Trustpilot removed 3.3 million fake reviews in 2024 per its own Transparency Report, yet the platform still receives more than 60 million new reviews a year and its ML fraud stack has documented blind spots on aged-account fraud, low-velocity coordinated attacks and language-mimicking fakes. That gap is why buyers, procurement teams and business owners need a manual forensic read — the automated filter catches volume attacks, not the quiet, patient fakes that actually hurt.
I'm Robiul, head of research at BGR Review. Our analysts file fake-review removal appeals on Trustpilot every week and the 8-signal checklist below is the same triage sheet we run before we open a case file. Two months ago we cleared 137 fabricated one-stars from a UK e-commerce profile using nothing more than this checklist plus the correct guideline citation.
Why fake Trustpilot reviews slip through in the first place
Trustpilot's automated moderation is genuinely strong against the obvious patterns: identical IPs, repeated wording, brand-new accounts posting on one profile, and traffic spikes above baseline velocity. The 2024 Transparency Report puts the platform's false-positive rate below 2% on flagged content — so when the model flags a review, it is right more than 98% of the time.
The problem is what the model doesn't flag. Aged accounts with real review history, reviews written in unique natural language, and fakes drip-fed at 2 to 3 per week from residential IPs all sit below the platform's automated thresholds. The UK CMA's 2023 investigation into fake reviews documented the same category-wide gap: platform ML catches volume, not craft. A human read of the profile catches craft in a way software still cannot.
The 8-signal forensic checklist
Run these eight signals on the individual review and the reviewer's history. Three or more matching is our internal threshold for opening a removal file. All eight are checkable in under two minutes on a public Trustpilot profile — no tools, no login, no paid access.
- Account age under 90 days. Click the reviewer's name and check when the account was created. A genuine reviewer's account usually predates the review being investigated by at least six months and often by years. Farmed accounts are typically created inside a 30 to 90-day window before the attack.
- Total review count of 1 to 3. Legitimate reviewers on Trustpilot average 8 to 20 reviews across unrelated brands over a multi-year history. A brand-new account with a single one-star on your profile and nothing else is the single most common fake pattern we see.
- Zero category diversity. A real reviewer writes about their airline, their broadband provider, their trainers and their electrician. A farmed account writes only about SaaS tools, or only about crypto brokers, or only about one competitive category. Total-brand list on one theme is a stronger signal than review count alone.
- Vague review body under 25 words. Genuine complaints name a product SKU, a staff first name, an order date, an amount or a specific incident. A fake reads like 'Terrible service, do not use, avoid at all costs' with no verifiable detail. Length alone is not a signal — some real one-stars are short — but length combined with zero specifics is.
- Timing cluster on the profile. Sort the profile's reviews by newest and look for 3 or more one-stars posted inside a 72-hour window that break the profile's normal cadence. A profile that averages 2 reviews a week receiving 8 one-stars in two days is a coordinated attack, not organic feedback.
- Impossible star-distribution shape. A profile whose distribution is 60% five-star, 5% four-star, 3% three-star, 2% two-star and 30% one-star is bimodal and mathematically unusual for a real business. Real profiles taper — 5-star is highest, 1-star second, and the middle bands are small but present. A missing middle is almost always planted one-stars, planted five-stars or both.
- Reviewer country flag mismatched to the market. A UK-only plumber receiving one-stars flagged from accounts in Bangladesh, Vietnam and the Philippines is not receiving reviews from customers — it is receiving farmed reviews from click-farm regions. Trustpilot displays the reviewer's country flag on every review; use it.
- Reviewer profile picture is a stock face or empty. AI-generated headshots (perfectly symmetrical face, unusual ear shape, blurred earring, generic corporate background) and blank-avatar defaults on brand-new accounts correlate strongly with farmed profiles. This is a supporting signal, not a decisive one — plenty of real reviewers use no avatar — but combined with signals 1 to 3 it becomes strong.
The single highest-leverage signal is category diversity in the reviewer's history. Fake reviewers get lazy about faking a plausible life outside the campaign they were paid for — and Trustpilot lets anyone see that history in two clicks.
Three real fake-review patterns we see every week
Detection is easier once you know the operator's playbook. From removal cases filed for BGR clients in the last twelve months, three patterns account for roughly 80% of confirmed fakes.
Pattern 1 — the aged-account fleet
Someone bought or built a fleet of 30 to 60 Trustpilot accounts more than two years ago and slowly grew each one with real reviews on real brands. The fleet is then rented per attack — 5 to 8 one-stars per profile per week, spread across the fleet, on whichever target the buyer specifies. Signals: accounts pass age and count checks, but category diversity is thin and the timing cluster on the target profile is the giveaway.
Pattern 2 — the offshore burst
20 to 100 brand-new accounts created inside a week and pointed at one profile with short vague one-stars. Country flags cluster in known click-farm regions. This one Trustpilot's automated filter usually catches — but usually is not always, especially for profiles under 200 total reviews where the baseline is too thin for the model.
Pattern 3 — the personal-grudge fabrication
A single real account — ex-employee, competitor, personal dispute — posts one detailed fabricated review. Signals 1 to 3 will not catch this because the account is real. The catches are internal: the incident described did not happen (no order, no customer record, no staff match), the tone is disproportionate, and the reviewer refuses any offline resolution. This is the slowest to remove but the most legally supportable via Guideline 4.6 (harmful/defamatory content).
How to report a fake Trustpilot review and actually get it removed
Reporting a fake review is not the same as getting it removed. Trustpilot rejects roughly half of business-flagged reviews on first submission — almost always because the flag cited the wrong guideline or included no evidence. A well-filed appeal is granted more than 70% of the time on the first pass. The mechanics that actually matter:
- Cite the exact guideline breached. Trustpilot removes under Guidelines 4.1 to 4.6: fake experience, off-topic, not about the goods or service, harmful/defamatory, breach of privacy, or promotional. Flagging without naming the specific guideline number is the most common reason for refusal.
- Attach evidence, not opinion. For a fake-experience claim, evidence is 'no order in this reviewer's name or email across our last 90 days of records' — not 'we don't think this is a real customer'. Screenshots of empty CRM searches are stronger than paragraphs of narrative.
- Flag through the Business Portal, not by replying to the review. Public replies do not initiate the removal workflow. Log in to business.trustpilot.com, open the review, click 'Flag' and complete the guideline-citation form.
- Give the integrity team 5 to 7 business days before escalating. Chasing on day two closes the file. If the initial flag is refused, you have one appeal — do not waste it with the same submission you filed the first time.
What buyers should do with all this
If you are researching a brand before you buy, do not stop at the TrustScore. Run the 90-second five-signal read for any Trustpilot profile from our authenticity guide on the profile itself, then apply the 8-signal checklist above to any specific review that swayed you. And when the profile shows obvious planted five-stars alongside obvious planted one-stars, the honest read is that you cannot trust the average in either direction — go to a category-appropriate second source (G2 for B2B software, Google Business Profile for local services) for a triangulated read.
For business owners, fake-review defence is a workflow, not a one-off. Weekly monitoring of new reviews, monthly audit of reviewer histories on any negative that meaningfully moves the average, and a documented 24-hour response cadence on genuine negatives — the same 4-part response framework we cover for real negative Trustpilot reviews — together do more for the profile's trust weight than any removal campaign in isolation.
Q.How do you know if a Trustpilot review is fake?
Run eight signals together: reviewer account age under 90 days, total review count of 1 to 3, no category diversity across the reviewer's history, vague body under 25 words, a timing cluster of similar reviews within 72 hours on the target profile, an unusual bimodal star distribution, reviewer country mismatched to the business's market, and a stock or blank profile picture. Three or more matching indicates a fake with high confidence.
Q.Can Trustpilot detect fake reviews automatically?
Partly. Trustpilot's ML fraud stack removed 3.3 million reviews in 2024 for authenticity violations, with a false-positive rate below 2% per its Transparency Report. Automated detection is strong against IP clustering, template similarity, brand-new accounts and velocity spikes, but weaker against aged-account fraud, unique-language fakes and low-velocity coordinated attacks. Manual reporting is still required for the patterns the model misses.
Q.How do I report a fake Trustpilot review as a business?
Log in to business.trustpilot.com, open the review and click Flag. Cite the specific Trustpilot Guideline breached — usually 4.1 (fake experience), 4.4 (harmful or defamatory) or 4.5 (breach of privacy) — and attach evidence such as an empty CRM search showing no matching order. Do not report through a public reply, and do not chase inside 5 to 7 business days.
Q.How long does Trustpilot take to remove a fake review?
Straightforward guideline breaches with strong evidence are typically actioned within 5 to 7 business days. Contested cases — where the reviewer disputes the flag — can extend to 3 to 6 weeks while Trustpilot's integrity team requests further information from both sides. Well-filed appeals with a specific guideline citation and evidence are granted on the first pass more than 70% of the time.
Q.Can I sue someone for leaving a fake Trustpilot review?
In principle yes, in practice rarely. UK defamation law and US state defamation statutes both cover fabricated review content, but proving identity behind an anonymous account usually requires a subpoena to Trustpilot and then to the ISP, which is time-consuming and expensive. The pragmatic path is guideline-based removal via Trustpilot first; litigation is a last resort where the reviewer is identifiable and the reputational damage is material.
Q.Are all short one-star Trustpilot reviews fake?
No. Plenty of genuinely unhappy customers write short angry one-stars with little detail. Length alone is not a fake signal — it becomes one only when combined with reviewer signals such as account age under 90 days, review count of 1 to 3, and zero category diversity. Treat the review body as one of eight signals, never as the deciding one.
The honest bottom line
Fake Trustpilot reviews are real, they slip past automated moderation more often than the platform admits publicly, and they are removable — but only when the person reporting them knows the eight forensic signals, cites the correct guideline and attaches real evidence. Two minutes per suspicious review, one well-filed flag per confirmed fake, and a monthly monitoring habit is the entire defensive playbook. Everything more elaborate than that is usually someone selling you a tool you do not need.
