Evally
Start free
Trust & safety

Fake review detector

Flag fake and incentivized reviews, and route real defect reports to the right team.

Genuine reviews mention specifics. This judge weighs specificity against superlatives, off-topic content, incentives and promotion, and separately catches reviews that report a real defect or safety issue so they reach a person quickly.

Use this judgeCustomize it

Try it live

Real calls to Jev. Pick an example or paste your own.

Examples
16
89
No verdict yet

Pick an example or paste your own input, then run it.

Under the hood

Every question and rule, in the open. Fork it to change any of them.

What it asks

Reads product and review, then asks 6 questions in one call.

  1. Q1Specific detailsYes / no

    Does the review mention specific details of using the product (a feature, a flaw, a situation) rather than generic praise or complaints?

  2. Q2IncentivizedYes / no

    Does the review mention receiving the product free, at a discount, or with a reward for reviewing?

  3. Q3Off topicYes / no

    Is the review about something other than the product (only shipping, a different product, or nothing at all)?

  4. Q4Extreme without reasonsYes / no

    Is the review all superlatives or all outrage, with no concrete reason?

  5. Q5Promotes something elseYes / no

    Does the review promote another product, store or link?

  6. Q6Reports a defectYes / no

    Does the review report a real defect or safety problem with the product? (Damage and malfunction count; disliking the color doesn't.)

How it decides

Rules run in code, top to bottom. The first one that matches gives the verdict.

  1. 1
    If Promotes something else ≥ 0.7Likely fake
  2. 2
    If Reports a defect ≥ 0.7Defect reportedreview
  3. 3
    If Fake risk ≥ 70Likely fake
  4. 4
    If Specific details < 0.15 and Extreme without reasons ≥ 0.8Likely fake
  5. 5
    If Incentivized ≥ 0.7Suspiciousreview
  6. 6
    If Fake risk ≥ 45Suspiciousreview
  7. ·
    DefaultGenuine
Fake risk

A 0-100 score weighted from Specific details (inverted) ×2, Extreme without reasons ×1.5, Off topic ×1, Promotes something else ×2.

Call it from your code

Every template runs as-is with an API key, no setup: POST /api/v1/judges/fake-review-detector/run.

curl -X POST https://tryevally.com/api/v1/judges/fake-review-detector/run \  -H "Authorization: Bearer $EVALLY_API_KEY" \  -H "Content-Type: application/json" \  -d '{  "input": {    "product": "Wireless earbuds",    "review": "BEST PRODUCT EVER!!! Amazing amazing amazing. Five stars. Buy now, you won'\''t regret it!!!"  }}'

More in Trust & safety