Skip to main content
FirstCoast.ai
All case studies
Case study: B2B marketing agency

AI Search Visibility Audits

Buyers increasingly ask AI engines instead of search boxes. We built a single audit covering Search Engine Optimization (SEO), Generative Engine Optimization (GEO), and Answer Engine Optimization (AEO), with 21 deterministic checks and every claim backed by evidence. It caught a bug that had made an agency's site invisible to every AI crawler.

How the AI search visibility audit worksBuyers increasingly ask AI engines instead of search boxes, so this audit measures what an engine can see. Collection is passive and read-only, and re-runnable on a schedule: robots and sitemap analysis, DNS and email-authentication lookups, page-speed APIs, and crawler-access probes, with nothing intrusive and nothing written. Those readings feed 21 deterministic checks across three pillars, each mapped to a discipline that measures independently of the others. Technical readiness is classic search engine optimisation: crawlability, structured data, and site speed. Content architecture is generative engine optimisation: whether the content itself is structured to be cited by generative engines. Agent readiness is answer engine optimisation: whether the business gets named when someone asks an AI a direct buying question, which an AI probe tests by sampling real buyer-phrased questions and recording the result as cited, mentioned, or absent. Every check then resolves to exactly one of three states. Pass and fail are counted in the score and ranked. Unmeasured is counted in neither: a check that cannot run is never recorded as a pass and never as a failure, and the audit library carries 162 automated tests so that honesty is enforced rather than promised. Both paths land in one report. Pass and fail become a scored, evidence-backed backlog where every item is ranked by impact against effort and split by who owns the fix, which makes it a work plan rather than a scare sheet. Unmeasured becomes the disclosure block stating what the audit could not measure, which the tests require to appear in every report. On the agency's own site the audit caught a one-character case-sensitivity bug in a geo-redirect that had exempted zero AI crawlers, leaving the homepage visible to Google and effectively invisible to every AI answer engine. It was fixed the same week and technical readiness moved from 63 to 80 with no content changes at all. The same run also surfaced a regional-tag misconfiguration that had invalidated language targeting across three related sites.COLLECTIONREAD-ONLYWhat it sees from outsideRobots and sitemapDNS and email authPage-speed APIsCrawler-access probesPASSIVE AND READ-ONLYre-runnable on a scheduleTHE 21 CHECKS, IN THREE PILLARSSEOTechnical readinesscrawlability, structured data, site speedGEOContent architecturestructured to be cited by generative enginesAEOAgent readinessdoes the business get named when askedan AI probe asks real buyer questionsCitedMentionedAbsentTHE VERDICT ON EVERY CHECKTHREE STATESPass, fail, or unmeasuredPassin the scoreFailscored and rankedUnmeasuredin neithera check that cannot run is nevera pass and never a failurethe audit library carries 162 testsHONESTY IS TEST-ENFORCEDPASS AND FAIL, SCOREDUNMEASURED, DISCLOSEDWHAT THE CLIENT GETSEVIDENCE-BACKEDA scored, ranked backlogevery item ranked by impact against effortand split by who owns the fixa work plan, not a scare sheetWHAT IT CAUGHT63 to 80technical readiness, zero content changesa one-character case bug had exemptedzero AI crawlers from a geo-redirectTHE DISCLOSURE BLOCKUnmeasuredwhat the audit could not measureREQUIRED IN EVERY REPORT

The challenge

  • A B2B agency needed to know how its sites, and its clients', appear when AI engines answer buyer questions
  • Most AI-visibility advice is vibes; the agency needed measurements that could be re-run and defended
  • Nobody knew that a routine infrastructure rule was quietly blocking AI crawlers

What we built

An audit engine of 21 deterministic checks across three pillars, each mapped to a discipline: technical readiness (classic SEO: crawlability, structured data, site speed), content architecture (GEO: whether the content itself is structured to be cited by generative engines), and conversational agent readiness (AEO: whether the business gets named when someone asks an AI a direct buying question). It produces a scored, evidence-backed report with a priority backlog ranked by impact against effort.

  • Every check is deterministic and repeatable; an AI probe samples real buyer-phrased questions and records cited, mentioned, or absent
  • A check that cannot run is reported as unmeasured, never counted as a pass or a failure, and that honesty is enforced by the test suite
  • One audit covers SEO, GEO, and AEO together, so a business sees where it ranks, gets cited, and gets recommended in a single report
  • Findings split by who owns each fix, so the report is a work plan, not a scare sheet

How it's built

A standalone audit library with 162 automated tests, compiled into a scheduled workflow. Collection is passive and read-only: robots and sitemap analysis, DNS and email-authentication lookups, page-speed APIs, and crawler-access probes. The disclosure blocks stating what the audit cannot measure are required by the tests to appear in every report.

Results

  • Caught a one-character case-sensitivity bug in a geo-redirect that had exempted zero AI crawlers, leaving the homepage visible to Google and effectively invisible to every AI answer engine
  • Fixed the same week with a tested replacement; technical readiness moved from 63 to 80 with zero content changes
  • Also surfaced a regional-tag misconfiguration that had invalidated language targeting across three related sites

Do you know what AI engines say about your business?

The audit measures it as cited, mentioned, or absent, and hands you the ranked fix list.