My Work

I Showed a 74/100 SEO Score Hiding Zero AI Citations

3 min

Last updated,

Summary

A 25-year call-center outsourcing advisory had a 74 out of 100 SEO health score, 1,168 ranking keywords and roughly 4,500 organic visits a month. Asked the six questions its buyers ask, AI assistants and Google AI Overviews cited it zero times, a 4% share of voice. I found all of it as a prospect, with no analytics access.

A well-run operation with a blind spot

This wasn’t a neglected site. Someone had done the work: clean structure, 470 indexed pages, more than a thousand ranking terms, a health score most B2B firms would frame. The organic traffic was steady and the business behind it was 25 years old.

The blind spot was where the buyers were going. A company deciding whether to outsource its call center now asks an assistant to shortlist advisors. That conversation doesn’t appear in a rank tracker.

A 74/100 health score measures how well a site is built for the search engine of 2015. It says nothing about whether an AI assistant will recommend the company in 2026.

Auditing without the keys

I had no Search Console, no analytics, no login. The company was a prospect, not a client.

That constraint shapes the method. I ran my Audit Engine in prospect mode, and the audit carries its own section on what it could and couldn’t see.

  • Ranked keywords and domain overview from a third-party data provider, to see what the site ranked for and how strong it was.
  • Live search results for six head, category and brand questions, with detection of whether Google showed an AI Overview and which domains it cited.
  • The same six questions to AI assistants, with the cited sources captured.
  • On-page scoring of the homepage and five key pages.
  • A live crawl of the raw HTML: structured data extraction, the robots and AI-crawler files, eleven child sitemaps, redirect chains, and the full set of security headers.

Everything a prospect can see, nothing a prospect can’t, and the limits of that written down.

Cited zero times

Across the six buyer questions, the firm was cited in zero AI answers. Its share of voice in AI, the fraction of citations going to it rather than competitors, was 4%.

Six questions is a small sample and AI answers move week to week, so the finding shipped with its sample size and date attached rather than as a headline: six questions, one date, zero citations, 4% share.

A 25-year firm with a 74/100 score was absent from the conversations where its next clients were being shortlisted.

What the score wasn’t measuring

The gap between the SEO score and the AI result came down to a few concrete things. The pages ranked but didn’t answer; a buyer’s question was addressed across three paragraphs rather than in the first one. The structured data was thin.

There was no plain-language file telling AI crawlers what the firm did and for whom. And the content organized itself around the firm’s services rather than around the questions a buyer asks an assistant.

None of that dents a health score. All of it determines whether a company gets cited.

What I delivered

Fifteen prioritized recommendations and a 30-60-90-day roadmap, ordered by what a small team could ship first. The technical fixes were specified with the evidence attached.

The content plan was built around the six buyer questions and the ones behind them. And because it ran on my Audit Engine in prospect mode, the same engine behind fifteen client audits, it went out alongside eight others in a single day.

At a glance

  • The niche: Call-center outsourcing advisory, 25 years in business, WordPress.
  • The problem: Strong traditional SEO (74/100, 1,168 keywords) with no AI visibility, and no analytics access to audit with.
  • What was built: My Audit Engine in prospect mode, 12 sections: third-party rank data, live search results with AI Overview detection, six buyer questions across assistants, on-page scoring, a full live crawl.
  • Implementation: Method limits documented in their own section; 15 prioritized recommendations and a 30-60-90 roadmap.
  • Results: Zero AI citations and a 4% share of voice against a 74/100 score; delivered alongside eight other proposals in a single day.
  • Built with: my Audit Engine in prospect mode.

Frequently asked questions

Yes, with stated limits. Third-party rank data, live search and AI results, and a crawl of the public site cover most of a technical and AI-visibility audit. What you cannot see is the site’s own click and conversion data, and a good audit says so in its own section.

Read next

Agencies charge $800 to $2,000 per article. I built a multi-tenant factory that produces SEO and AI-search articles for about $12 to $15 a month per client, queues up to 1,000 a week, and gates every piece with a 30-point scorecard plus a link check that overrides the AI. Then I built the gate that can block our own work.

4 min

A gatekeeper checking scrolls before a messenger may pass

A DTC dress-shirt brand's Search Console showed 17,077 clicks from 1.39 million impressions in 50 days. A brand-versus-non-brand split showed 84% of those clicks were people typing the brand name, and its category term ranked at position 5.6 with a 0.37% click rate. Same trap I'd found at a marketplace weeks earlier.

2 min

A busy tailor's courtyard with an empty road beyond the gate

A 12-location therapy practice needed per-market SEO and AI visibility in a vertical where a wrong claim is a compliance problem. It became the reference implementation for my Audit Engine's growth-plan layer, and the engagement where every defect the system could produce was caught and turned into a permanent check.

2 min

Twelve small herb gardens connected by paths across a valley

Pitching SEO and AI-search work to eight accounts meant eight discovery cycles. I ran one live outside-in scan across all eight in a day, for about 2,100 rank-data credits, and deployed nine client-ready proposals the same afternoon. Three of the eight ranked for zero keywords.

2 min

A surveyor walking between distant estates at the end of the day

SEO delivery is normally a shared template and a shared team. Every client of mine runs in a dedicated environment with its own tooling and context layer. I built a hosted, ten-section systems tour to show a private equity firm's Head of AI how it works, with every figure dated and sourced and the unbuilt parts labeled. He probed the isolation model and didn't break it.

2 min

A patron being shown a workshop of classical automata

A Detroit-metro home-services brand consolidating two longstanding trade businesses under a new name became a full tenant of my Article Factory: its own isolated project, a deployed dashboard, and a complete eight-file context layer including who the business is not for. The build took one command and about fifteen minutes.

2 min

A steward handing a key at the door of a newly opened house