Buy on Amazon with verified reviews.
We score review authenticity (Auth%), track price trends, and flag suspicious patterns — so you buy with confidence.
- ✓ Review Authenticity Score
- ✓ Price Trend Tracking
- ✓ Custom Alerts & Watchlists
How it works
Paste an Amazon link or ASIN and start a free scan.
We detect review authenticity patterns and price trends.
See Auth%, red flags, and price history to buy with confidence.
Sample report
See how CrowdTruth scores review authenticity (Auth%), flags risky patterns, summarizes buyer sentiment, and tracks price history.
Air Purifier HEPA H13 (Sample)
Trends & distributions
Buyer topics (what verified buyers say)
Flags & risk signals
Higher-Auth% similar picks
Methodology (quick)
Auth% combines signals including verified-buyer ratio, review timing bursts, language similarity, reviewer reputation, incentive patterns, and cross-listing leakage. Read full methodology
Methodology
How we calculate the Review Authenticity Score (Auth%), derive confidence intervals, identify risk signals, and compute price & trending insights. Clear, practical, and auditable.
Overview
Auth% estimates how trustworthy a product’s reviews are (0–100) using a weighted ensemble of signals spanning reviewer behavior, timing, language, and listing integrity. It complements (not replaces) star ratings: high ratings can coexist with lower Auth% if manipulation is likely. We show tiers for quick reading and confidence bands where applicable.
Auth% is probabilistic and relies on available data. Use it alongside price history, flags, and your own judgment.
Signals (what we measure)
Verified-buyer share, account tenure, cross-product diversity, abnormal posting velocity.
Sudden spikes near promos; decay curves; weekend/holiday asymmetries vs baseline.
Template-like phrasing, repeated n-grams across reviewers, sentiment skew without variance.
Coupon/giveaway phrasing, “free for review”, undisclosed affiliate-style language.
Merged/variant listings where reviews mismatch features/ASINs (evidence of review transfer).
Natural disagreement vs. monolithic praise; topic polarity spread and entropy.
Sample-size & minimums
- Minimum recent reviews: 50 (or 20 with a wider confidence band).
- Recency window: Typically 90 days; older reviews are down-weighted.
- Variant handling: Variants are de-duplicated; signals computed per canonical listing.
Scoring & thresholds (how Auth% is computed)
Auth% is a normalized 0–100 composite built from standardized signal sub-scores. At a high level:
Auth% = 100 × σ( w₁·RVR + w₂·TB + w₃·LD + w₄·IH + w₅·CL + w₆·SV + b )
σ is a squashing function that maps the linear combination into [0,1]; weights (wᵢ) are tuned to balance false positives/negatives. Weights may be adjusted as the system learns.
High verified-buyer share and credible reviewer history buoy Auth%.
Burst intensity vs baseline; promo-aware to reduce false alarms.
Repetitive language and incentive hints strongly reduce Auth%.
Cross-listing anomalies and unnatural sentiment uniformity lower trust.
Tiers (display & decisions)
Tiers are guardrails, not absolutes. Always consider context (category norms, review volume, price movement).
Confidence & error bars
We expose a confidence interval (CI) when sample sizes are small or signals conflict. CI widens as uncertainty grows.
Interpretation: “We estimate true authenticity at ~78%, with plausible range 68–86% at current sample size.”
Trending & category ranks
We label “Trending” when there’s a recent combination of review activity, stable or improving Auth%, and meaningful price movement.
Spike vs. trailing baseline, promo-aware.
Flat or rising Auth% preferred.
Relative to 30–90d average.
Category rank blends Auth% × volume with recency weighting; extremely volatile items may be temporarily damped.
Data handling & integrity
- We do not modify or remove reviews. We analyze publicly available text/metadata and render our own summary.
- Canonicalization: Variants & merged listings are reconciled; mis-attached reviews trigger cross-listing flags.
- Recency: Older reviews are down-weighted; we prefer last 90 days for core scoring.
- Censorship resistance: Disagreeing topics are preserved; low-volume dissent increases variance, not hidden.
- Reproducibility: Re-running on the same snapshot yields stable results (within CI), barring data changes.
Known limitations
- Data availability: Private/hidden/filtered reviews are out of scope.
- Rapid promotions: Genuine promo bursts may resemble manipulation; we mitigate but cannot fully eliminate overlap.
- Category quirks: Some categories (e.g., consumables) have distinct baselines; we normalize but edge cases remain.
- Language/model drift: New spam tactics evolve; weights and detectors are updated periodically.
- Small samples: Wider CI; we’ll indicate when conclusions are tentative.
Glossary
- Auth%
- Estimated authenticity of a review set (0–100). Tiers: 0–39 risky, 40–59 mixed, 60–79 reliable, 80–100 very strong.
- Verified buyer
- Account with purchase evidence for the reviewed item.
- Review burst
- Abnormal, concentrated inflow of new reviews vs baseline.
- Linguistic duplication
- Repeated phrasing patterns across different reviewers indicating coordination.
- Cross-listing leakage
- Review text referencing features of other variants/listings; signals review transfer.
FAQ
What is Auth%?
Auth% estimates how trustworthy a product’s reviews are (0–100) using signals like reviewer history, timing bursts, linguistic duplication, verified-buyer share, and cross-listing leakage.
What thresholds do you use?
0–39 risky, 40–59 mixed, 60–79 reliable, 80–100 very strong.
Do you edit reviews?
No. We only analyze and summarize.
Could high ratings have low Auth%?
Yes. Ratings reflect satisfaction; Auth% reflects trust in the review set.
Does Auth% guarantee truth?
No. It’s probabilistic; use with context and price trends.
External product data from Amazon. Content is informational only and not medical advice. Amazon® is a registered trademark of Amazon.com, Inc. or its affiliates. If we participate in the Amazon Associates Program, we may earn from qualifying purchases.
Simple, fair pricing
Start free. Upgrade for unlimited scans, historical trends, compare, alerts, and exports. Cancel anytime.
Free
- 50 free scans on signup
- Review Authenticity Score (Auth%)
- Basic price trend
- Sample report
Starter
- 200 scans / month
- Historical trends (90 days)
- Watchlists & basic alerts
- Priority fetching
Pro
- 1,000 scans / month
- Compare up to 10 products
- CSV/PDF export
- Email digests (Top 5 trusted deals)
Teams
- Unlimited scans
- API & data feed access
- Role-based access & SSO
- SLA & support
Compare features
| Feature | Free | Starter | Pro | Teams |
|---|---|---|---|---|
| Scans / month | 50 signup bonus | 200 | 1,000 | Unlimited |
| Auth% & flags | ✓ | ✓ | ✓ | ✓ |
| Historical trends | Basic | 90 days | 365 days | 365 days |
| Compare products | — | Up to 5 | Up to 10 | Up to 20 |
| Exports (CSV/PDF) | — | — | ✓ | ✓ |
| Watchlists & alerts | — | Basic | Advanced | Advanced |
| API access | — | — | — | ✓ |
| Support | Community | Standard | Priority | SLA |
Pricing FAQ
What counts as a scan?
One product authenticity evaluation, including flags and trend snippets. Re-scans within a short window may be cached and won’t double-count.
Do I need a credit card for the free tier?
No. Start free with 50 scans on signup.
Can I cancel anytime?
Yes. Plans are month-to-month. Annual plans are prepaid with a discount.
Do you offer refunds?
If we experience sustained downtime or a defect prevents usage, contact support for a resolution.
Do you support teams?
Yes — Teams adds seats, roles, SSO, API, and SLA. Contact sales.
External product data from Amazon. Content is informational only and not medical advice. Amazon® is a registered trademark of Amazon.com, Inc. or its affiliates. If we participate in the Amazon Associates Program, we may earn from qualifying purchases.