How our scoring works
We score every website on three things — Technical Health, AI Readiness and AI Visibility. Here’s exactly what each one measures, how it’s calculated, and — just as important — what it doesn’t promise. No black box.
How does a scan produce three scores?
When you run a scan, we fetch your site and analyse it across several areas — technical setup, SEO tags, keyword gaps, AI readability and content — using a language model that reads the evidence from your pages against a fixed rubric. Two of the three scores come out of that analysis. The third — AI Visibility — comes from putting real questions to ChatGPT, Claude and Gemini and recording what they actually say.
Technical Health
Technical Health (0–100) measures whether search engines and AI crawlers can reach, read and index your pages — and whether the foundations are in place for you to be found and recommended at all. It’s judged against a fixed rubric, not opinion, and it draws on your site’s crawlability, robots.txt, sitemap, canonical tags, title and meta tags, structured data, and AI-readability signals.
Important: this is technical SEO and AI-readability health — not a Lighthouse-style measured page-speed score. A high number means your site is well set up to be read and indexed, not that it loads in a particular number of milliseconds.
| Band | What it means |
|---|---|
| 85–100 | No critical or high technical issues; title, description and structured data present and well-formed; strong AI-platform readiness. Minor gaps only. |
| 70–84 | No issue that blocks indexing; at most a couple of high-impact ones; core meta present; moderate gaps. |
| 55–69 | Several high-impact issues, or weak/missing core meta; substantial gaps; fair-to-poor AI readiness. |
| 40–54 | At least one indexing-blocking issue, or core meta largely missing. |
| 0–39 | Multiple critical issues. |
Run-to-run noise: because the score is judged by a language model, it can move by a few points between scans even when nothing on your site changed. We treat a move of 3 points or fewer as “no material change”, and moves of up to 6 have been seen. Read the trend, not a single scan.
AI Readiness
AI Readiness (0–100) measures how well your pages give an AI assistant what it needs to understand and quote you accurately. We rate your site separately for ChatGPT, Gemini and Claude — each as poor, fair, good or excellent — then average the three. So the score is a multiple of the mapping below, and 80 means “good across all three”.
| Rating | Maps to | What earns it |
|---|---|---|
| Poor | 40 | Content is invisible to crawlers, or there’s no meaningful text. |
| Fair | 60 | Readable but generic — no question-phrased content, no direct answers, thin structured data, weak entity clarity. |
| Good | 80 | Answer-first copy or FAQs present, valid relevant structured data, clear entity and brand info, descriptive media. |
| Excellent | 100 | All of “good”, across the site’s key pages — not just the homepage — with the depth a reviewer would call complete. |
This is on-page readiness — what your site gives an assistant to work with — not whether assistants actually recommend you. “Excellent” is deliberately hard to earn: the rubric asks for complete coverage across your key pages, and tells the reviewer to choose the lower level when in doubt. Most good, well-kept sites sit at “good”.
AI Visibility
AI Visibility is about what the assistants actually say — measured, not guessed. There are two related measures, and it’s worth being precise about which is which.
1. Can the assistants describe your business? (the dashboard score)
For each of ChatGPT, Claude and Gemini, we ask around ten questions that already contain your business name — “What is [your business]?”, “What does [your business] do?”, and a few industry-specific ones — and count how often the answer mentions you, mapping that onto 0–100. It measures how clearly the assistants can describe you once you’re named: your entity clarity. It does not measure whether an assistant recommends you to someone who didn’t name you.
2. Do the assistants name you to a buyer? (the mini scan & Monitor check)
This is the honest “does AI recommend you” test. We take one buyer-style question derived from your own homepage — with no brand name in it — and ask each assistant, both with live web access and cold. The verdict is decided by code, not the model grading itself: are you named, which other businesses were named, and what was cited. If an assistant can’t ground its answer in the live web, we record “unavailable” rather than pass off a from-memory answer as current. It’s what Monitor’s “first named” alert fires on.
The free mini scan grade
The mini scan also shows a quick letter-of-the-law grade — Leader (85+), Challenger (60–84) or At Risk (under 60) — from a simple, fixed checklist. It’s deliberately lighter than the full scan, and these bands are their own thing, separate from the three dashboard scores:
| Check | Weight |
|---|---|
| AI-crawler access | 25 |
| Structured data (JSON-LD) | 20 |
| Title tag | 15 |
| Meta description | 15 |
| Readable without JavaScript | 15 |
| H1 heading | 10 |
What the scores don’t promise
- No guaranteed rankings or recommendations — nobody can honestly promise those.
- A high AI Visibility (describe) score is not the same as being recommended to a new customer.
- Technical Health is not a measured page-speed score.
- Scores are not exactly reproducible — expect small run-to-run movement; we flag only changes that clear the noise floor.
- “Excellent” AI Readiness is intentionally hard to earn; “good” is a strong, healthy result.
- We never quote result numbers or timeframes we can’t stand behind.
Scoring — common questions
Do a few points’ change between scans mean something?
Not on their own. The two on-page scores are judged by a language model against a fixed rubric. We pin the sampling to be as consistent as possible, but run-to-run moves of a few points with no change to your site are normal. We treat a move of 3 points or less as “no material change”, and we’ve seen up to 6. Look at the direction over several scans, not a single number.
Is Technical Health a site-speed score?
No. It measures technical SEO and AI-readability health — crawlability, indexing, robots.txt, sitemaps, canonicals, meta tags, structured data and how well your pages are set up for AI to read. It is not a Lighthouse-style measured performance score, so don’t read it as “how fast my site loads”.
Does a high AI Visibility score mean AI recommends my business?
Not by itself — and this is the most important distinction. The dashboard AI Visibility score measures how clearly ChatGPT, Claude and Gemini can describe your business once it’s named. Whether an assistant recommends you to a buyer who didn’t name you is a separate, harder question — that’s what our buyer-question check (in the mini scan and Monitor) measures.
Can you guarantee I’ll rank or get recommended?
No, and you should be wary of anyone who does. Search engines and AI systems decide what to show, and being named for questions that don’t mention you depends on signals from across the web, not just your own pages. We measure honestly and tell you what to change; we don’t sell guarantees.
See your three scores
Run a free 30-second mini scan, or get the full three-score report and prioritised plan with a subscription.
New to the terms? The AI search glossary explains them in plain English.