Methodology
How the Laudia Score works
Your Laudia Score is one clear number for how easily AI assistants can find and recommend your brand. Here is what each part of the report means in plain terms, and why it matters for getting named when buyers ask.
Safe by design. Any change Laudia recommends is labeled and applied only with your approval, and we never add crawler-blocking rules or medical claims, so improving your Score never puts your existing search rankings at risk.
Last updated July 26, 2026
What you receive
Every audit ends in a one-page report: your number and level, the technical signals with what we found, and how AI answered when we asked about you, with the evidence behind each line. Here is a full sample for a fictional client.
Your readiness, as one number
The headline number is how ready your site is for AI search: can the assistants reach it, understand what you offer, find your pages, and read a plain summary of your brand. We grade how well, not just yes or no. The same site always gives the same Score, so your progress stays comparable over time. The higher it is, the more easily engines can read you and put you in their answers.
Four signals make up the number, weighted by how much each one shapes whether an assistant can actually use you: whether AI crawlers can reach your site, whether your structured data describes you clearly, whether your pages are discoverable through a sitemap, and whether you publish a plain-text summary for language models. Out of 100 points, crawler access counts 35, structured data 35, sitemap discoverability 20, and the emerging plain-text convention 10. We also separate the crawlers that decide whether you can be quoted in an answer from the ones that only gather training data: blocking the first can quietly keep you out of AI answers and counts for more, while blocking the second is a legitimate choice and counts for less. A missing robots file is treated as fully open, never penalized.
The Score dial and level
Where you stand, in four levels
Your number lands in one of four levels: Not yet recognized, Emerging, Recommended, or Cum laude. You see at a glance whether AI already puts you forward or overlooks you. A low level is where you start, with a clear path up and exactly where to focus first.
The four levels
How we handle what we can't read
If a site blocks our reader or an engine is briefly unavailable, we mark that signal as not measured and leave it out, instead of guessing. Your number reflects only what we could actually check.
If we could read fewer than two of the four signals, we publish no number at all and say the result is not scored, rather than a confident headline built on a single signal. When we could read most of your site but not all of it, we mark the result provisional and cap the level we show, so a partial read never flatters you.
The technical signals
What AI says when buyers ask
Separately from the number, we ask ChatGPT, Claude, Perplexity, and Gemini five real buyer questions, the kind your customers actually type, and show how often you come up, with the answer we got as proof. You see exactly where you are recommended, and where a competitor is named instead of you.
We compare you within your real market, not against the whole world. A neighbourhood studio is measured against other options in its city, a national online brand against national peers, and only a genuinely global brand against global names, so a small Swiss brand is not marked down for failing to outrank a giant it never competes with. And when we show who was recommended instead of you, we only name a competitor that at least two different assistants named independently. One model's offhand list is noise, and we never invent a rival.
The five questions and who got named
A number you can trust over time
Every report carries the version of the method behind it, so your numbers stay comparable as you improve, and every finding comes with the evidence we based it on. You can prove progress month over month and stand behind every result with your team or a customer.
Each engine is queried at a pinned model version, recorded alongside the result, so a change in an underlying model is a deliberate, dated update on our side rather than silent drift in your numbers.
The version line
