Five Stars, Thin Evidence: Auditing App Store Ratings Before the CMA Comes Calling
Photo: Brion VIBBER, CC BY-SA 4.0, via Wikimedia Commons
Every app team believes its reviews are genuine. Very few can prove it in a morning.
That gap used to be a marketing footnote. Under the Digital Markets, Competition and Consumers Act 2024, with the consumer enforcement regime live since April 2025, the CMA can investigate and fine a trader directly — no court trip required — with penalties reaching 10% of global turnover. Reviews are explicitly in the crosshairs, and an app store rating is a review display like any other.
So treat this as a launch-readiness item. Here's how to run the audit while it's still cheap.
Four banned practices, in plain English
The unfair commercial practices chapter of the DMCC sets out review behaviour that is prohibited outright. Paraphrased, you cannot:
- Write or commission fake reviews — including paying a service to post them.
- Incentivise anyone to write a review without taking reasonable steps to make sure it reflects a real experience.
- Publish reviews without reasonable steps to check they're genuine.
- Misrepresent reviews — selective display, misleading aggregate scores, or editorial tweaks that change what a reviewer meant.
The CMA's guidance on unfair commercial practices, published in 2025, is the document to read alongside this. It has views on review gating — selectively inviting only the customers you expect to be happy — and those views are not friendly.
Where app teams actually leak
Almost never in a dramatic way. It's the small stuff.
Prompt copy with a nudge. 'Enjoying the app? Leave us five stars!' is a request for a rating, not an honest one. Both Apple and Google prohibit incentivised reviews, and frameworks like StoreKit's review prompt and the Play In-App Review API have their own rate limits and rules — but the copy around them is where teams get creative and where the evidence sits.
Gating by sentiment. A pre-prompt that asks 'are you happy?' and only routes the happy ones to the store is a textbook example of the pattern the CMA has flagged.
The vendor you inherited. A review platform bought by a previous marketing lead, whose contract promises 'managed review generation' in language nobody has read since.
Seeded reviews at launch. Colleagues, agency staff, friends of the founders, sometimes with a note in a Slack channel that says 'just for day one'. That Slack channel is discoverable.
Review swaps and 'ASO' bundles. Anything that exchanges a rating for a benefit — credits, entries, access — is a problem, however it's dressed up.
Influencer CTAs. A creator saying 'and drop us five stars in the App Store' is an incentivised review if there's a paid relationship behind the video. Your creator briefs are review marketing.
The stale aggregate. Showing a 4.8 on your marketing site from a filtered subset of reviews, or a score that hasn't been refreshed since a major version. Misrepresentation doesn't stop at the store boundary.
The audit, in an afternoon
Start with an inventory. Every place a star, score, quote or review count appears: the App Store listing, Google Play listing, your website, paid landing pages, email templates, in-app prompts, social bios and any retail or partner listing you've supplied copy to.
Then walk the funnel:
- Read your in-app prompt copy out loud. Does it ask for a feeling or a rating?
- Open the prompt logic in the codebase or config. Is there sentiment branching before the store link?
- List every vendor touching reviews or ratings. For each, get the contract, the statement of work and the current campaign brief.
- Search internal chat for 'review', 'stars', 'rating' and 'testimonial'. You'll find things.
- Look at the last three months of published reviews for a pattern — same-day bursts, identical phrasing, accounts with one review.
- Check whether you moderate reviews you host yourself. Deleting the bad ones is not moderation.
The paper trail you want on file
If the CMA, a client's procurement team or a nervous investor asks, the answer should be a folder, not a shrug. Dated evidence of:
- A written review policy covering staff, agencies and creators: no incentives, no gating, no fabricated content.
- Vendor due diligence — who they are, what they do, what they've contractually promised, and how you'd exit.
- The 'reasonable steps' record. What you actually do to check reviews are genuine: moderation rules, fraud signals, takedown of suspicious batches, and who owns it.
- A decision log for anything you've removed or queried, with reasons.
- Training evidence. A short, dated briefing to whoever writes prompts, briefs creators or runs paid social — plus a note in onboarding for new starters.
- A complaints route for review disputes, and a named owner with a job title, not a shared alias.
Fixing it without tanking conversion
The fear is that honesty costs you stars. In practice the fixes are cheap and mostly invisible to users:
- Rewrite prompts to ask for a rating without steering the outcome. Keep the timing and the frequency; change the wording.
- Remove sentiment gating. If you want to hear from unhappy users, route them to support, not away from the store.
- Replace the launch-week seeding plan with a real tester programme and an invite to people who actually used the build.
- Put review language into creator briefs as a red line, and check the videos before they go live.
- Refresh aggregates on a schedule, and label the source and date.
- Respond to reviews, including the bad ones. A visible reply habit is decent evidence of good faith.
None of that requires a rebuild. It requires someone to own it.
The Tuesday-afternoon test
Which is where this lands: if a regulator asked on Tuesday afternoon to see how you ensure your ratings are genuine, could you produce the documents before the kettle boils? If the answer is no, the work isn't the review platform. It's the folder.