--- title: Website cloaking, switch by switch · Serpward url: https://serpward.com/journal/website-cloaking/ description: The signals a site uses to serve one page to a crawler and another to a visitor: user agent, referrer, network, country, visit — and how each is caught. date: 2026-09-11 source: Serpward (serpward.com) --- Method · 11 Sep 2026 · 4 min read # Website cloaking, switch by switch Every form of cloaking comes down to one decision made on the server: which visitor is this. Five things it can look at, and what each one takes to detect. Descriptions of website cloaking tend to stop at the idea: one page for the search engine, another for the visitor. The idea is easy. What decides whether a particular case can be caught — and by whom — is the switch the server is actually using. There are five in common use. They differ enormously in how hard they are to see from outside. ## The user agent Every request states what software is making it. Serving a different page to anything calling itself Googlebot is the oldest form of cloaking and the one every guide describes. It is also the one that survives least. Anyone can set their browser to announce the crawler's name and reload the page, and the difference appears immediately. A cloaking operation that relies on the user agent alone is either old, careless, or aimed at something other than search. **To detect:** change the user agent in the browser's developer tools and compare. Costs nothing, proves little when the result is clean. ## The referrer The request also carries where the visitor came from. A page can behave one way for someone arriving from a search result and another for someone who typed the address, and the second case is exactly how most people check a suspicious page. This is the switch behind the most frustrating pattern in enforcement: the reviewer opens the link from the notice, arrives with no referrer, and is shown a blank page. The visitor who arrived from the search result an hour earlier was sent to an offer in another language. **To detect:** request the page twice, once with a search result as the referrer and once without. No special access required — only the discipline to do both. ## The network the request comes from This is where cloaking becomes serious. The server looks at the address making the request, and decides by network: addresses belonging to search engines get one page, everyone else gets another. Pretending to be the crawler is useless here, because the name is not what is being checked. A request from an ordinary connection is served the visitor version no matter what it calls itself. **To detect:** the request has to come from somewhere else. Google's own testing tools fetch a page from Google's infrastructure, and a script running on Google's cloud does the same. Neither is the real crawler, but both come from the right address space — and that is what the switch is looking at. The limit is worth stating plainly: content served only to the crawler's own addresses cannot be requested by anybody outside the search engine. It leaves a trace rather than proof — the search result quotes text that is nowhere on the page you are shown. ## The country The same address requested from two countries returning two pages is not necessarily deception: local languages, local prices and local availability all look like this. It becomes cloaking when the choice of country is the product. A page that serves the material to visitors from the markets worth monetising and a harmless page to everyone else is cloaking by geography, and it defeats checks made from the wrong country — including, routinely, the country the person checking happens to be sitting in. **To detect:** request from the country the page is aimed at. This is the one switch where the checker's own location is part of the answer, which is why a check that always looks from a single country can report a clean page honestly and still be wrong. ## The visit A cookie set on the first request is enough to make the page behave differently on the second. The material appears once, to a first-time visitor, and never again — so the person who opens the link to confirm a colleague's report sees nothing, and concludes the report was mistaken. **To detect:** request the page with a clean session, and again with the session that has already visited. Any comparison that reuses one browser state misses this entirely. ## Rendering, which is usually not cloaking at all The last case is the one that produces most false accusations. A page assembled by JavaScript can legitimately look different to a crawler that does not run it or is blocked from loading what it needs. The difference is real and the intent is absent. Telling the two apart is a question of direction: a rendering fault usually leaves the crawler with *less* — a skeleton, a missing section, an empty container. Cloaking leaves the crawler with *more*, and specifically with more of what the ranking depends on. ## Reading a comparison Two versions side by side are not yet a verdict, and the mistakes are predictable. A refusal is not a substitution: if one request is answered with a block page and another with a not-found page, nobody was shown the page at all, and the difference is between two refusals. The strong signal is asymmetric — one side shown a page, the other shown a wall. Text that differs is worth less than it looks until you know how much the page differs from itself between two identical visits; plenty of pages rotate their contents on their own. And a different final address for one of the visitors outweighs any amount of differing text, because it is the whole mechanism in a single observation. A [cloaking checker](https://serpward.com/check/) that makes those requests and compares the answers is a starting point. What it can and cannot reach — and why the crawler's own addresses are out of everyone's reach — is set out in [how to view a page as Googlebot](https://serpward.com/view-page-as-googlebot/). Read next [ Study ### How often cloaking actually happens ](https://serpward.com/journal/how-often-cloaking-happens/) [ Method ### What is cloaking in SEO ](https://serpward.com/journal/what-is-cloaking-in-seo/)