- Google evaluates consistency: what its systems observe versus what real users see, not one isolated signal.
- The compared dimensions include page content, destination domain, user-facing claims and the post-click experience.
- A persistent mismatch, not a one-off difference, is what draws policy scrutiny and account-level review.
- Legitimate variation — language, region, device, personalization — stays inside policy when the underlying offer is equivalent.
What Google actually compares
Google's ad review combines automated crawlers, classifier models and periodic human evaluation. Each of them forms a picture of your landing page, and the picture is then compared against the picture real users get. The comparison is not limited to visible text. It covers the destination domain and its ownership signals, the main content and its topic, the claims made in the ad versus the claims on the page, the products or services offered, and the post-click flow such as forms, downloads and checkout steps.
When the two pictures line up, nothing happens. When they diverge on something material — the topic, the offer, the business behind the page — the divergence itself becomes the signal. That is why searching for a single trick misses the point: the system is measuring agreement, and agreement is a property of your whole funnel rather than of one rule.
Where the policy language draws the line
Google's published policies group this area under misrepresentation and abuse of the ad network. The relevant idea is intent to present different content to Google's systems than to users in order to get an ad approved that would not otherwise be. Our article on the circumventing-systems policy covers the wording in detail; the practical reading is that what matters is whether the user-facing experience matches what was evaluated.
Notice what the line is not. It is not a ban on having different versions of a page. Localized pages, language variants, device-specific layouts and logged-in versus logged-out views are ordinary. The concern begins when the version shown to evaluators is materially different in offer or topic from the version shown to the audience the ad was bought to reach.
Signals that make a mismatch visible
Mismatches surface through several channels at once, which is why single-point workarounds age quickly. Crawler fetches from Google-owned infrastructure are compared with fetches from ordinary residential and mobile connections. Post-click behavior is compared with the page content: if users consistently land on something different from what was reviewed, engagement and complaint patterns diverge. User reports and advertiser-verification checks add a slower but more reliable layer.
Our detection overview explains how ad platforms combine these layers. The takeaway for a team is simple: assume every layer will eventually be observed, and design so that the observed layers agree. A setup that only holds up if nobody ever compares two views is fragile by construction.
Variation that stays inside policy
Plenty of legitimate practices show different visitors different pages. Region and language routing sends a visitor to the version in their language with local pricing. Device routing serves a lighter layout to mobile. Personalization changes the headline or featured product based on the ad the visitor clicked. In each case the underlying offer, the business identity and the topic stay the same, so the evaluated view and the user view agree on everything material.
A useful test is substitution: if a Google evaluator were shown the user's version, would they reach the same conclusion about the business and the offer? If yes, the variation is presentation. If no, it is a different page pretending to be the same one, and that is the case the policy is written for. Our comparison of routing and personalization boundaries goes deeper on where teams tend to cross over without noticing.
A self-check before an evaluation does it for you
Run the comparison yourself. Fetch the page the way a crawler would and the way a real mobile visitor would, from a few regions, and diff the results on topic, offer, business identity and post-click flow. Then read the ad copy side by side with the page. Any claim that appears in the ad but cannot be found by a neutral visitor is a mismatch waiting to be flagged, whatever routing sits in front of it.
Log the result, repeat it after every material change, and treat unexplained differences as defects to fix rather than quirks to tolerate. Our guide to checking whether a page is cloaking lists concrete steps, and the landing-page requirements article covers the baseline expectations Google publishes for destination pages.