Cloaking means presenting one version of a page to search engine crawlers and a different version to human visitors, typically to rank for content users never see. Google's spam policies prohibit it. Common forms include serving keyword-rich text to Googlebot while users get a different page, or showing crawlers a page about one topic and redirecting users elsewhere.
What counts as cloaking
The test is intent and mismatch: does the crawler see content the user would not, in a way that misleads?
- Serving a page about one topic to crawlers and another to users
- Inserting text or keywords only when the requester is a search engine
- Redirecting users, but not crawlers, to a different destination
What is not cloaking
Adapting a page for device type, language or location is fine when crawlers and users with the same characteristics see the same content. Paywalled content is allowed when it is marked up with Google's paywalled content structured data. Dynamic rendering, where crawlers get a pre-rendered version of the same JavaScript page, was accepted as long as the content matched, though Google now describes it as a workaround rather than a recommended approach.
Accidental cloaking
Sites sometimes cloak without meaning to. Bot detection that blocks or serves an error page to Googlebot, A/B testing tools that always serve crawlers one variant, or a hacked site that injects spam only for crawlers can all create mismatches. URL Inspection in Search Console shows what Google actually received, which is the fastest way to check.
Example
A security rule that hid the page from Google
Hypothetical vendor Brackenhull added aggressive bot protection, which served Googlebot a challenge page while users saw the normal site. It was not intentional cloaking, but Google was indexing the challenge page. Allowing verified Googlebot through and checking with URL Inspection fixed the mismatch.
Common mistakes
Long-running A/B tests that treat crawlers differently
Google's testing guidance says not to cloak during tests: crawlers should be treated like any other visitor, and tests should run only as long as necessary.
Questions
No, as long as crawlers are treated like users with the same attributes and the core content is not deceptively different.
