Review Verification, Moderation & Appeals Policy

_Last updated: 3 August 2026_

This policy explains how content moves through TrustRating: how reviews are verified and screened, how our TrustGuard system and human moderators work, what enforcement actions exist, and how both reviewers and businesses can dispute decisions. It is part of the Terms of Service and implements the commitments in the Reviewer Guidelines, the Acceptable Use Policy and the Business Terms of Service. Our aim is a process that is fast for honest content, hard for fraud, and fair to everyone in between — and one you can read in full, because a moderation system that works only while secret is not working.

1. The review pipeline

Every submitted review follows the same path, regardless of who wrote it and which company it concerns. Submission — the review is received with its rating, text, date of experience and, where it originates from a business invitation, the invitation token that verifies the transaction. Eligibility checks — the account must exist in good standing and have a verified email address; reviews from unverified or review-banned accounts are not published. Email verification is a deliberately low bar with a deliberately high yield: it costs an honest reviewer thirty seconds once, and it costs a thousand-account bot farm a thousand working mailboxes, which is why it is the pipeline's first gate rather than an afterthought. Automated screening — TrustGuard (described below) scores the review for fraud and policy risk. Clean reviews are published immediately; flagged ones are quarantined for human review before they can affect anything. Publication — the review appears on the company profile, enters the TrustScore calculation, and becomes visible on the reviewer's public profile. Life after publication — the review can be edited or deleted by its author, replied to by the business, flagged by anyone, and re-examined whenever new evidence arrives. Nothing about this pipeline can be bought: there is no priority lane, no paid cleanup, no plan that changes how moderation treats a company's reviews — see the Content Integrity, Scoring & IP Policy.

2. TrustGuard — automated screening

TrustGuard is our fraud-prevention layer. It evaluates signals across four groups: account signals (age, verification status, history, prior sanctions, unusual registration patterns); behavioural signals (submission velocity, coordination between accounts, device and network patterns consistent with bot activity or review farms); content signals (duplicated or templated text, machine-generated boilerplate presented as experience, language mismatches, spam and prohibited content); and context signals (bursts of reviews for one company suggesting bombing or astroturfing, correlations with invitation data, relationships between reviewer and company suggesting self-reviews or competitor attacks). Each review receives a risk assessment. Low-risk reviews publish instantly. High-risk reviews are quarantined into a moderation queue — they are not shown publicly and do not touch the TrustScore until a human decides. Companies experiencing a coordinated attack are protected in the same way: an anomalous flood of one-star reviews is held for investigation rather than instantly destroying a score, exactly as a flood of five-star praise is.

TrustGuard is automation with a human backstop: it can quarantine and prioritise, but consequential decisions — removals with account sanctions, suspensions, bans — are made or confirmed by human moderators, and every one of them is appealable (Section 6). The system is tuned to the asymmetry of its errors: wrongly quarantining an honest review delays it briefly and a human releases it; wrongly publishing a fraud campaign poisons a score that thousands of buyers rely on. When signals are ambiguous, holding for a human look is the designed behaviour, not overreach — and reviewers whose honest review was briefly held lose nothing but a short delay, since publication timing has no effect on the review's weight.

Settled cases stay settled. Once a review or account has been through moderation and a decision has been made, the same evidence does not re-trigger the same case — re-examination happens when new evidence or a new report with new substance arrives. This is what keeps the queues meaningful for moderators and the outcomes stable for everyone else.

3. Flags and reports

Any user can flag a published review, with a reason, directly on the page. A business can report reviews of its own profile from the business panel, with evidence. Flags from multiple independent users automatically escalate a review into the moderation queue. Moderators see the flag reasons, the reviewer's history, the invitation status and the TrustGuard assessment together, and decide with the full picture. Flagging is confidential — the review's author sees that their review is under review, not who flagged it. Abusive flagging (coordinated campaigns against honest reviews, retaliatory flags) is itself sanctioned under the Acceptable Use Policy.

For businesses, the correct route to challenge a review is the report function plus, where relevant, a request to the reviewer for proof of experience. A documented request that goes unanswered for a reasonable period allows removal of the review under Section 2 of the Reviewer Guidelines. Legal threats against reviewers sent through our platform, and off-platform pressure to delete truthful reviews, violate the Business Terms of Service.

4. Human moderation

Moderators are TrustRating staff bound by confidentiality and conflict-of-interest rules: no moderator handles a case involving a company they have a personal or financial relationship with, and — critically — moderation sits organisationally apart from sales: nobody whose job is revenue has moderation tools, and nobody moderating a case knows or cares what plan the company is on, because the tooling does not surface it as a decision factor. Moderators work from documented internal criteria that mirror the published policies — this document, the Guidelines and the AUP — so that the rules we enforce are the rules you can read. Decisions record which rule was applied. Where a case is ambiguous (a heated but arguably genuine review, a borderline privacy disclosure), moderators err on the side of keeping honest speech online, editing nothing, and using the least severe adequate measure — asking for evidence or redacting a specific detail rather than removing an entire review.

5. Enforcement actions

Depending on severity and history, we may: decline to publish a review that fails screening; hide a review pending investigation (it drops out of public view and the score temporarily); remove a review that violates the rules — removal recalculates the company's TrustScore automatically; redact specific unlawful details (for example a private phone number) where the rest of the review is legitimate, marking the edit; label accounts for transparency (for example "review-banned" or "suspended" states visible where relevant); restrict features — a review ban stops publishing new reviews while leaving the account otherwise intact; suspend an account temporarily; terminate an account and its linked identities for grave or repeated violations; and for businesses, apply the additional consequences described in the Business Terms of Service, up to revoking claims and subscriptions. Fraud rings and extortion are additionally preserved as evidence and may be reported to law enforcement.

We do not punish honest mistakes the way we punish fraud. A first-time borderline review typically results in education or a request to edit; a purchased review campaign results in every fake review being removed and every involved account terminated.

6. Appeals

Every substantive decision — a removed or hidden review, a declined publication, a review ban, a suspension, a rejected business report — can be appealed by the affected party. Appeals are submitted through the contact page or the notification you received, within 6 months of the decision. A useful appeal says why the decision was wrong: what rule was misapplied, what evidence was missed, what context changes the picture — and attaches the proof. "I disagree" restates the situation; "the order number attached shows the transaction the review describes" reverses decisions. An appeal is reviewed by a person not involved in the original decision, with authority to overturn it. You will receive a reasoned outcome; where an appeal succeeds we restore the content or account and correct any affected TrustScore retroactively — and where an appeal reveals that our rule or its application is systematically wrong, we fix the rule, not just the case. If you have exhausted the appeal and remain unsatisfied, you retain any rights available under the law of your country, including — for EU users under the Digital Services Act — the right to choose a certified out-of-court dispute-settlement body and the right to complain to your Digital Services Coordinator.

7. Scores during moderation, and what records we keep

Moderation states map to score effects mechanically, so it is worth stating them plainly. A review pending screening has no effect on the TrustScore — it does not exist publicly yet. A quarantined or hidden review is removed from the calculation for as long as it is out of public view; if it is reinstated, its effect returns as if it had never left. A removed review leaves the calculation permanently, and the score recalculates at once. A restored review (after a successful appeal) re-enters both the page and the calculation retroactively. The symmetry is deliberate: no moderation state exists in which a review counts against a company while being invisible, or is visible while not counting. Businesses watching their score move during an investigation are seeing this mechanism, not a penalty.

We keep records of moderation: the decision, the rule applied, the evidence considered, the moderator's reasoning for non-obvious calls, and the appeal history. These records exist so that decisions are consistent across time and moderators, so appeals can actually re-examine something, and so patterns (a business filing dozens of meritless reports, an account repeatedly skirting the same rule) are visible as patterns. Retention of these records is described in the Privacy Policy; they are internal, disclosed only through your own decision notices, appeals, and where the law compels more.

8. Timelines

Speed matters differently in different directions, and our targets reflect that: publication screening is near-instant for the overwhelming majority of reviews — honest reviews should never languish in a queue; quarantine decisions (whether a TrustGuard-held review publishes or not) are made promptly, since both the reviewer and the company deserve an answer; flag-escalated reviews are handled in the order and urgency their category warrants — a privacy disclosure outranks a tone complaint; evidence requests to reviewers state their response window, long enough to be fair to people who do not check email daily; and appeals are answered as quickly as genuine re-examination allows — an appeal that gets a same-minute template answer was not re-examined, and we would rather be a day slower and actually look. Where volumes spike (a viral company, a coordinated attack), triage protects the same priorities: unlawful content first, active manipulation second, everything else in turn.

9. Government and legal orders

Where a court or competent authority orders removal of content, we comply to the extent the order is valid and binding, restrict the effect to the territory the order covers where possible, and — unless legally forbidden — notify the affected user and record the removal reason. A private legal threat is not a court order: a lawyer's letter demanding removal of a review receives the same evidence-based process as any other report, because if letterheads removed reviews, every dissatisfied company would buy stationery instead of improving service. Reports of manifestly illegal content from any source are prioritised. Intellectual-property complaints follow the dedicated takedown procedure in the Content Integrity, Scoring & IP Policy.

10. Transparency

We publish how the score works (methodology), we label verified reviews, moderation states and sanctioned accounts, we tell affected users which rule was applied and how to appeal, and we aggregate what we learn about fraud patterns in public trust reports. What we do not publish are the exact detection thresholds — advertising them would be a manual for evading them.

11. Changes

We refine this policy as fraud evolves and as legal duties (including the DSA) develop. Material changes are announced as described in the Terms of Service, and the "Last updated" date above always identifies the current version.