How we built an AI website audit tool in a day and what it does for clients

Have you ever wondered how much your website is really saying about you? Not just the words on the page but the accessibility gaps, the broken links, the security headers nobody checked, the cookie banner that doesn't quite do what it claims, and the tone of voice that drifts a little further from your brand with every new page someone publishes.

We wondered the same thing. So on March 13, 2026, at our internal hackathon a team of four set out to answer it. By the end of the day they had a working prototype. Today that prototype is Comprend WEB: Website Audit and Benchmarking – the platform our specialists can use to scan a site, then read, interpret, verify and turn into expert recommendations for our clients.

This is the story of how it happened, what WEB does and why we think the way it was built matters as much as what the tool enables.

A working product before the day was over

Hackathons usually produce a slide deck and a promise. This one produced software that ran. The reason was AI. The team didn't spend the day writing a boilerplate and wiring up plumbing by hand. They worked alongside AI coding tools, describing what they wanted and reviewing what came back, which let four people move at the pace of a much larger group. Instead of a mock-up, they presented a prototype that actually crawled a website and returned real insights.

That's the part we keep coming back to. A small team, one day and a live demo at the end of it. It's a concrete example of something we talk about a lot at Comprend: AI certainly doesn't replace expertise, it compresses the distance between an idea and a working version of it. The judges agreed and this project won the hackathon.

From prototype to production

Winning a hackathon is one thing. Running something in a live workflow for clients is another.

A demo can cut corners. A service we put our name to can't. So after the hackathon we took the prototype and did the less glamorous work: hardening it, securing it and making it reliable enough to trust with real client websites.

We didn't ship everything at once. WEB is built around separate audit "profiles" and we move each one from experimental to production on its own timeline as it proves itself. The first profile reached its first production version in about one month. The rest follow as they're ready, which means the platform keeps getting more capable without us ever betting a client deliverable on untested code.

Seven expert lenses on one website

WEB looks at a website the way a team of specialists would, through seven distinct profiles. Each one focuses on a different question.

UX Competence Lead reviews accessibility, usability and overall experience, including WCAG accessibility checks on every page, plus performance and SEO signals.

Content Strategist analyses content quality, messaging, tone of voice, readability and structure.

Technical Developer runs a technical health check: page performance, Core Web Vitals, HTML validity, broken links and images and security hygiene developers can fix in code.

Security Analyst identifies vulnerabilities and risk exposure, from TLS configuration and security headers to exposed files and common attack vectors.

Load Tester measures how the site holds up under traffic.

Privacy Audit checks cookie and tracking compliance per page, including whether the "reject" button on a consent banner actually works.

Site Inventory counts pages, inventories linked documents and assets, and identifies the underlying platform, hosting and CMS.

Underneath each profile sits a set of established, industry-standard scanning tools. That's the engine. What makes WEB more than a wrapper around those tools is the layer on top: AI reads the raw, technical output, summarises it in plain language, explains what each finding means and recommends what to do about it. A report from WEB doesn't just tell a specialist that a website has 200 issues. It tells the specialist which ones matter, why, and what to fix first.

Where it gets interesting: the Content Strategist

The profile we've invested the most in is the Content Strategist, and it shows where this approach can go.

We can give WEB your entire brand platform – your key messages, your audiences, your Style Guide and tone of voice, your priority pages and the content rules you hold yourself to. WEB then reads every page on your site and scores it against that platform. It tells us which pages are on-brand and which have drifted, where the tone slips, and which audiences a page is actually written for versus the one it should serve.

For a large site that's the kind of review that would take a content team weeks of reading. WEB does the first pass across hundreds, or even thousands, of pages and hands the strategist a per-page action list to review and verify. The judgement stays human. The reading at scale doesn't have to be.

Built to grow with our clients

WEB is deliberately easy to extend. Adding a new scanner or a new capability is straightforward, which means when a client has a specific need, we can usually build for it and it won't fall out of scope.

One example we're working on now: the EU's new Green Claims Directive. Companies will need to make sure the environmental claims on their websites are actually substantiated and documented. We want WEB to scan for those claims and help us verify that what they say is backed up by evidence before a regulator asks them to.

It's worth being clear about how clients work with WEB. They don't log in and run it themselves, that's by design. We run the platform, our specialists review and verify, and then we engage in forward-leaning conversations. Our clients get specialist-guided insights and the recommendations without having to operate the tooling. We already have clients working with us in this workflow today.

What we take from it

WEB is a useful tool but the story we like most is how it got here. A real value-adding software, built by a small team in under a day. Because AI let them spend their time on the idea instead of the scaffolding, we could focus the saved bandwidth on making it production-ready through the unglamorous, careful work that turns a clever demo into something you can stand behind.

That's the combination we're betting on: specialists who know what good looks like, AI that helps them get there faster and the discipline to ship something we trust to enable the strategic conversations we have with our clients.

To find out what a WEB audit could reveal about your website, contact Tina Söderlund at [email protected]. For more on the technical side of how WEB was built, get in touch with Johan Yourstone at [email protected].