A WebSite Auditor alternative that publishes its own calibration

We ran our crawler against a real WebSite Auditor export and wrote down every disagreement, including the categories where we have nothing to offer. Here is the honest version of the comparison.

500 / 500

pages fetched, 0 failed

531

page WebSite Auditor export, 2026-05-18

1,376

issues in that export

33 / 33

categories, parser exact

A native crawl of gplcoffee.dev, 500 / 500 pages fetched, against a WebSite Auditor export of 2026-05-18, under the compatibility preset: 33 / 33 categories parser-exact on a 1,376-finding export. Fourteen categories matched outright, including robots, sitemap, www, http/https and custom 404 handling, and every remaining delta got one line explaining it. The full delta table is on the case study.

What WebSite Auditor does that we do not

Three of these came out of the calibration itself: they are categories in the export our catalog has no answer for.

It is a desktop application you license: The report we calibrated against arrived as a file exported from somebody's machine on 2026-05-18, which is also how our import works: you upload the Site Audit (details) export and we parse it. If you want the tool itself, its rank tracking and its link tools, that is a licence to buy and a machine to run it on, and CrawlSpec is not a replacement for either.
It validates HTML and CSS against W3C: Two of its categories, W3C HTML errors/warnings and W3C CSS errors/warnings, have no equivalent in our catalog and we do not fake one. If a validator pass is what you are shopping for, we are the wrong instrument.
It reports a mobile-friendly verdict: Its Mobile friendly category has no native equivalent here either. We crawl one user agent, identify ourselves on every request, and report what the HTML says; we do not render the page on a phone and grade it.

What we do that a desktop report cannot

Each one is a property of where the audit lives, not a feature bolted onto a document.

The audit is state, not a document: A finding is a row that remembers the crawl that first saw it and the crawl that last saw it, so "this started three weeks ago" is a query rather than a diff of two PDFs. An export is a photograph of one afternoon.
Re-crawls run against the same project: Crawl again and the new run lands in the same history, on the same URL inventory, with run-over-run comparison built in. Nothing has to be re-uploaded, re-named or re-reconciled by hand.
A card you actually fixed stops coming back: Recommendation cards are rebuilt from the findings the NEWEST crawl re-observed, so a template you fixed produces no card on the next run. Four checks go further and close their underlying findings outright when the run re-probes the target and finds it healthy: resource 4xx, resource 5xx, broken link and broken image.
Everything is readable by an agent: The same audit is a Bearer-token JSON API and an MCP server over stdio, so a coding agent reads the fix, writes it in your repo, and the next crawl decides whether it worked. A desktop report has to be read by a person first.
Briefs come out of the same evidence: Search Console queries and the keyword map feed topic clusters, and a content brief is written from what the crawl already knows about the template. Marking a brief applied is the one write the API allows, and it closes the card for that brief.

The 4 categories we do not compute

Of every category in that export, these are the ones with no equivalent in our issue catalog. Our importer knows it: they sit on the parser's known-unmappable list, so a reconciliation reports them as absent rather than as a match.

Pages with 301 redirect
Mobile friendly
W3C HTML errors/warnings
W3C CSS errors/warnings

You do not have to choose on our word

Import your own WebSite Auditor Site Audit (details) export - the detailed report, not the summary - and we reconcile it against a live crawl. That is the same reconciliation we published on our own site: your export on one side, a live crawl on the other, every category lined up and every delta explained in one line.

Import your own export What a plan costs The 11 agent tools

Point it at your own site first

A free scan reads 25 pages with the same catalog and hands back 3 fixes with the prompt to act on them. No account, no card. What the free scan does