Grader tools have become the front door to AEO: paste your domain, wait twenty seconds, get a score out of 100. They work as a quick diagnostic — but only if you understand what they measure and what they quietly skip.
What a well-built score measures
An honest grade combines two blocks that should never be blended blindly:
1. Technical readiness (what you control today). Whether AI crawlers can read and understand you:
- Valid structured data (Organization, Product, FAQ, Article).
- Content rendered in the served HTML, not only after JavaScript.
- Robots.txt and headers that do not block the AI crawlers you want.
- Direct answers above the fold, not three paragraphs of preamble.
- An llms.txt summarising your proposition and key pages.
2. Assistant awareness (what others say about you). Whether real category prompts name you, in what context, and against which competitors.
A grader that only inspects the technical block will hand you an 80 even if no assistant has ever mentioned you. That is the most expensive confusion in this market.
Why two graders give you different scores
- Different weighting: some put 70% on technical signals, others on mentions.
- Different prompts: if the set does not match your category, the score means nothing.
- Different models: not all of them query the same assistants, or with live browsing.
- Different moment: generative answers vary between runs.
Never compare scores across tools. Compare your own score against itself over time, using one methodology.
How to read your score honestly
- Below 50: there is a structural problem, almost always awareness. Assistants have no material to talk about you.
- Between 50 and 75: you show up in some prompts, usually branded or very specific ones. Category and comparison coverage is missing.
- Above 75: you appear consistently. The work turns defensive: sentiment, position, and not losing ground to competitors.
Always read the breakdown, never the headline number. A 62 hiding 90 technical and 20 awareness demands a completely different plan than a 62 the other way round.
From score to plan: four moves
- Fix the technical layer first if crawling is blocked or structured data is missing: it is cheap and unblocks everything else.
- Publish the missing page for the three category questions where you never appear.
- Get into the sources the AI cites: comparisons, directories, roundups and industry forums.
- Re-measure in four weeks with identical prompts and compare the breakdown, not the headline.
The limit of any grader
A score is a photograph, not a film. Its real value shows up when you repeat it: the trend tells you whether your content is working; the isolated grade only tells you where you stood on a Tuesday.
You can run an AI visibility audit of your domain with no signup and see the split between technical readiness and AI awareness. Then go deeper with the 30-day AEO audit.