Jake Diego.
Case study

An AI-visibility audit of this site: find the gaps, fix them, measure the result

The clearest proof of a capability is using it on yourself. I ran a structured AI-visibility audit on this very site, found where AI answer engines could not read it, and shipped the fixes. Here is the before and after.

What is an AI-visibility audit, and why run one on your own site?

An AI-visibility audit checks how well a site is positioned to be found, trusted, and accurately quoted by AI answer engines such as ChatGPT, Perplexity, and Google AI Overviews.

The finding on this site was specific: strong, citable content sitting on almost no machine-readable plumbing. The writing was already unusually well suited to how AI systems read the web, clean semantic HTML, one heading per page, real and dated figures like the $30,590 in retained revenue. What was missing was the layer those same systems now rely on to decide who to cite: no robots.txt, no sitemap, no schema, no canonical tags. The raw material was excellent and the wiring that carries it to an answer engine was not there.

How was the audit run?

As a six-stage multi-agent pipeline built to catch weak claims and invented numbers before they reached a conclusion, not as one reviewer's opinion.

Stage 1 · 4 parallel domain audits
Domain 1
AI content indexing
Domain 2
Vector & embedding readiness
Domain 3
GEO citation
Domain 4
AEO snippets & FAQ
Stage 2 · independent validation
Validation
Re-ran every file read and grep from Stage 1, trusting none of its prose
Stages 3–5 · expert debate
Round 1
Prioritization debate, argued from evidence
Round 2
Fresh data gathered between rounds, re-argued
Stage 6 · final validation gateGO
4 audits independent domain passes, each catching what the others missed
0 invented fabricated or misquoted figures found across all four audits
15 of 15 demo pages confirmed to share one identical title, verified in the final gate
GO final gate verdict, with every action item lineaged to a real finding

What did the audit find, and what changed?

A strong content foundation with a near-total gap in the machine-readable layer, all of it additive to fix rather than a rewrite. Here is the before and after, domain by domain.

DomainBefore the auditAfter the fixes
Indexing plumbing No robots.txt, no sitemap, no canonical tags robots.txt welcoming ten AI and search crawlers, a focused sitemap, canonical on every page
Structured data Zero JSON-LD schema anywhere on the site Person, WebSite, Article, and FAQPage schema, with a safe non-invented sameAs
Number integrity Real $30,590 sat next to an illustrative $60,000 behind a fragile trailing label; a duplicate A/B stat panel shipped twice in raw HTML Illustrative labels front-loaded as the first words, one canonical stat panel
FAQ and snippets No FAQ, no question-shaped headings Five headings rewritten as questions, a six-question FAQ, matching FAQPage schema
Comparison framing Nothing named or contrasted an alternative A "How this is different" section with four comparison cards
Topical clustering Case studies siloed, no sideways links "Related systems" cross-links tying the portfolio into one cluster

Every figure above traces to a finding confirmed twice against the actual site files. No gap was asserted on faith, and no number was invented to make the story cleaner.

What was actually shipped?

Seven focused changes, each one additive and traceable to a confirmed finding.

  • Indexing plumbingAdded a robots.txt that explicitly allows ten major AI and search crawlers (including GPTBot, ClaudeBot, PerplexityBot, and Google-Extended), a sitemap covering the real content pages, and a canonical tag on every top-level page.
  • Number integrityFront-loaded "Illustrative example, not a verified client outcome" as the literal first words of the example stat block, so a text extractor cannot mistake the illustrative $60,000 for the real $30,590 result.
  • A/B panel retiredRemoved the duplicate stat panel and its picker script so only one canonical set of numbers now ships in the raw HTML.
  • Question headings and FAQRewrote five declarative headings into question form with a one-sentence lead answer, and added a visible six-question FAQ, all reusing copy already on the page with zero new claims.
  • Cross-linksAdded "Related systems" blocks to the case studies and the cost-tracker so the portfolio reads as one coherent set of measured systems.
  • Comparison contentAdded a "How this is different" section contrasting the approach with agencies, off-the-shelf tools, freelancer marketplaces, and no-code platforms.
  • Structured dataAdded Person and WebSite schema to the homepage, Article schema to each case study, and FAQPage schema paired with the visible FAQ, using an empty-by-honest-default sameAs rather than inventing links.

What is the outcome?

The site is now positioned to be found, trusted, and quoted correctly by AI answer engines, where before it was largely invisible to them.

These fixes shipped recently, so this write-up reports what changed and why, not a traffic or citation result. There is no before-and-after visibility metric to claim yet, and inventing one would defeat the entire point of an audit built to reject unverifiable numbers.

What can be said honestly is concrete. An AI crawler that previously had to discover pages by link-luck now has an explicit welcome and a map. An engine assessing who Jake Diego is now has a machine-readable entity to name instead of unstructured text it cannot confidently attribute. And the one figure the whole brand rests on, the real $30,590, is no longer one careless extraction away from being reported as the fictional $60,000. The audit closed on a GO precisely because the upside was high, the risk of change was low, and none of it required rewriting what the site says.

How does this fit the rest of the work?

This is one capability demonstrated under the existing "AI ops systems" brand, not a separate service line.

The same discipline runs through every system on this site: measured baselines, claims traced to source, and results reported honestly rather than estimated. This audit is that method turned on the site itself, which makes it a working example of the systematic, measured approach rather than a new offering bolted on beside the others.

Related systems

Customer-Success Command Center · a command center that replaced five spreadsheets with one screen and credited $30,590 in retained revenue.

Market Intelligence Pipeline · an automated four-stage pipeline that turns raw public data into a ranked opportunity list.

Price History Tracker · stitches every price sheet into one self-refreshing time series, built in a day.

Video-to-Documentation Pipeline · turns one monthly walkthrough recording into four polished, branded release documents automatically.

Back to projects