What a website audit finds, 2,000 pages of VisitDenmark

18 July 2026

It's one thing to describe what a full-site audit does. It's another to watch one land. So we pointed a Webperf Audit at a real, public, well-run site, visitdenmark.com, let it crawl 2,000 pages, and read the results the way you would read your own.

VisitDenmark is, on the whole, a fast and carefully built site. That's rather the point of using it. An audit isn't a search for a disaster. It's a map, and the interesting thing about a good site's map is how a long list of findings turns out to be a short list of causes.

The headline came back as four numbers, measured at the 75th percentile across every page: performance 69, accessibility 94, SEO 92, best practice 100. Underneath them sat 35 distinct issue types and just over 37,000 individual error instances. That last number sounds alarming until you see where it comes from.

The lab said 69; the real users said green

The most useful thing in this whole audit is a disagreement.

Lighthouse, testing each page under controlled lab conditions, rated performance a middling 69. But the CrUX field data, measured from actual Chrome users across the whole origin, told a calmer story. Largest Contentful Paint 1.1 seconds. Cumulative Layout Shift 0.000, which is as clean as that metric goes. Interaction to Next Paint 184 milliseconds, inside the good range.

This is exactly why an audit shows you both, side by side. Lab data tells you where the technical headroom is; field data tells you what your visitors actually feel today. VisitDenmark's visitors are having a genuinely fast time of it. The lab findings aren't an emergency, they're the instructions for keeping it that way as the site grows. A team that only read the 69, or only read the green field data, would draw the wrong conclusion. Together they say: fast now, and here's how to protect it.

A big number that's really a small one

Those 37,000 error instances feel like 37,000 problems. They aren't. VisitDenmark runs on a component framework, so the same button, the same heading, the same script ships on thousands of pages at once. Count problems and you drown. Count causes and you get a to-do list that fits on a napkin. This is the pages-affected lens doing its work.

The heaviest performance findings each land on all 2,000 pages, because they live in one shared bundle:

• Unused JavaScript, the single biggest opportunity. One chunk shipped 197 KiB of code the page never runs, 56% of the file. Another wasted 95 KiB. Google Tag Manager alone added 81 KiB of dead weight.
• Legacy JavaScript, polyfills and transpilation sent to modern browsers that stopped needing them years ago.
• Unused CSS and render-blocking requests, styles loaded up front that the page doesn't use yet.
• Image delivery, flagged on 1,945 pages, where images could be compressed, resized, or served as AVIF or WebP.

None of that is a rebuild. Split the bundle, drop the legacy polyfills, tighten the images, and the score lifts on thousands of pages from a handful of changes at the source.

Accessibility: a strong 94, with one fix worth doing first

A 94 is a good score, and the findings are precise, each tied to WCAG, EN 301 549 and Section 508 so the work maps to the standards a public-sector site is held to, not just a number:

• Critical: the mobile search button has no accessible name. It ships on every page with no text and no aria-label, so a screen reader can only announce button. One component, 2,000 pages, a few minutes to fix.
• Serious: 1,103 pages have image-carousel links that are icon-only, with no readable text for a screen reader to speak.
• Moderate: headings skip from h2 to h4, breaking the outline that assistive tech relies on to navigate.

The pattern repeats: the fix is a single component, and clearing it clears every page it appears on.

SEO: mostly a matter of link text

SEO scored 92, and it's blunt about why. On 1,998 pages, links read read more or click here, which tells neither a search engine nor a screen-reader user where the link goes. A handful more pages have links a crawler can't follow, and two robots.txt syntax errors. Descriptive link text is one template change with an outsized return, better for search and accessibility in the same edit.

Privacy: consent works, headers don't

The data-protection check runs in a real headless browser, not a header sniff, and it had good news first. Not a single page set a cookie before consent. Getting that right is the hard part, and VisitDenmark has.

The rest is a tidy list. Every page loads third parties, seven domains in all, and Google Tag Manager is flagged as a known tracker that ought to be consent-gated before it loads rather than on arrival. And the security headers are missing site-wide: no Content-Security-Policy, no HSTS, no X-Content-Type-Options, no X-Frame-Options, no Permissions-Policy, and a weak Referrer-Policy. Those are server-configuration changes, set once and applied to the whole site.

Best practice scored a perfect 100 and still surfaced something useful, JavaScript console errors on 28 pages, including a batch of broken image requests to an external host. The quiet kind of breakage you'd never spot without testing every page.

What you can expect from your own

Strip away the specifics and this is the shape of every audit you'd order:

• Every page tested, not a sample, across performance, accessibility, SEO and best practice.
• Real data-protection testing in a headless browser, cookies, consent, third parties, trackers, TLS and headers.
• Real-user field data next to the lab scores, so you can tell a test-bench problem from a visitor one.
• Findings sorted by severity and reach, with the shared source named, so you fix once instead of page by page.
• Standards mapping for accessibility, and dashboards plus raw JSON you keep and own.

VisitDenmark came out of this looking like what it is, a good site with a short, clear list of high-leverage fixes. That's the honest promise of an audit. Not a verdict, but a map, with the few changes that move thousands of pages marked at the top.

We can run the same audit against your site and show you the map. See the plans, or email hello@webperf.cloud.