Google Says It Blocked 160 Million Spam Ratings in 2025—So Why Are App Store Reviews Getting Less Reliable in 2026?
Google’s enforcement numbers are huge—but users judge what they can actually see. In 2026, believable manipulation, review bombing, and ranking design make “trust” feel shakier, not stronger.
By TheMurrow Editorial
March 28, 2026
Key Points
- 1Interrogate the metric: “160 million blocked” signals enforcement scale, not that the reviews you see in 2026 are trustworthy.
- 2Expect subtler manipulation: AI makes believable persuasion cheaper, while review bombing turns ratings into a competitive battleground.
- 3Read reviews tactically: prioritize recency, scan 2–4 stars, and treat “Most helpful” sorting as curated, not neutral.
Google says it blocked 160 million spam ratings and reviews on Google Play in 2025. If you’ve been squinting at app-store feedback in 2026 and thinking, Sure doesn’t feel like it, you’re not alone.
The tension sits at the heart of modern platform trust: enforcement numbers can be enormous and still fail to answer the question users actually care about—whether the reviews that remain are representative, recent, and written by real people acting in good faith.
Google’s figure comes from its Feb. 19, 2026 Android and Play ecosystem safety update, where it also reported stopping 1.75 million+ policy-violating apps from being published and banning 80,000+ “bad” developer accounts in 2025. Engadget and TechRadar relayed the same metrics, framing them as evidence that Google is leaning heavily on AI and automation to police its store. The scale is real. So is the skepticism.
“A platform can block 160 million bad reviews and still leave users wondering whether the visible ones tell the truth.”
— — TheMurrow Editorial
What’s changed isn’t only the volume of spam. It’s the kind of persuasion that slips past filters, the way ratings are weaponized, and the design choices that make even authentic reviews less useful. Google’s number matters—but not for the reason many people assume.
What Google actually said—and what “160 million blocked” does (and doesn’t) mean
Google’s claim is specific: in 2025, it blocked 160 million spam ratings and reviews, including both inflated and deflated attempts. The phrasing matters. Google isn’t saying there were 160 million fake reviews visible to users. It’s saying protections prevented those submissions from taking effect—at some point in the pipeline.
The same blog post includes three other statistics that help frame the scale of Play’s enforcement:
- 1.75 million+ policy-violating apps prevented from being published in 2025
- 80,000+ bad developer accounts banned for trying to publish harmful apps
- 0.5-star average rating drop prevented for apps targeted by review bombing
Those numbers are platform-wide. They describe a system managing billions of interactions across regions, languages, categories, and business models. They also invite a basic journalistic question: blocked how?
The same blog post includes three other statistics that help frame the scale of Play’s enforcement:
- 1.75 million+ policy-violating apps prevented from being published in 2025
- 80,000+ bad developer accounts banned for trying to publish harmful apps
- 0.5-star average rating drop prevented for apps targeted by review bombing
Those numbers are platform-wide. They describe a system managing billions of interactions across regions, languages, categories, and business models. They also invite a basic journalistic question: blocked how?
160 million
Spam ratings and reviews Google says it blocked on Google Play in 2025 (including inflated and deflated attempts).
1.75 million+
Policy-violating apps Google says it prevented from being published on Google Play in 2025.
80,000+
“Bad” developer accounts Google says it banned in 2025 for trying to publish harmful apps.
“Blocked” isn’t a consumer-facing reliability score
A “blocked” review might mean it never posted. It might mean it posted and was removed quickly. Public summaries don’t fully define the operational details, and that’s where user perception diverges from enforcement math.
From a reader’s perspective, “review reliability” usually means something like:
- Are the top reviews authentic and written by real users?
- Are the reviews recent enough to match the current version of the app?
- Do ratings reflect typical experiences, not coordinated campaigns?
- Can people trust sorting systems like Most helpful?
Google’s statistic signals effort and scale, not necessarily the quality of what you see on an app page today. The distinction isn’t pedantic; it’s the difference between a security report and a consumer guarantee.
From a reader’s perspective, “review reliability” usually means something like:
- Are the top reviews authentic and written by real users?
- Are the reviews recent enough to match the current version of the app?
- Do ratings reflect typical experiences, not coordinated campaigns?
- Can people trust sorting systems like Most helpful?
Google’s statistic signals effort and scale, not necessarily the quality of what you see on an app page today. The distinction isn’t pedantic; it’s the difference between a security report and a consumer guarantee.
“Enforcement stats measure what a platform stopped. Users judge what the platform shows.”
— — TheMurrow Editorial
The arms race has moved from obvious spam to believable persuasion
A decade ago, fake reviews often looked like spam because they were spam: repetitive phrases, generic praise, broken English, suspicious timing. In 2026, the more consequential problem is cheaper “good writing.”
Commentary in outlets like Forbes has warned that generative AI lowers the cost of producing reviews that sound human—specific enough to feel credible, varied enough to avoid simple pattern matching. A fake review doesn’t need to be perfect. It only needs to be plausible at a glance.
Commentary in outlets like Forbes has warned that generative AI lowers the cost of producing reviews that sound human—specific enough to feel credible, varied enough to avoid simple pattern matching. A fake review doesn’t need to be perfect. It only needs to be plausible at a glance.
Why plausible fakes are harder to filter—and easier to trust
Platforms can detect certain signals: bursts from new accounts, identical text, unusual device fingerprints, coordinated patterns. But AI-generated text can be:
- Unique every time while still carrying the same marketing message
- Detailed (feature mentions, timelines, complaints) without being real
- Localized to region and language, complicating moderation consistency
This is why the “blocked 160 million” metric can rise even as users feel review sections are deteriorating. Platforms may be stopping an ocean of low-effort junk, while higher-effort manipulation—less frequent, more strategic—slips through.
Engadget’s coverage of Google’s annual safety update emphasizes AI’s role in enforcement. The uncomfortable symmetry is that AI also makes manipulation cheaper. The same technology that helps stop spam can also help write it.
- Unique every time while still carrying the same marketing message
- Detailed (feature mentions, timelines, complaints) without being real
- Localized to region and language, complicating moderation consistency
This is why the “blocked 160 million” metric can rise even as users feel review sections are deteriorating. Platforms may be stopping an ocean of low-effort junk, while higher-effort manipulation—less frequent, more strategic—slips through.
Engadget’s coverage of Google’s annual safety update emphasizes AI’s role in enforcement. The uncomfortable symmetry is that AI also makes manipulation cheaper. The same technology that helps stop spam can also help write it.
What this means for readers choosing apps
For users, the practical implication is simple: a handful of convincing reviews can outweigh hundreds of honest but vague ones. If the most visible comments sound “real,” many people stop investigating.
Reliability, in 2026, often fails quietly. The reviews don’t scream “fake.” They whisper “probably.”
Reliability, in 2026, often fails quietly. The reviews don’t scream “fake.” They whisper “probably.”
Review manipulation isn’t just fake 5-stars—it’s coordinated 1-star warfare
Google didn’t only tout blocked spam. It also said it prevented an average 0.5-star rating drop for apps targeted by review bombing. That sentence does a lot of work. It implies review bombing is common enough—and measurable enough—that Google can quantify its impact.
Review bombing changes the trust problem in two ways. First, it normalizes the idea that ratings can be driven by motives unrelated to the app’s quality. Second, it forces platforms into visible intervention, which can itself reduce trust.
Review bombing changes the trust problem in two ways. First, it normalizes the idea that ratings can be driven by motives unrelated to the app’s quality. Second, it forces platforms into visible intervention, which can itself reduce trust.
0.5-star
Average rating drop Google says it prevented for apps targeted by review bombing in 2025.
When platforms intervene, users see curation—sometimes unfairly
If Google suppresses a bombing campaign, some users will say: good, the store protected an app from harassment. Others will say: the store manipulated the rating.
Both reactions can coexist because they stem from the same reality: review systems are no longer passive reflections of sentiment. They are moderated arenas.
And even if platforms blunt the worst attacks, users may still encounter:
- Residual low-star waves that look organic enough
- Mixed signals where legitimate criticism gets downranked
- Controversies where people assume the platform “protected” a developer
Google’s own framing—“inflated and deflated” spam—shows it sees manipulation in both directions. That should make consumers more sophisticated, not more cynical: a five-star wall can be engineered, and so can a one-star cliff.
Both reactions can coexist because they stem from the same reality: review systems are no longer passive reflections of sentiment. They are moderated arenas.
And even if platforms blunt the worst attacks, users may still encounter:
- Residual low-star waves that look organic enough
- Mixed signals where legitimate criticism gets downranked
- Controversies where people assume the platform “protected” a developer
Google’s own framing—“inflated and deflated” spam—shows it sees manipulation in both directions. That should make consumers more sophisticated, not more cynical: a five-star wall can be engineered, and so can a one-star cliff.
“Once ratings become a battleground, every number looks like strategy.”
— — TheMurrow Editorial
Metrics can mislead without definitions: what counts as spam, and who decides?
The 160 million figure is impressive, but it raises basic questions that public reporting rarely answers. Security metrics are not self-explanatory. They depend on definitions.
The missing details that change the story
Google’s summary doesn’t fully explain, at least in the public-facing material:
- Whether “blocked” means never posted or removed after posting
- What portion was caught by automation versus user reports
- How enforcement varies by language, region, and app category
- How edge cases are handled—especially gray areas like incentivized-but-not-disclosed reviews or campaigns using real accounts
Those details matter because modern manipulation often avoids the most obvious policy violations. A campaign can use legitimate devices and accounts, post varied text, and still be coordinated. The intent is deceptive even if the inputs look “human.”
- Whether “blocked” means never posted or removed after posting
- What portion was caught by automation versus user reports
- How enforcement varies by language, region, and app category
- How edge cases are handled—especially gray areas like incentivized-but-not-disclosed reviews or campaigns using real accounts
Those details matter because modern manipulation often avoids the most obvious policy violations. A campaign can use legitimate devices and accounts, post varied text, and still be coordinated. The intent is deceptive even if the inputs look “human.”
Why users feel the gap between policy and experience
Platforms can be technically correct—blocking millions of violations—while leaving consumers with a feed of reviews that are:
- Authentic but unhelpful
- Helpful but outdated
- Recent but unrepresentative of average use
None of those problems show up in a spam-blocking counter. That counter measures enforcement, not the epistemic quality of the review section.
From a reader’s standpoint, the most honest way to interpret Google’s number is: Google is spending heavily on the problem. It is not: the problem is solved.
- Authentic but unhelpful
- Helpful but outdated
- Recent but unrepresentative of average use
None of those problems show up in a spam-blocking counter. That counter measures enforcement, not the epistemic quality of the review section.
From a reader’s standpoint, the most honest way to interpret Google’s number is: Google is spending heavily on the problem. It is not: the problem is solved.
Key Insight
“160 million blocked” is an enforcement metric, not a promise that the reviews you see are representative, recent, or useful.
Cleaning up the store can raise the stakes—and the incentive to game it
Google Play is also shrinking. TechCrunch, citing Appfigures, reported that Play’s app count fell from about 3.4 million to about 1.8 million from the start of 2024 to April 2025—a 47% decline—attributed to stricter policies and enforcement.
A cleaner store sounds like an unambiguous win. In many ways it is. Fewer junk apps should mean less noise, fewer scams, and better baseline quality. But the competitive dynamics shift when supply contracts.
A cleaner store sounds like an unambiguous win. In many ways it is. Fewer junk apps should mean less noise, fewer scams, and better baseline quality. But the competitive dynamics shift when supply contracts.
47%
TechCrunch (citing Appfigures) reported Google Play’s app count fell from ~3.4M to ~1.8M from early 2024 to April 2025.
Fewer apps can mean fiercer competition for attention
When marginal spam apps are removed, the remaining developers often face:
- More pressure to climb rankings and search results
- Higher customer acquisition costs
- Greater dependence on ratings as conversion signals
That environment can intensify the temptation to manipulate reviews. Not necessarily through cartoonish fake accounts, but through subtler levers: coordinated communities, “review encouragement” that skirts disclosure rules, or attempts to bury rivals under negative sentiment.
The point isn’t that cleanup causes manipulation. The point is that a healthier marketplace can still be a more competitive one—and competition has a long history of producing incentives to cheat.
- More pressure to climb rankings and search results
- Higher customer acquisition costs
- Greater dependence on ratings as conversion signals
That environment can intensify the temptation to manipulate reviews. Not necessarily through cartoonish fake accounts, but through subtler levers: coordinated communities, “review encouragement” that skirts disclosure rules, or attempts to bury rivals under negative sentiment.
The point isn’t that cleanup causes manipulation. The point is that a healthier marketplace can still be a more competitive one—and competition has a long history of producing incentives to cheat.
What readers should take from the app-count decline
The 47% reduction is a reminder that Google Play is actively reshaping itself. That helps explain why enforcement numbers are so large: the platform is not only reacting to bad behavior; it is tightening entry and survival.
In that context, “160 million blocked reviews” reads less like a victory lap and more like a progress report from a store in constant triage.
In that context, “160 million blocked reviews” reads less like a victory lap and more like a progress report from a store in constant triage.
Even “real” reviews can be unreliable: design choices that distort usefulness
A crucial distinction gets lost in most debates: authenticity is not the same as reliability.
A review can be written by a genuine user and still mislead you—because the app changed, the business model changed, or the platform’s interface made certain feedback more visible than others.
A review can be written by a genuine user and still mislead you—because the app changed, the business model changed, or the platform’s interface made certain feedback more visible than others.
Version drift: the app you download isn’t the app they reviewed
Many apps evolve quickly. Features disappear, subscriptions replace one-time pricing, ads become more aggressive, or a new owner changes priorities. Reviews, meanwhile, can remain frozen in time.
Readers experience this as a betrayal: thousands of four-star reviews, then a new install that feels like a different product. Nothing “fake” happened. The signal just aged.
Readers experience this as a betrayal: thousands of four-star reviews, then a new install that feels like a different product. Nothing “fake” happened. The signal just aged.
Prompting mechanics: the store asks at the happiest moment
In-app rating prompts are notorious for capturing users when they’re most likely to say yes—after a win, a completed task, or a seamless onboarding moment. That can skew distributions upward even when long-term satisfaction is mediocre.
Again, this isn’t necessarily fraud. It’s architecture.
Again, this isn’t necessarily fraud. It’s architecture.
Sorting systems: “Most helpful” is a power tool
What you see first shapes what you believe. Platforms rank and filter reviews because raw chronological lists are unusable at scale. But ranking becomes a second battleground.
Apple’s Digital Services Act risk assessment report (non-confidential) offers a useful window into how a major platform thinks about this. Apple describes ordering reviews by “helpfulness,” considering factors such as source, quality, thoroughness, and timeliness, and relying on monitoring and reporting workflows. The important takeaway isn’t that Apple is better or worse. It’s that the ordering system is editorial—it decides which reality most users encounter.
Google’s anti-spam number doesn’t capture any of these design-driven reliability problems, yet these are the problems people complain about most.
Apple’s Digital Services Act risk assessment report (non-confidential) offers a useful window into how a major platform thinks about this. Apple describes ordering reviews by “helpfulness,” considering factors such as source, quality, thoroughness, and timeliness, and relying on monitoring and reporting workflows. The important takeaway isn’t that Apple is better or worse. It’s that the ordering system is editorial—it decides which reality most users encounter.
Google’s anti-spam number doesn’t capture any of these design-driven reliability problems, yet these are the problems people complain about most.
Key Takeaway
Authentic reviews can still be unreliable when apps change fast and platforms curate visibility through prompts, filters, and “helpfulness” ranking.
Google vs. Apple: why users compare them, and what the comparison misses
Even when the headline is about Google, readers naturally triangulate: Is the App Store any better? The honest answer is that both ecosystems face similar pressures—enormous scale, high incentives, and a constant cat-and-mouse game.
Apple’s DSA risk report underscores a shared truth: app stores don’t simply host reviews; they curate them through ranking, moderation, and reporting systems. That curation can protect users from manipulation, but it also makes “trust” feel like something granted by the platform rather than earned by the crowd.
Apple’s DSA risk report underscores a shared truth: app stores don’t simply host reviews; they curate them through ranking, moderation, and reporting systems. That curation can protect users from manipulation, but it also makes “trust” feel like something granted by the platform rather than earned by the crowd.
The more platforms curate, the more they must explain
Google’s public metrics are strong on scale and weaker on the user-facing mechanics behind them. Apple’s report provides more explicit detail on ranking considerations—at least in the portion cited—highlighting that review ordering is not neutral.
Both approaches invite the same consumer question: Why am I seeing these reviews first?
The comparison also risks a false binary. Trust isn’t just about which store blocks more spam. Trust is about whether a platform helps users form accurate expectations—especially when app quality can swing wildly between versions.
Both approaches invite the same consumer question: Why am I seeing these reviews first?
The comparison also risks a false binary. Trust isn’t just about which store blocks more spam. Trust is about whether a platform helps users form accurate expectations—especially when app quality can swing wildly between versions.
How “trust” gets shaped in both major app stores
Before
- Google—publishes big enforcement totals; user-facing mechanics behind “blocked” and visibility are less defined in public summaries
After
- Apple—DSA risk report describes “helpfulness” ordering factors like source
- quality
- thoroughness
- timeliness
Practical ways to read app reviews in 2026 without getting played
Consumers can’t audit Google’s detection systems. But readers can change how they interpret the evidence in front of them. The goal isn’t paranoia; it’s discipline.
A smarter review checklist (fast, realistic, effective)
When deciding whether to download—or pay—use a few checks that cut through both spam and “authentic but unhelpful” noise:
Review checklist for 2026
- ✓Filter for recency if the store allows it. Version drift makes older reviews less predictive.
- ✓Read a spread: 2-star to 4-star reviews often contain the most actionable detail.
- ✓Look for specific constraints (device model, region, subscription tier, accessibility needs). Specificity can be faked, but patterns across many reviewers are harder to stage.
- ✓Watch for review clustering: sudden bursts around updates, controversies, or promotions can distort the signal.
- ✓Compare reviews to the app’s recent change log and business model. A spike in complaints about paywalls often correlates with monetization shifts.
What developers and platforms should hear in user skepticism
If Google wants the “160 million blocked” figure to rebuild confidence, not just impress policymakers, it may need to publish clearer definitions: what “blocked” includes, how ranking and review visibility work, and how it avoids suppressing legitimate negative feedback in the name of anti-spam.
Users aren’t asking for perfection. They’re asking for legibility—rules they can understand and outcomes that match lived experience.
Users aren’t asking for perfection. They’re asking for legibility—rules they can understand and outcomes that match lived experience.
Conclusion: 160 million blocked reviews is real progress—and also the wrong comfort
Google’s 2025 numbers—160 million spam reviews blocked, 1.75 million+ policy-violating apps stopped, 80,000+ bad developer accounts banned, and an average 0.5-star drop prevented from review bombing—describe a store fighting at scale. They do not, by themselves, guarantee that what you see in 2026 is trustworthy.
The deeper story is structural. Persuasion has gotten cheaper. Manipulation has gotten subtler. Ratings have become a lever in competitive conflict. And even honest feedback can mislead when platforms surface it through ranking systems that optimize for “helpfulness,” engagement, or velocity rather than representativeness.
App reviews still matter, but they work best when treated as one signal among several—and when readers remember that a five-star average isn’t a verdict. It’s a negotiation between users, developers, and the platform’s invisible hand.
The deeper story is structural. Persuasion has gotten cheaper. Manipulation has gotten subtler. Ratings have become a lever in competitive conflict. And even honest feedback can mislead when platforms surface it through ranking systems that optimize for “helpfulness,” engagement, or velocity rather than representativeness.
App reviews still matter, but they work best when treated as one signal among several—and when readers remember that a five-star average isn’t a verdict. It’s a negotiation between users, developers, and the platform’s invisible hand.
T
About the Author
TheMurrow Editorial is a writer for TheMurrow covering reviews.
Frequently Asked Questions
Did Google say there were 160 million fake reviews on Google Play?
No. Google said it blocked 160 million spam ratings and reviews in 2025. “Blocked” suggests the platform prevented those submissions from affecting app pages, but the public summary does not fully specify whether all were stopped before posting or removed after appearing.
Why do app reviews still feel unreliable if Google is blocking more spam?
Because “spam blocked” doesn’t measure whether the remaining reviews are representative, recent, or useful. Reviews can be authentic but outdated (version drift), skewed by in-app prompts, or distorted by ranking systems that decide which reviews most users see first.
What is review bombing, and how does it affect ratings?
Review bombing is a coordinated wave of negative reviews—often driven by controversy, competitors, or organized campaigns rather than typical user experience. Google said it prevented an average 0.5-star rating drop for apps targeted by bombing, implying it actively mitigates such attacks.
Are AI-generated fake reviews a real concern in app stores?
Yes. Commentary has warned that generative AI can produce credible-sounding reviews at low cost, making manipulation harder to detect. Even if platforms block large volumes of obvious spam, smaller amounts of more persuasive, varied text can slip through and shape perception.
Does a smaller Google Play store mean reviews are more trustworthy?
Not necessarily. TechCrunch reported Play’s app count fell from ~3.4M to ~1.8M from early 2024 to April 2025 (a 47% drop), linked to stricter enforcement. A cleaner marketplace can still be highly competitive, and high stakes can increase incentives to manipulate ratings.
How can I quickly tell if an app’s reviews are worth trusting?
Focus on recent reviews, read across mid-range ratings (2–4 stars), and look for repeated themes across many users rather than one dramatic story. Check whether complaints match recent updates or monetization changes. Treat “Most helpful” as curated, not neutral.
