Why not just ask an AI?
Fair question. You can paste your URL into an AI assistant and get an answer in seconds, at no cost.
We use AI too. It reads our measurements, interprets them in context, and writes your remediation guide. So this isn't a page about AI being wrong.
It's about what has to happen before AI is useful.
Run the test yourself
Don't take our word for it.
- Ask an AI assistant to analyse your site and name the three things you should fix first.
- Open a new conversation. Ask for exactly the same thing.
- Compare the two lists.
Then run SiteMerit twice.
Both AI answers may be perfectly reasonable. That's the problem. You can't tell which one to act on, and neither can the person who runs it after you. A method you can't repeat isn't a baseline — it's an opinion with good grammar.
Five differences that matter
| AI assistant | SiteMerit | |
|---|---|---|
| Method | Chosen fresh on every run — it may measure, it may estimate, and you rarely know which | Fixed and versioned: the same checks, the same weights, every time |
| Repeatability | Varies between runs and between phrasings | Same input, same result |
| Comparability | Two runs used different methods, so the gap between them means nothing | Two reports share one scale while the method version is the same |
| The output | A conversation that ends with the tab | A permanent, dated, shareable report |
| Accountability | A disclaimer | A company standing behind the number |
Where AI sits in our engine
The free sample runs the same fixed, versioned checks on the page delivered by the server. In the paid audit, we load the page as a browser would, and observe what it executes. The two layers are reported separately, and page execution does not change the score.
Then the semantic layer reads that evidence and interprets it — what matters here, what's noise, what to fix first. Your remediation guide is written by AI, on top of the measurements.
The order is the entire point. Interpretation without measurement is a conversation. Interpretation grounded in measurement is an audit.
What we don't claim
We think you learn more about an audit tool from what it refuses to promise.
- We don't certify compliance. Our accessibility checks are aligned with WCAG 2.1 and 2.2, and they cover a small part of it. That is a signal, not a legal verdict.
- We don't guarantee rankings. Nobody can.
- We don't crawl your entire site. We analyse the page you submit.
- We don't tune weights to produce a flattering score.
- We don't call the product free. Your sample is free; the full report is paid, and you see the price before you buy.
If a tool won't tell you where it stops, the number it gives you is worth less.
See it on your own site
Run a free sample. If the evidence is worth having, the full report is one click away — no account, no subscription.
Start audit