We were our own first client

Before offering the AI Visibility Audit to anyone else, we ran it on ourselves. Our old website scored 35/100 — "invisible to AI". One rebuild later, measured with the exact same 33 tests: 90/100, AI-ready.

3590/ 100 AI Readiness Score

The starting point

In early July 2026, appsynth.pl was a typical software-house website. It looked fine to humans — and was nearly invisible to machines. The audit made the problem concrete: no structured data at all (0/15), no question-and-answer content structure (0/15), key content only appearing after JavaScript ran (7/15), weak entity signals (3/10) and a thin technical foundation (2/12). Total: 35/100 — the "invisible to AI" band of our scale.

That is exactly the profile we see across the market: our benchmark of 447 Norwegian trade businesses has a median score of around 70, and the most common failures are the same ones our own site had.

What we changed

The fix list came straight out of the audit — the same way it does for clients:

  • Content in raw HTML. The site is statically rendered: every word is readable without executing JavaScript, which AI crawlers do not run.
  • Structured data. JSON-LD describing the organisation (name, registration numbers, address, services) and an FAQPage — so AI does not have to guess who we are.
  • Question-shaped content. A real FAQ section, descriptive headings, one H1 per page, clean heading hierarchy.
  • Unambiguous entity. Consistent company name across title, structured data and footer; registration numbers and full address in the HTML.
  • Technical foundation. sitemap.xml, canonical URLs, hreflang between the English and Norwegian versions, forced HTTPS — and speed (largest contentful paint around 150 ms).

The result, category by category

A. Accessibility for AI crawlers2325 / 25
B. Content without JavaScript715 / 15
C. Structured data (JSON-LD)013 / 15
D. Content structure015 / 15
E. Entity clarity310 / 10
F. Technical foundation212 / 12
G. Actual AI visibility00 / 8
before rebuildafter rebuild

Both measurements: July 2026, methodology v0.1 — the same version before and after, as always. How the score works: read the full methodology.

What about actual AI answers?

Category G — what AI assistants actually say about a business — scored 0 in both measurements, and we left it that way on purpose. Technical readiness is immediate; AI knowledge is not. Assistants keep serving cached or trained impressions of a site for weeks after a rebuild, and pretending otherwise would be exactly the kind of overpromising our methodology exists to prevent.

The lag is now closing on schedule: as of mid-July 2026, ChatGPT reads and correctly describes the new site — where it previously showed a stale cached version. The live G tests are the next measurement, and this page will be updated with the results, whichever way they land.

The same playbook we sell

Nothing about this rebuild was special: audit, prioritized fix list, implementation, re-measurement with the same tests. That is the exact process we run for clients — starting with a free mini-audit of your website.

Get a free mini-audit → How we measure