We Studied 135 Deleted Google Reviews on 1 Profile. Here Is What They Had in Common
135 Google reviews vanished from 1 business profile in a single day. We had a copy of every one of them. So we read them all and looked for what they shared.
Key takeaways
- All 135 missing reviews were 4 or 5 stars. Not 1 was negative.
- 21 of them had no text at all. So the words in the reviews were not the reason.
- The reviews were not copies. They were written by real, different people.
- 85 percent of the text reviews named a staff member. Google’s rules do not allow a business to ask for that.
- 86 of the reviews came in 1 2-week window, often minutes apart, right after the tour ended.
- A few reviewer accounts were employees or other businesses. That is against Google’s rules too.
- 134 of the 135 went missing in the same second. That means the count must be checked before anyone calls it a deletion.
GBPPromote saves a copy of every review on a profile. When Google stops sending a review back, the copy stays. That gave us the full record for all 135: star rating, full text, date written, owner reply, and the exact second it went missing. This is 1 real profile in our own dashboard, not a survey.
The profile
The business is a horse and carriage tour company in a busy tourist city in the United States. It has tens of thousands of public reviews and a long history on Google. It is not a new listing and it is not a spam listing. Guides take groups on a 1-hour ride, tell stories about the city, and the ride ends back at the barn.
That matters. This is the kind of profile most owners think is safe.
Finding 1: Not 1 of the 135 reviews was negative
132 of the 135 reviews were 5 stars. The other 3 were 4 stars. There was no 3, 2 or 1-star review in the set.
Figure 1. Star rating of the 135 missing reviews
Count of reviews by rating
If you picked 135 reviews at random from a big profile, some would be low ratings. This set has none.
Whatever pulled these reviews out of the profile pulled positive reviews only.
Finding 2: 21 of the reviews had no text at all
21 reviews were a star rating and nothing else. No words, no names, no keywords.
Figure 2. What the 135 reviews contained
Reviews with text, and reviews with an owner reply
Reviews with text and without text both went missing. Reviews with a reply and without a reply both went missing. Neither one is the cause.
This changes how you should think about review removal. A review with no text cannot be removed for what it says. So the reason cannot be bad wording, AI writing or keyword stuffing. If there is a trigger, it sits outside the review text.
Finding 3: The reviews were not copies of each other
We compared all 114 text reviews with each other. There was no shared template. No sentence showed up again and again.
Figure 3. Review length
Number of reviews by word count (rounded groups)
The middle review is about 22 words. The shortest text review is 2 words. The longest is about 200. A template would give a tight cluster, not this spread.
The reviews had typos. They had emoji. They had complaints inside the praise: hard seats, a tour that was too short, drunk people in a passing car. Some reviewers could not remember the guide’s name and said so. That is what real customer writing looks like. Google did not remove these because they looked fake on the page.
Finding 4: 85 percent of the text reviews named an employee
This is the pattern that stood out most.
Figure 4. How many text reviews named a guide or a horse
Out of 114 reviews with text
- Named 2 or more (guide and horse): about 60 reviews, 53%
- Named 1: about 37 reviews, 32%
- Named no one: about 17 reviews, 15%
The same guide-and-horse pairs came up again and again from different customers.
Most reviews followed the same shape even though the words were different: name the guide and the horse, say they were amazing, say the guide knew the city’s history, say “highly recommend”.
Figure 5. Most repeated words and phrases
Number of text reviews (out of 114) that contain each term
We left the word “horse” out of this chart. It appeared in 36 reviews, but this is a horse carriage business, so that word tells us nothing. The other words do: different sentences, same talking points. That is what a prompted review looks like, even when every customer is real.
This does not prove the reviews are fake. It shows customers were writing about the same things, in the same order. That can happen on its own when everyone gets the same tour. It also happens when a business tells customers what to write.
Finding 5: 86 of the reviews came in a 2-week window
About 117 of the 135 reviews were written between July and early September. About 86 of them landed between 25 July and 9 August.
Figure 6. When the missing reviews were written
Number of reviews by year
13 reviews are from 2017 to 2025. 122 are from 2026, and 117 of those are from July to September.
Figure 7. The 2-week burst, day by day
Reviews per day, 29 July to 9 August 2026 (busiest stretch in the set)
13 reviews in 1 day, all later removed. A busy day on its own breaks no rule. A busy day plus the other patterns is what Google’s systems look at.
Reviews also came in small groups, a few minutes apart. Here are 3 real examples from the set.
Figure 8. Reviews posted minutes apart
3 groups from the set. Each group is people from the same tour reviewing right after the ride.
| Date | How many reviews | Time between first and last | What they shared |
|---|---|---|---|
| 5 September | 2 | 2 minutes 20 seconds | Both named the same 2 guides |
| 9 August | 3 | 9 minutes | All named the same guide. 2 reviewers had the same surname |
| 1 August | 4 | 6 minutes | Posted around midnight, right after evening tours ended |
Across the full set we counted about 18 pairs of reviews posted within 10 minutes of each other.
The simplest explanation is that customers were reviewing while still standing at the barn, often as a group, right after the ride.
Finding 6: A few reviewer accounts looked wrong
Most reviewers were ordinary people with ordinary names. A few were not.
- 2 reviews came from accounts named after other businesses, not people.
- 2 reviewer names matched the names of guides who work at the company.
- 1 person seems to have reviewed twice from 2 accounts, 2 days apart.
- About 10 accounts were blank or throwaway handles with no real name.
Google’s rules do not allow employees, family of employees, or business accounts to review the business. A handful of these inside a burst of reviews is enough to put the whole burst under review.
Finding 7: Owner replies had nothing to do with it
About 93 of the 135 reviews had a reply from the business. About 42 did not. Both groups went missing (see Figure 2). The replies were polite and varied. When a review goes, its reply goes with it. The reply is not the cause.
Summary of the 7 findings
132 five-star and 3 four-star. Zero low ratings. Whatever removed them targeted positive reviews.
A review with no words cannot be removed for its words. The cause is outside the text.
114 different reviews, 2 to 200 words, with typos and complaints. Real people wrote them.
Google bans asking customers to name staff. This is the strongest link to a rule.
86 reviews in 2 weeks, 13 in 1 day, many posted minutes apart while customers were still on site.
Business accounts, employees and a duplicate reviewer sat inside the burst.
Reviews with and without an owner reply went missing at the same rate.
What Google’s policy says
Google’s rules allow a business to ask for reviews. They ban asking for reviews that include specific content, and they name content that identifies a staff member as an example. They ban asking staff to collect a set number of reviews. They ban pushing customers to review while still on the premises. And they say reviews that show unusual rating patterns may be removed.
Put the findings next to those rules:
| Google bans | What we saw | See |
|---|---|---|
| Asking for content that names a staff member | 85 percent of text reviews named a guide | Figure 4 |
| Pushing customers to review on the premises | 18 pairs of reviews minutes apart, right after rides | Figure 8 |
| Staff review quotas | The same guides praised again and again in a 2-week burst | Figures 5, 7 |
| Employee or business-account reviews | Reviewer names matching guides, business accounts reviewing | Finding 6 |
| Unusual rating patterns | 132 of 135 were 5 stars | Figure 1 |
Every finding lines up with a rule. The words in the reviews were fine. The way the reviews were collected is what put them at risk. You can read the full policy on Google’s prohibited and restricted content page.
The finding that should worry every review tool
1 more data point. 134 of the 135 reviews were noticed missing at the exact same second. Not the same hour. The same second.
Figure 9. When the reviews were written vs when they went missing
Blue dots: years the reviews were written. Navy dot: the 1 second all 134 were found missing.
Reviews from 7 different years, 1 detection time. That pattern has 2 possible causes, and only 1 of them is Google.
Google does not look at a 2017 rating and a 2026 rating and remove both in the same instant. When a tracker sees that, it has to consider a second explanation: Google’s review API sent back a short list on that sync, and everything not on the list looked deleted.
For this profile, the 13 older reviews from 2017 to 2025 are the ones to check first. If they are still live, the true removal count is smaller and the story is about the July burst only. If they are gone too, the pattern above is the best explanation Google’s rules can give.
What to do if this is your business
- Stop asking customers to name the guide, driver or staff member. Ask for an honest review and leave it there.
- Never ask for 5 stars. Never give staff a review number to hit.
- Move the ask off the premises. Send the review link by email or text after the customer has left. The review generation tool in GBPPromote sends timed email sequences, which is the collection method Google’s rules are built for. If you still want a code at the counter, use a review QR code that does not steer wording or stars.
- Check whether any employees or their family have reviewed the business. Remove those reviews yourself.
- Watch review speed. 13 reviews in a day on a profile that normally gets 2 is a signal, even when every review is real. A free Google Business Profile audit shows your recent review pace next to your rating trend.
- Do not trust a “deleted” count from any tool until it has been checked twice.
If you run reviews for more than 1 location, the same checks apply to every profile you manage. GBPPromote’s GMB management software runs review tracking, review requests and profile audits across all locations from 1 dashboard, and it is included in every plan of our local SEO software for agencies.
Method
Source: 1 Google Business Profile managed inside GBPPromote. Data: the 135 reviews the platform stopped receiving from Google on 11 September 2026, with rating, text, date written, owner reply and detection time for each. Analysis: a manual read of all 135 reviews plus counts of rating, length, repeated terms, named staff, posting time gaps and reviewer account type. Counts marked “about” are rounded from a manual tally. No reviewer names, business name or review text are reproduced here.
Conclusion
135 positive reviews vanished from 1 profile. They were not copies. They were not written by AI. 21 of them had no words at all. What they shared was the way they were collected: a guide named in almost every review, reviews posted within minutes of the ride, a 2-week spike, and a few accounts that should never have reviewed at all. Each of those things sits next to a line in Google’s rules.
2 lessons come out of this. For business owners: the safest review is the one you asked for by email after the visit, with no script and no name to mention. For anyone who tracks reviews with software: a big number that appears in 1 second is a reason to check, not a reason to panic. GBPPromote keeps the copy, flags the spike, and lets you confirm before you act.