Can Google detect fake reviews?

Google's detection works less on what a review says and more on everything around it. The account that posted it: how old, how real its history looks, what else it has reviewed. The device and network: reviews for one business arriving from the same handful of devices or the business's own connection. The graph: clusters of accounts that review the same unrelated businesses in sequence, which is the fingerprint of a review farm. And velocity: a profile that gathers more reviews in a week than in its whole prior life.

Text plays a supporting role. Interchangeable superlatives with no job detail, the same phrasing recycled across businesses, and now the mass-produced cadence of AI-written batches all raise scores, but behavioural signals do the heavy lifting because they are far harder to fake at scale.

Detection is demonstrably imperfect, which is why the Competition and Markets Authority investigated Google and, in January 2025, took undertakings requiring stronger detection, sanctions against businesses and reviewers, and three years of reporting to the regulator. The practical consequence for UK businesses is that the filters have tightened and keep tightening.

The uncomfortable edge case is that the same velocity signals can catch honest businesses. Ask two hundred past customers in one enthusiastic weekend and the resulting spike looks, from the outside, like a purchase. Genuine review campaigns should drip over weeks precisely because the detector cannot see your good intentions, only your pattern.

The asymmetry worth remembering: a fake review only has to be caught once, including retrospectively in a later sweep, while a genuine review from a real customer survives every sweep forever. Detection risk is entirely a problem for one side of this.