AI can read every word
Before, the content only appeared after the browser ran extra code, and AI doesn't run it. Now every word is in plain text from the start.
Case study · AppSynth
We were our own first client. 35 → 90.
Before we offered the AI Visibility Audit to anyone, we ran it on our own website. The old site scored 35 out of 100, which our scale calls "invisible to AI". After one rebuild, measured with the same 35 tests, it scored 90. The measurements were taken before we moved from appsynth.pl to appsynth.eu.
+55points, same 35 tests
150 mslargest contentful paint
#1in Gemini · not yet in ChatGPT or Perplexity
35before
90after
Starting point · 35/100
Invisible to AI.
In early July 2026, our website was a typical software-house website. People liked how it looked. Machines could barely see it. AI could hardly read the content and had no idea who we were, where we worked or what we offered. The technical basics were thin too. That added up to 35 out of 100.
We see the same profile across the market. Our benchmark of 419 Norwegian trade websites has a median score of 70, and the most common failures are the ones our own site had.
Content only visible after JavaScript
No clear identification of the company
No structured data, no FAQ
Weak technical foundation
What we changed
Five changes, one rebuild.
The fix list came straight out of the audit, the same one clients get. It came down to five things:
AI knows who we are
We labelled the company name, registration number, address and services, so AI doesn't have to guess or mix us up with someone else.
Answers to the questions people ask
A proper FAQ section, and headings phrased the way a customer would ask.
The same name everywhere
The company name reads the same in the title, on the page and in the footer. Small differences confuse people and AI alike.
A fast, tidy site
The site loads almost instantly, always runs over HTTPS, and has the technical basics a search engine expects.
The technical details, if you want them
- Content in raw HTML. The site is statically rendered, so every word can be read without running JavaScript. AI crawlers don't run it.
- Structured data. JSON-LD for the organisation (name, registration numbers, address, services), plus an FAQPage.
- Question-shaped content. An FAQ section, descriptive headings, one H1 per page and no skipped heading levels.
- Unambiguous entity. The company name is the same in the title, the structured data and the footer. Registration numbers and the full address are in the HTML.
- Technical foundation. sitemap.xml, canonical URLs, hreflang between the English and Norwegian versions, and forced HTTPS. Largest contentful paint is around 150 ms.
Category by category
Where the 55 points came from.
We took both measurements in July 2026 with methodology v0.1. Before and after always use the same version. Category G stayed at zero on day one, because AI assistants need time to update what they know.
Live AI tests · six days after launch
5 of 8 points in the first week.
Category G measures what AI assistants say about a business. It scored 0 in both measurements, and we left it there on purpose. Technical fixes count from the day they go live. What AI knows takes longer: assistants keep working from cached or trained impressions of a site for weeks after a rebuild. Claiming otherwise would be the kind of overpromising our methodology is meant to prevent.
Update, 16 July 2026: six days after the rebuild reached AI assistants, we ran the standard category G test. That means three fixed questions in ChatGPT, Perplexity and Gemini, with screenshots kept. A check passes at 2 of 3 engines.
Result: 5 of 8 category G points, six days in. From here, monthly monitoring asks the same three questions, word for word, every month. That trend line is what our monitoring service gives you.
Does AI cite our domain?
Pass (3/3). All three engines cite appsynth.pl and summarise the services correctly.
Does AI know the company?
Pass (2/3). ChatGPT and Gemini describe the business accurately, down to the methodology. Perplexity still confuses us with companies that have similar names.
Does AI recommend us in our category?
Not yet (1/3). Asked to recommend agencies for AI visibility in Norway, ChatGPT and Perplexity list competitors. Gemini already puts AppSynth first. That's below our own 2-of-3 bar, so the check stays failed, but it went from 0 of 3 engines to 1 of 3 within a week.
Your site next
Run the same audit on your website.
There was nothing special about this rebuild. We ran the audit, worked through the fix list in order, then measured again with the same tests. Clients get the same process, and it starts with a free mini-audit of your website.