A plumbing firm in Bristol lost three enquiries in a fortnight after a competitor posted five fake one-star reviews in 48 hours. The owner spotted the pattern eventually — but only after the damage was done. This is happening to UK small businesses every week, and most have no process to catch it early.
The problem has two sides: protecting yourself from fake negative reviews, and making sure you are not accidentally sitting next to fabricated five-star reviews that regulators are now actively pursuing. Both carry real risk. Both are now manageable with the right tools.
Coordinated review attacks have become more sophisticated because the platforms have got better at detecting the obvious ones. Reviews now come from aged accounts, vary in language, and stagger their posting times. Google and Trustpilot have improved their automated detection. Fraudsters have improved faster.
The UK Competition and Markets Authority fined several companies in 2024 for fake review manipulation, and enforcement is increasing. That means there is legal risk on both sides: being attacked, and unknowingly hosting or benefiting from fake positive reviews on your own profile.
AI does not replace your judgement here — it scales it. A trained model can analyse thousands of reviews in seconds and flag patterns that a human would miss after ten minutes of reading. The three most useful applications are anomaly detection (sudden spikes in volume, clusters posted within minutes of each other, reviewers with no prior activity), linguistic analysis (shared phrasing or syntax across supposedly independent authors — a tell-tale sign of templated content), and sentiment inconsistency (star ratings that contradict the written content, common in poorly executed attacks).
Tools like Trustpilot's own fraud detection, ReviewTrackers, and Podium offer some of this natively. Businesses dealing with more targeted attacks are running their review data through AI models with custom prompts that flag linguistic fingerprints across batches of reviews. It takes five minutes and gives you something concrete to take to Google's review removal process.
The earliest warning sign is velocity. Three or more reviews in a single day from accounts with no prior review history should be treated as suspicious until proven otherwise. Real unhappy customers describe specific interactions — a name, a date, a job that went wrong. Fake reviews are vague. When three reviews use similar vague language in the same week, that pattern is worth acting on immediately.
Fighting back effectively means giving platforms a case, not a complaint. Google removes reviews when you provide a clear, evidenced submission: confirmation the reviewer has no record as a customer, screenshots showing the velocity pattern with timestamps, linguistic analysis showing similarities across reviews, and any broader context about competitive activity. Vague reports go nowhere. Detailed ones succeed far more often. For Trustpilot or Checkatrade, a formal written complaint to their fraud team gets treated entirely differently to a clicked report button.
This cuts both ways. If you have ever used a service that generated reviews, or incentivised customers with discounts for five-star reviews, you are exposed. The CMA 2024 enforcement action named specific practices now considered deceptive, and small businesses are not exempt. A clean review profile with genuine variance — some fours, an occasional three, real specificity in the language — is also more persuasive to prospective customers than a suspiciously perfect average.
The businesses that handle this well are the ones who treat it as a process, not a crisis response. A monthly export of your review data, a basic velocity check, and a ready-made submission template for Google removal requests — built into a repeatable workflow — means you catch an attack before the damage compounds, not weeks after a competitor has flagged it to you.
—
If your business has been hit by suspicious reviews, or you want to build a monitoring process that catches problems before they escalate, we can help. Start at aias.co/contact.