How it works
Evidence first. Recommendations second.
Most automated website audits work backwards: they decide what to sell you, then generate findings that justify it. This one runs the other way round, and the order is the whole point.
The pipeline
Each stage only ever passes forward what the stage before it could establish.
Website data
We fetch your homepage and up to three more pages a customer would naturally visit — contact, services, booking — plus your robots.txt and sitemap.xml. Only publicly accessible pages, exactly as any visitor or search engine would see them.
Structured analysis
Each page is read into structured facts: page titles and descriptions, heading structure, links and where they go, forms and the fields in them, phone and email links, images, scripts, structured data, and the third-party tools your site loads.
Verified observations
Facts are separated by how certain they are. Something present in your page source is verified. Something we looked for and did not find is "not detected". Something a website cannot show at all is flagged as needing investigation.
Opportunity detection
A fixed set of rules turns observations into findings. Each rule must be able to point at the specific evidence that triggered it. A rule that cannot cite evidence does not produce a finding.
AI writing
Only now does a language model get involved, and only to phrase findings for your business. It never sees your raw pages, so it cannot "notice" a problem no rule found, and it cannot add a finding, a number, or a claim. Its output is checked against the findings we sent it.
Report generation
The report is assembled and stored. Your copy does not change afterwards, even if we improve the rules next week — the report you paid for is the report you keep.
Reading your report
Three labels, three different levels of certainty.
This distinction is the most important thing in the report. Ignore it and you will over-read what we found.
We saw this directly. It is present in — or provably absent from — the HTML we fetched, and the report quotes the evidence.
Example: "No <form> element with contact
fields was present in the HTML of your homepage."
We looked and did not find it. That is not the same as it not existing — tools loaded after page load, or behind a tag manager, are invisible to any scanner.
Example: "No scheduling widget was detected. A booking tool loaded another way would not be visible to this scan."
A website cannot show this. What happens after an enquiry arrives, who answers the phone at seven in the evening, how a quote gets written.
Example: "The site publishes a phone number. What happens to an unanswered call is not observable from a website."
Limits
What this scan cannot do
Worth stating plainly, because an automated tool that claims no limits is telling you something about itself.
- It sees a handful of pages, not your whole site.
- It reads the HTML your server returns, so anything rendered later by JavaScript may be missed.
- It cannot measure Core Web Vitals. Where speed is mentioned, it is one request from one place, and the report says so.
- It cannot see inside your business: your volumes, your close rate, your team's workload, or what actually happens to an enquiry.
- It cannot tell you what a change is worth in money. Nobody can, without your numbers.
The first four are why the detailed report ends with a conversation: a person has to ask. The fifth is why no report we produce will ever contain a revenue guarantee.
Find your next AI opportunity
Start with a free scan. See where your business could benefit from better technology, automation, and AI.