How we check websites: we load the same public URL as different visitors and compare what comes back. Below is exactly what we run.
Views
Five requests per check, one per view, from one datacenter location (Cloudflare).
| view | what we vary |
|---|---|
| Desktop Chrome | user agent (the baseline) |
| iPhone Safari | user agent |
| Googlebot | user agent (Googlebot Desktop string) |
| Facebook crawler | user agent (facebookexternalhit/1.1) |
| Ad click | referrer https://www.google.com/ and gclid, utm_source=google, utm_medium=cpc on the URL |
Same Accept and Accept-Language on every view. Exact strings: website cloaking checker. JavaScript is not run; script signals are read from the page source. Country and residential views, and a real browser, are a deep scan, by contact.
Compare
We compare each view against Desktop Chrome: final URL and redirect chain, status, title, visible text (3-word overlap), link domains, form and checkout, size, and script signals. A difference is a lead, not a verdict. Pages change for normal reasons (language, device, tests). What counts as cloaking is in the rules: website cloaking.
Evidence
Results are shown to you and not stored. We keep counters only: a hashed visitor id, the site name and the day, to enforce limits. Re-run the check, or reproduce it by hand with the curl guide below.
Limits
- A user agent is a label. A site that checks the visitor’s IP can tell our Googlebot label from the real one: IP-based cloaking.
- One IP type, one place, per check. 8 seconds and 2 MB per view. 5 checks per visitor per day, 6 per site per hour.
- No JavaScript run: a swap after load shows only as a script signal.
- Public pages only.
What we never do
- No logins. We do not sign in, create accounts or fill forms.
- No personal data. Checks read public web pages, not people.
Reproduce any check by hand: curl guide. Ask us to drop a page: removal request.