CrawlClick

Cloaking

Serving different content to a crawler than to a human visitor, usually to obtain ranking or placement that the real page would not earn.

Also called: content cloaking

Cloaking is showing a crawler something different from what a person sees. Search engines have treated it as a serious violation for decades, and AI engines inherit that stance.

Is inserting content for AI crawlers cloaking?

It depends entirely on whether the difference deceives. Serving a crawler a page stuffed with keywords that a reader never sees is cloaking. Serving a labelled sponsored passage that a reader would also see, and that is honestly marked, is advertising.

Where is the line in practice?

The workable test is whether a reader shown both versions would feel misled. Hidden text, invisible keywords, and claims presented to machines but withheld from people all fail that test. Labelled paid content that survives being read aloud does not.

What are the consequences of getting it wrong?

Delisting. An engine that concludes a domain is manipulating its answers can remove the domain from consideration entirely, which costs far more than any short-term placement gained.

How do search engines define it?

By intent and effect rather than by mechanics. Google’s spam policy describes cloaking as presenting different content to users and search engines in order to manipulate rankings and mislead users. Both halves matter: serving different bytes is not the offence, and neither is serving a machine something a person cannot see. The offence is doing it to deceive.

What is the standard example of the opposite?

A subscription paywall. Google states that a paywall is not cloaking where it can see the full content just as a subscriber can, and where the publisher follows the flexible sampling guidance — which in practice means declaring the arrangement in structured data rather than leaving it to be inferred. The mechanics are identical to the prohibited case; the declaration is the difference. Handling AI crawlers on paywalled content covers how that declaration is made and what no AI vendor has published an equivalent of.

Where does that leave inserted content?

On the right side of the line when it is labelled and on the wrong side when it is not. A passage marked as sponsored, in wording a reader would also see, is advertising: the reader knows, the engine knows, and nothing is being concealed from either. The same passage served only to crawlers, or served to readers with the label stripped, is the prohibited case regardless of how accurate the claim inside it is.

What is definitely not permitted?

Hidden text, invisible keywords, off-screen blocks, text coloured to match the background, and content injected purely because the user-agent looked like a bot. None of these have a disclosed version, because concealment is the entire mechanism. They are also the reason the whole category is regarded with suspicion, which is why a disclosed market has to be visibly different rather than merely differently intentioned.

What else should I read?