answer-first and chunk self-containment (checks 4–5)

For each key page, read the opening: does it give a direct 40 to 60 word answer to the page's primary question, starting with a definition or verdict, or does it begin with an introduction? Then sample sections with the rest of the page hidden: does each name its entity in the first sentence, answer its heading, and stand alone with no cross-references? List the failing pages.

This is where most sites fail: findable, not citable. The failing pages that already rank are the first rewrites.

  • Opening: direct answer, 40–60 words, definition or verdict first
  • Headings: questions or claims, never bare labels
  • Sections: entity named, question answered, no above or below
  • List: failing pages, ranked by current search position

Thursday: verifiability (check 7)

Sample the claims on the key pages and confirm each is a product mechanism, shown arithmetic, or a named public source. Any statistic, study, customer, case study, or quote with no source is a fail, and the most serious finding in the audit: engines cross-check sources, and an uncorroborated claim marks the page and the domain as a risk. Remove or source every one before any new page publishes.

Record what was found and what was done; the record is part of the domain's trust.

Friday: fan-out coverage and the measurement baseline (checks 8–9)

For each primary question the site wants to be cited for, list the five to nine likely sub-queries and mark which are answered in a self-contained section somewhere in the cluster. The gaps are the content plan. Then write the prompt set from the primary questions and their sub-queries, sample it once across every engine that matters, recording citation presence, position, competing sources, and answer text with the date, and set the weekly schedule.

Without Friday's baseline nothing done afterwards can be judged; it is the day that cannot be skipped.

One questionfrom a buyerwhatsectionhowsectioncostsectionrisksectionversussectionwhensectionsetupsectionEACH SUB-QUERY RETRIEVES PASSAGES ON ITS OWN
Figure 1Friday's coverage check in one picture: each primary question fans into sub-queries, and each sub-query needs a self-contained section somewhere in the cluster. The unanswered ones are the plan.

What does done look like on Friday evening?

Done is a short document, dated: nine checks with pass or fail and the evidence, technical fails already fixed, the entity findings with the unifications made, the list of failing pages ranked by position, the verifiability findings with what was removed or sourced, the fan-out gaps as a content plan, and a sampled baseline with a schedule. A Search Visibility Digital FTE runs this audit as a mission and produces exactly this document; the week is the same either way.

The next quarter's work is now a list, in order, with a way to measure it.

What should you refuse to do during the audit?

Refuse to publish anything new until Thursday's verifiability fails are fixed. Refuse to add schema for content not on the page to pass a check. Refuse to skip Friday because the week ran long; the baseline is the point. And refuse to treat any single check as the whole answer; the nine together are the audit, and the order matters.

Run it again next quarter against this one, and the site's progress becomes a record.

Frequently asked questions

What is checked on Monday?

That content and schema appear without JavaScript, and that the schema is complete, valid, and describes only what is on the page.

Which check is most serious?

Thursday's verifiability. Invented claims mark a page and domain as risky and are fixed before anything new publishes.

What is the most common result?

Passing the technical checks and failing the structure checks on Wednesday: findable but not citable.

Why is Friday essential?

The sampled measurement baseline is what makes every later change judgeable. Without it, nothing can be compared.

Can a Digital FTE run the audit?

Yes. A Search Visibility FTE runs it as a mission and produces the same dated document with its findings and plan.

  • Playbook
  • AI Search
  • Audit
  • GEO
  • One week
Further reading

Keep going.

Related guides on hiring and managing Digital FTEs.

Explainer · AI Search · 5 min

An AI Search Visibility Audit in Nine Checks

A repeatable audit for whether a site is ready to be cited by AI engines: crawlability without JavaScript, schema, llms.txt, answer-first structure, chunk self-containment, entity consistency, verifiability, fan-out coverage, and a measurement baseline.

Read

Playbook · AI Search · 5 min

Playbook: From Audit to First Citations in 90 Days

A quarter with a Search Visibility FTE in three phases: fix the technical foundation and record the baseline, rewrite the pages that already rank, build the cluster from the gaps, and read the first honest comparison.

Read

Checklist · AI Search · 5 min

Checklist: Ten Checks Before a Piece Publishes

The pre-publish checklist for any page, guide, or whitepaper: answer-first opening, question headings, self-contained sections, fan-out coverage, information gain, verifiable claims, voice-ready FAQs, video-ready H2s, three related links, complete metadata.

Read

Ready to hire your first Digital FTE?

Pick a role, connect the tools it needs, and review its first proposals this week.