Fake-Review Signals Across 825 Places: Our Data, October 2026

Seven in ten places we have analyzed come back Clean. As of October 2026, 570 of 825 analyzed Google Maps places carry that verdict, 255 show at least one manipulation signal, and 28 reach Likely Manipulated. Our catalog has grown more than threefold since the July release, and the headline rate moved with it.

This is the second dated release from our own data. Every figure comes from completed analyses, not a survey, and the July numbers are the baseline.

#How many analyzed places show manipulation signals in October 2026?

About 31%. Of 825 places with a completed analysis, 255 sit outside the Clean band. The full split:

The July release listed three verdicts. The solicitation campaign verdict is a separate category for places whose flagged reviews look like staff steering real customers toward named five-star reviews, which is a different thing from buying fakes. The review drive explainer covers why we separate the two. The average score across all 825 places is 29.4 out of 100, down from 32.1 in July.

#Did the share of manipulated places go up or down since July?

Neither, strictly. The July release counted 242 places and found 33.5% with at least one signal and 12.4% Likely Manipulated.¹ Today 30.9% have a signal and 3.4% are Likely Manipulated. The absolute number of Likely Manipulated places is about the same (30 then, 28 now) while the denominator is 3.4 times larger.

That tells us the first catalog was not representative, and the second is only somewhat closer. The additions skew American, and 500 of the 859 places in the database (58%) are in the United States. Read the drop as a change in what we sampled. It is not evidence that manipulation fell by two thirds.

#What does the score distribution look like now?

In July the histogram had two humps: a big cluster at 10-19 and a second one at 60-79, holding 54 of 242 places. The second hump is now a tail. Places scoring 60 or above are 64 of 825 (7.8%), against 22% in July. The shape suggests that manipulation is concentrated in a minority of listings, and that adding more ordinary places dilutes the share without changing how extreme the extreme cases are. The ones at the top still score in the 80s.

#Which fake-review signal is most common?

Low-activity reviewer accounts, by a wide margin.

Of 618,257 reviews examined, 125,714 (20.3%) carry at least one flag. That share is identical to July's, even though the pool of reviews grew from 241,041. A single review can carry more than one signal, so the column does not sum to the flagged total.

Low-activity accounts account for 111,527 flags, roughly 89% of flagged reviews, up from about 76% in July. Our explainer on that signal says why it is both the most common and the weakest on its own: plenty of honest reviewers write two reviews a year. Duplicate text, once a headline signal, is now the rarest of the five at 480 flags.

#Where does the dataset lean?

Heavily American and heavily food. After the US, the largest countries are Poland (47 places), France (38), Germany (36), and the United Kingdom (29). Restaurants lead the categories with 69 places, followed by hamburger restaurants (44), American restaurants (25), and fast food, diners and Italian restaurants at 15 each.

Two consequences for anyone quoting these numbers. First, the verdict rates describe this catalog, not Google Maps as a whole. Second, the country and city comparisons we have published (by country and by US city) rest on small samples outside the US. You can browse every analyzed place, and the evidence behind each verdict, in the places catalog.

#What should you take from this?

Three things hold up across both releases. Most places look like what they claim to be. Roughly a fifth of individual reviews trip a flag, and almost all of those flags are thin reviewer histories rather than bursts or copied text. And the places at the high end of the scale stay rare and stay extreme, whatever the size of the catalog.

None of this says a given business bought reviews. A score describes patterns in public data. To see the evidence for a specific listing, paste the link into our free checker, and read the report rather than the verdict word alone.

#Methodology note

All figures are from completed analyses as of October 19, 2026. Each place counts once, using its latest completed analysis. Scores come from rule-based scorers over public review text, ratings, timestamps and reviewer profile data. Coverage is Google Maps only. The July figures quoted here are from the July release, and the earlier verdict counts there used a catalog of 242 places.

#Frequently asked questions

#Is 31% of Google Maps manipulated?

No. It is 31% of the 825 places in this catalog, which skews toward US restaurants. A rate for all of Google Maps would need a random sample of listings, and we do not have one.

#Why did the Likely Manipulated share fall from 12.4% to 3.4%?

The catalog grew from 242 to 825 places, and the added places skew toward the US. The count of Likely Manipulated places barely moved (30 to 28), so the denominator changed, not the places at the top.

#What is the solicitation campaign verdict?

It flags places where staff appear to steer real customers toward named five-star reviews in waves. It is a different pattern from buying fakes, and the reports label it separately.

#Which signal should I check first myself?

Reviewer history. Open a few reviewer profiles and see how many reviews each account has. Thin accounts were the most common flag in our data by far.