Methodology

How the Rated Stores Fake Review Detector estimates review authenticity.

Two-stage hybrid scoring

Every review is scored on a 0-100 scale where 0 = clearly genuine and 100 = clearly fake. The final score is a weighted combination of two independent signals:

1. Heuristic pattern analysis (30% weight) — regex-based detection of linguistic patterns common in paid review farms: superlative bursts, generic praise phrases, ALL-CAPS, repeated punctuation, short text with extreme sentiment, missing specific product detail.
2. AI semantic analysis (70% weight) — Google Gemini model evaluates plausibility, scripted tone, AI-generation patterns, sentiment polarity without supporting detail, and personal-experience markers.

Verdict thresholds

• Likely genuine: score ≤ 30
• Uncertain: 31-64
• Likely fake: ≥ 65

URL analysis

When you paste a URL (Trustpilot, Google Maps, Yelp, TripAdvisor, Etsy, Judge.me, Amazon) we first check whether the business is already indexed in Rated Stores. If it is, we aggregate the heuristic score across the 50 most-recent reviews already in our database. If not, you'll see a guidance message — paste individual reviews for instant analysis.

This is a limited check. Full audits (sentiment trends, source breakdown, competitor benchmarking) require a Rated Stores account.

Limitations

• No detector is 100% accurate. Treat results as one input alongside reviewer history and purchase verification.
• Genuine reviews can use generic praise phrases. Single-review verdicts are noisier than aggregate scores.
• AI scoring depends on model availability; if unavailable, only the heuristic component is used.
• Non-English reviews are scored less accurately — heuristics are English-tuned.

The verified-customer ceiling

The single most reliable signal of authenticity is purchase verification — confirming the reviewer actually bought the product. Rated Stores enforces this for first-party reviews and surfaces it as a badge in widgets and on company pages.

How to cite this tool

Journalists, researchers, and analysts citing a Rated Stores fake-review analysis should link to the shareable result URL (format: /tools/fake-review-detector/r/[slug]) and reference this methodology page. Each analysis carries a timestamp and is reproducible from the original input.

UK / EU fake-review regulation

The UK Digital Markets, Competition and Consumers Act 2024 (effective 2025) prohibits fake reviews. Platforms have a duty to take reasonable steps to remove them. Rated Stores' detector is one component of a compliance posture — verified-customer enforcement and human moderation are the others.