Google's 2026 Review Filter: What Changed

We track Google review filter behaviour across 640 Google Business Profiles that we scrape weekly for benchmarking, plus the 214 removal cases and 3,100 review-buy orders we shipped in 2025 and 2026. Something changed in Q1 2026. Filter rates on brand-new reviews across our monitored profiles jumped from a rolling 8% to a rolling 22% inside eight weeks, then settled around 17% by April. This article explains what we saw, what Google confirmed publicly, and what the operational changes mean for anyone posting new reviews - whether through customer requests, employee campaigns, or a paid provider.

Perves runs the QA side and Adam runs the removal side; the tests below were run across live client profiles with matched control conditions. Numbers are ours, not Google's.

What Google publicly confirmed changed

Google's Business Profile Help centre updated its "About fake reviews" page in February 2026, and Product Manager comments in the Google Business Profile Community forum confirmed three specific changes:

1. Machine learning models were retrained on 2025 review data to identify AI-generated language patterns, including the tell-tale rhythms of GPT and Gemini-generated text.
2. Automated systems now use "additional signals" (Google's wording) to detect reviews from accounts with unusual posting patterns, including sudden geographic mismatches and cross-industry review histories.
3. Business owner reporting was streamlined so flagged reviews now route to a triage team faster, with a target resolution of 3 to 7 days instead of the previous 7 to 21.

What Google did not confirm publicly, but is visible in the filter behaviour, is a much tighter velocity threshold and stronger IP clustering. That is what changed the numbers on the ground.

The three levers that tightened

1. Velocity thresholds dropped

The old rule of thumb was that a Google Business Profile could absorb 5 to 8 new reviews per day without triggering filter suspicion, provided the profile had a healthy long-term review history. In our 2026 data, that number has dropped to 2 to 3 per day for SMBs with fewer than 200 total reviews. Larger profiles (500+ reviews) still absorb 5 to 7 without issue. Push above the threshold for two consecutive days and Google soft-filters roughly 30 to 45% of the new reviews within 14 days.

2. IP and device clustering is stricter

Google has always looked at IP clustering to spot review farms, but the 2026 update expanded the fingerprint to include browser canvas hashes, WebGL renderer, timezone-plus-language mismatches, and reused device identifiers via mobile app signals. In our tests, 20 reviews posted from a single residential IP over 10 days filtered at 68%; the same 20 posted from 20 different residential IPs across 20 different cities filtered at 11%.

3. AI-generated text is scored separately

This is the biggest quiet change. Reviews now pass through an AI-detection classifier before the general filter. Text that scores above roughly 0.7 on the classifier (our estimate from filter behaviour, not a public number) gets down-weighted in the Places ranking algorithm and is more likely to be removed on report. Reviews written by real customers score below 0.3 in most cases; reviews generated in bulk from a single LLM prompt score above 0.8 almost universally.

The single fastest way to trigger the new filter is to post reviews written from one AI prompt in one sitting from one IP. That was a sloppy pattern before; now it is a suicide pattern. - Perves, Head of Fulfillment & QA, BGR Review.

What we tested in Q2 2026

Between April and June 2026 we ran controlled posting tests across 24 client Google Business Profiles. Same profiles, same industries, same reviewer pool structure. What varied was velocity, IP diversity, device diversity, and text authorship. Numbers below are the 30-day filter rate - the share of posted reviews that Google either soft-filtered or removed.

The takeaway: velocity and diversity matter, but text authorship is the single largest variable. A slow-dripped LLM review from a diverse IP pool filters six times more often than the same volume in fully human text.

What still works in the new filter

Reviews that consistently stick share the same pattern. Match these six conditions and 30-day filter rates stay under 5%.

1. Aged accounts. The reviewer's Google account is at least 12 months old with 4+ prior reviews across at least 2 categories.
2. Geographic proximity. The reviewer's prior reviews cluster within 100 km of the target business, or the reviewer's Google Maps location history places them near the business.
3. Human-written text. Text that scores under 0.3 on the AI classifier. Short, specific, dated details ("visited last Thursday for a wax", "delivered a Samsung TV to our office in Wembley") score lower than long, adjective-heavy reviews.
4. Photos. At least 30% of reviews in a posting batch include a photo taken on-site. Filter rates drop by roughly 40% when photos are attached.
5. Response from the business. The business responds publicly to every review within 48 hours. Response cadence signals authenticity to both users and to the filter.
6. Drip pacing. 2 to 3 reviews per day for SMBs, 5 to 7 for large profiles, spread across time zones and days of the week.

What breaks under the new filter

• Bulk providers that post 50 reviews in 48 hours from a small IP pool. Filter rates run 55 to 80% inside a quarter, plus a soft profile suspension risk.
• Employee campaigns from office WiFi. Same building, same IP, same time window. Even genuine reviews from real employees can filter. Have staff post from home or mobile data, on different days.
• Copy-paste template reviews. Even with different accounts, near-duplicate text triggers the near-duplicate filter. Every review needs unique phrasing.
• Reviews from accounts with no photo, no name, no prior activity. Empty accounts filter faster than they did in 2025.
• AI-generated bulk reviews. Covered above. The classifier is the fastest-moving change and the hardest to work around at scale.

The BGR posting rules for 2026

Our internal posting checklist changed in March 2026 to match the new filter behaviour. Every review that leaves our desk now passes these six checks.

1. Reviewer account age is 12+ months and has 4+ prior reviews across 2+ categories.
2. Reviewer IP is residential, geographically appropriate, and unique to that review that week.
3. Text is human-written by our writing team, targeting under 0.3 on our internal AI classifier which mirrors Google's public tells.
4. At least 30% of a posting batch includes on-site photos or matching evidence.
5. Drip pacing respects the profile's existing review velocity, never exceeding 2 to 3 new reviews per day for SMB profiles.
6. Every review is monitored for 30 days; any filtered review is replaced free under the 30-day replacement guarantee.

If you use our buy Google reviews service, this is the same checklist that runs on every order. The 30-day replacement guarantee is what makes the pricing work: we absorb the filter risk, not the client.

What to do if your existing reviews got filtered

1. Do not repost the same content from the same accounts. That trips the filter faster.
2. Audit which reviews filtered. Google Business Profile shows total review count; compare against your posted list to find the delta.
3. Categorize by pattern. Same IP, same day, same account age band, same text style. The pattern tells you what tripped.
4. Replace with a different pattern. Different accounts, different IPs, different pacing, human-written text.
5. Wait 14 days between the takedown and the replacement post to let the filter's memory of the pattern age out.
6. Layer in real customer requests alongside any paid replacements. The mix of paid plus real always beats paid alone against the new filter.

What this means for review buying in 2026

The market split. Cheap bulk providers ($1 to $3 per review) are essentially unusable for anything except one-off spam campaigns. Their filter rate crossed 50% in Q2 2026 and their profile suspension rate rose in parallel. Paid providers that run drip delivery, aged residential accounts, and human-written text ($15 to $45 per review) still post reliably at under 8% filter rate.

If a provider cannot answer any of these questions cleanly, they cannot survive the 2026 filter:

• What is the minimum reviewer account age you use?
• Do you use residential IPs and what's the geographic diversity?
• Is your review text human-written or LLM-generated?
• What is your written replacement guarantee if reviews filter?
• How do you handle profile suspension if it occurs?

Our answers, in order: 12+ months, yes and matched to target region, human-written by our team, 30-day free replacement, and we run a suspension recovery workflow at no cost when it happens on our orders. That is what the market looks like now.

What to expect next

Based on the direction of the filter changes, three things are likely in the next 12 months.

1. Tighter AI-text detection. The classifier will get better and threshold-lower. LLM-generated bulk reviews will approach 100% filter rate by late 2026.
2. Cross-platform identity checks. Google is likely to weight signals from Google Maps location history, Android device signals, and Chrome browsing history more heavily to verify reviewer authenticity for logged-in accounts.
3. Faster reporting cycles. The Feb 2026 change moved reporting to 3 to 7 day resolution. That will likely continue to compress toward 24 to 48 hours.

The strategic implication is clear: reviews that look and behave like real customer reviews - because they were written by humans, from real accounts, at natural pacing, with photos - are the only reviews that survive long-term. Everything else is on a shrinking runway.

Frequently Asked Questions

What are the key changes in Google's 2026 review filter?

Google's 2026 review filter brought three main changes: tighter velocity thresholds (especially for smaller businesses), stricter IP and device clustering detection to combat review farms, and enhanced machine learning models for identifying AI-generated review content. These updates aim to improve review authenticity.

How many reviews can a Google Business Profile now receive daily without filtering?

For most small to medium businesses (under 200 reviews), the safe daily limit is now 2 to 3 reviews. Larger profiles with over 500 reviews can still absorb 5 to 7 daily without significant filtering. Exceeding these limits for multiple days increases the soft-filter rate.

How does Google detect AI-generated reviews now?

Google retrained its machine learning models using 2025 review data to specifically identify language patterns characteristic of AI platforms like GPT and Gemini. Reviews scoring above 0.7 on Google's internal AI likelihood classifier are more likely to be down-ranked or removed due to suspected AI generation.

What is IP and device clustering, and how has it changed?

IP and device clustering refers to Google's method of identifying reviews originating from the same network or device. The 2026 update expanded this to include browser canvas hashes, WebGL renderer, timezone/language mismatches, and reused mobile device identifiers, making it harder for review farms to bypass filters.

What types of reviews are most likely to be filtered by the new system?

Reviews most likely to be filtered include those posted in bulk, from a small pool of IP addresses, or those exhibiting patterns of AI-generated text. Any sudden surge in reviews or reviews from accounts with unusual geographic or industry review histories are also at higher risk of filtering.