Your whole website, checked in 20 seconds.
One URL, seven checks at once: AI search, legal basics, accessibility, SEO, AI agents, email, and link previews. Then Claude reads every finding and tells you what to fix first, and why.
Takes 10 to 30 seconds. Nothing is stored. The written analysis comes from the Claude API (Anthropic), from the findings only.
What this checks
AI search visibility
Whether ChatGPT, Perplexity, and Google’s AI can read, understand, and cite the page.
Legal basics
Google Fonts, Impressum, privacy policy, trackers before consent, stale legal texts.
Accessibility (BFSG)
A WCAG 2.1 AA scan in a real browser, the standard the BFSG asks for since June 2025.
SEO basics
Crawlability, title, description, headings, and internal links.
AI agents
Whether browser agents like ChatGPT agent can find, name, and click the controls.
Email trust
SPF, DKIM, and DMARC: whether anyone can send mail in your name.
Link previews
How the page looks when shared on LinkedIn, X, WhatsApp, and Slack.
What to fix first
Claude reads every finding and ranks the fixes by what they are worth to the business, with the effort.
One report, many fixes. Want them done?
We fix what the check finds: legal basics, accessibility, AI visibility, and the systems behind the site, built and handed over running.
Questions about the Website Check
What does the Website Check do?
It runs seven of our checks on one URL at the same time and puts them into one report: AI search visibility, German legal basics, WCAG accessibility in a real browser, SEO basics, AI agent readiness, email authentication, and link previews. Each area shows its score and the most important open findings, and links to the full tool.
What is a good score?
In October 2026 we ran this check on 144 small business websites from Germany, Austria, and Switzerland. The median scored 79, twelve reached 90 or more, none passed 94. Each area tile in the report shows the median of that study next to your score; the study itself is linked under From the blog.
Who writes the analysis at the top?
A language model (Claude from Anthropic) reads the findings of the seven checks, nothing else, and ranks what to fix first by what it is worth to the business: legal exposure and lost customers before cosmetics. It cannot add findings of its own; every point traces back to a check in the report. If the model is unavailable, a rule-based ranking takes its place.
Is anything stored?
No. The page is fetched for this run only. For the written analysis, the findings (not the page itself) go to the Claude API (Anthropic). Nothing is kept on kant.dev.
What does it not cover?
Page speed and what cookies do after consent each need up to a minute of browser time, so they have their own tools: the Website Speed Check and the Cookie Scanner. The report links to both.
From the blog
- German small business websites: 150 checked
We ran 150 small business websites through seven checks: 55% cite outdated law, 76% do not stop forged email, 89% fail automated accessibility tests.
