Fraud and Spam Prevention Policy

Last updated: May 12, 2026

BetterReviews uses a multi-layer system to prevent fake, spam, and abusive reviews from being collected, published, or distributed. This page describes what the platform actually does today. It is intended as a reference for merchants, customers, partners, and review aggregators (including Google Merchant Center) who want to understand our integrity controls.

1. Verified-Buyer Enforcement

Reviews collected through BetterReviews' AI conversation flow originate from a unique, unguessable token tied to a real Shopify order. The customer must arrive through a post-purchase review request email; the conversation cannot start without a valid order ID, customer ID, and product ID matched against the merchant's Shopify store.

• 100% of reviews collected through the AI conversation flow are from verified buyers — they cannot be initiated without a valid review-request token.
• Tokens are bound to a specific order, customer, and product set, are time-limited, and are revoked or marked redeemed once used.
• Tokens cannot be re-used to submit reviews against products outside the original order.
• Reviews submitted through the optional public form (no review-request token) are stored as is_verified_buyer = false and never receive a verified-buyer badge.

2. AI-Powered Review Analysis

Every submitted review is processed by an AI model that returns structured integrity signals. The signals drive the moderation pipeline described in the next section. We use established third-party large language models, called from our backend over an authenticated internal channel; specific providers may change over time as model quality evolves.

• Spam detection — boolean flag with confidence (low / medium / high). Targets fake, keyword-stuffed, or copy-pasted content.
• Abuse detection — boolean flag with confidence (low / medium / high). Targets profanity, hate speech, slurs, and personal attacks. Frustrated all-caps writing alone does not qualify as abuse.
• Support-ticket detection — flags reviews that describe a defect, shipping issue, or customer-support complaint, so the message can be routed to the merchant rather than published as a review.
• Quality score — integer 1–10 reflecting how detailed, specific, and useful the review is.
• Sentiment, marketing potential, keywords, and other structural signals that feed the merchant's moderation queue and analytics.

3. Automated Moderation Pipeline

After analysis runs, every review is passed through the moderation decision pipeline. There are four possible outcomes:

• Auto-rejected — high-confidence spam or high-confidence abuse is rejected by the system and never published. The review is retained for audit, but it is invisible on storefronts and merchant dashboards as a publishable item.
• Flagged for manual review — low- or medium-confidence spam or abuse signals send the review to a "flagged" queue that requires explicit merchant action before publication.
• Routed to the merchant's support team via a bundled email cascade — reviews classified as support tickets are held in a multi-stage pending pipeline (awaiting CS response → optional hold → indefinite hold if no action). The merchant's customer-support inbox receives a bundled action email with Approve & reply privately, Hold 5 days, or Reject buttons; up to three nudges fire over 9 days if no rep clicks. Reviews in this cascade are never published while in this state. Customers receive a CAN-SPAM-compliant private reply only when a rep explicitly clicks the Reply Privately button — the merchant cannot bulk-trigger customer outreach from this pipeline. Two transactional ops notifications also flow on the operations-recipient side and are NOT marketing emails under CAN-SPAM: (1) the one-time intro email sent before the bulk-migration of pre-existing support-routed reviews into this cascade, and (2) the unified-intro email sent once per recipient when the merchant's legacy Settings → Support recipients are backfilled into the unified Notifications tab. Both carry the merchant's postal address + a one-click revoke link in compliance with Gmail bulk-sender requirements.
• Held pending — reviews that score below the merchant's auto-publish threshold remain in a pending queue until the merchant approves them. The merchant chooses how strict the threshold is via an auto-moderation preset.

Reviews that pass all of the above and meet or exceed the merchant's threshold are auto-approved and published.

4. Merchant Moderation Tools

• Merchants can approve, reject, archive, or delete any review through their dashboard. Merchants can also opt in to per-pending-review email notifications that surface the same approve / reply / archive actions to their customer-support team, with each email link bound to a specific review, a specific action, and a constant-time-matched recipient hash — so a forwarded notification cannot be used to act on a review by anyone other than the intended recipient.
• Rejected reviews are retained for audit and compliance purposes but are never displayed publicly.
• Merchants can change their auto-moderation preset; the system will re-evaluate previously held reviews against the new threshold without re-running the LLM. This includes the historical pending backlog from the manual period for reviews collected through BetterReviews' AI conversation flow or public submission. Reviews imported from other platforms (Okendo, Judge.me, CSV) are not retroactively re-scored. High-confidence spam, abuse, and support-ticket gates apply at all times. Reviews analyzed before our updated scoring rules (April 2026) are surfaced to the merchant for manual review rather than auto-published.
• System-applied moderation decisions are recorded with the decision and reason on the review's analysis record, so the rationale for an auto-reject, flag, or auto-approve can be reviewed after the fact. All merchant moderation actions (dashboard or email-driven) are recorded in an append-only audit log that is enforced at the database level and cannot be edited by the merchant.

5. Email-Channel Abuse Controls

Review-request emails are the primary collection channel. We apply additional controls to prevent the channel from being used to send unwanted mail:

• Hard-bounce suppression — email addresses with a hard bounce are added to a suppression list, and any pending review-request emails for that address are cancelled automatically.
• Spam-complaint suppression — recorded complaints (via the email-provider feedback loop) suppress the address and cancel pending sends in the same way.
• Per-merchant warmup throttling — every merchant has a daily send cap that ramps gradually, preventing bulk-mailing patterns.
• Automatic halt on poor deliverability — once a merchant has sent at least 50 emails in the previous UTC day, we calculate the hard-bounce rate and the spam-complaint rate. A bounce rate above 2% or a complaint rate above 0.1% automatically pauses outbound sends for that merchant and alerts our operations team.
• Age-guard cancellation — review requests older than 14 days are cancelled rather than sent.

6. Content Sanitization and Integrity

• All free-text fields (review title, content, reviewer name) are sanitized before storage; HTML and script tags are stripped.
• Reviews are immutable to the customer after submission. Customers cannot edit a review they have already submitted; they can only request removal.
• Server-only fields (such as the Shopify customer ID linked to a review) cannot be set from a public submission — they are always derived from the validated review-request token.
• Public submission endpoints are rate-limited by real client IP, with the trusted-proxy chain validated at the edge so submission floods cannot be hidden behind a proxy.
• Uploaded photos are validated server-side for type and size; only formats that pass validation are stored.

7. Customer Terms and Consent

Customers see a privacy and AI-processing disclosure at the point of submission, with a link to our Privacy Policy and Terms of Service. Submitting fake, misleading, or fraudulent reviews is explicitly prohibited under our Terms of Service. By submitting a review, customers acknowledge AI processing of the review for quality assessment, sentiment analysis, spam detection, and content moderation, as described in our Privacy Policy and Terms of Service.

8. Syndication to Product-Rating Services

Merchants can choose to syndicate approved reviews to third-party product-rating services through BetterReviews integrations (currently Google Shopping; Meta Shops is planned). When a merchant enables one of these integrations, the data shared per review is limited to: displayed reviewer name, review title and body, star rating, verified-buyer status, the product the review is associated with, and the submission timestamp. Reviewer email, order ID, conversation transcripts, AI analysis, Shopify customer ID, and uploaded media are never shared with these services.

Each integration is opt-in per merchant and can be disabled at any time. Disabling an integration stops new fetches; the third-party service may retain its previously-fetched copy of the feed under its own retention policy until it next crawls and receives a 404. Redactions and review-removal requests propagate to the syndication feed within 24 hours and become visible to third parties on their next crawl. The full data-flow disclosure for customers lives in our Privacy Policy under "Syndication to third-party product-rating services."

9. Reporting Fraud

If you believe a specific review on a merchant's store is fraudulent, or if you suspect a merchant is using BetterReviews to publish fake reviews, please email support@betterreviews.app with the relevant URL or order details. We review fraud reports as quickly as possible and prioritize them above routine support requests. Verified violations result in review removal, merchant warnings, or account termination depending on severity.

10. Updates

We may update this policy as our integrity controls evolve. Material changes will be reflected in the "last updated" date at the top of this page.