The Revdict fake review dataset
Every Google Maps place in our public catalog, added up. These are live figures, updated hourly from the latest analysis of each place. For a figure that will not change, cite a dated report.
Updated 22 November 2026, 12:16 UTC · CC BY 4.0
Download
CSV · JSON — the aggregates on this page. No business-level rows, reviewer names or review texts.
Dated reports: State of Fake Reviews on Google Maps: 2026 H2
Places analyzed
995
Reviews examined
709,271
Reviews flagged
150,806 (21.3%)
Places not Clean
349 (35.1%)
45 countries. Average manipulation score 31.6/100 (median 30). The median place has 22.3% of its reviews flagged. Analyses run 2026-07-14 to 2026-11-22.
Verdict split
Latest verdict per place. What each verdict means is on the methodology page.
| Verdict | Places | Share | Bar |
|---|---|---|---|
| Clean | 646 | 64.9% | |
| Review Drive | 19 | 1.9% | |
| Suspicious | 290 | 29.1% | |
| Likely Manipulated | 40 | 4.0% |
Score distribution
Places per 10-point band of the 0–100 manipulation score.
| Score | Places | Bar |
|---|---|---|
| 0-9 | 187 | |
| 10-19 | 143 | |
| 20-29 | 155 | |
| 30-39 | 161 | |
| 40-49 | 155 | |
| 50-59 | 108 | |
| 60-69 | 46 | |
| 70-79 | 31 | |
| 80-89 | 9 | |
| 90-100 | 0 |
Top signals
Flagged reviews per signal. A review can carry more than one signal, so the rows add up to more than the flagged total.
| Signal | Reviews | Bar |
|---|---|---|
| Low-activity account | 133,237 | |
| Review burst | 22,717 | |
| Review drive | 3,818 | |
| Duplicate text | 739 | |
| Generic account | 532 |
By country
The 20 largest groups with at least 5 places.
| Country | Places | Reviews | Flagged | Avg score | Not Clean |
|---|---|---|---|---|---|
| United States | 494 | 455,656 | 17.3% | 24.3 | 20.6% |
| Poland | 65 | 27,656 | 20.6% | 35.8 | 35.4% |
| Indonesia | 59 | 24,800 | 33.2% | 46.1 | 69.5% |
| France | 55 | 22,889 | 21.8% | 35.4 | 38.2% |
| Germany | 50 | 21,039 | 19.4% | 29.3 | 26.0% |
| India | 43 | 25,460 | 48.1% | 52.6 | 86.0% |
| United Kingdom | 33 | 21,208 | 20.7% | 29.5 | 36.4% |
| Spain | 26 | 16,827 | 21.4% | 36.5 | 42.3% |
| Canada | 24 | 13,634 | 28.9% | 36.6 | 41.7% |
| Australia | 15 | 6,075 | 21.3% | 36.9 | 46.7% |
| Saudi Arabia | 12 | 5,704 | 19.7% | 33.2 | 41.7% |
| Türkiye | 12 | 7,837 | 42.5% | 49.7 | 83.3% |
| Greece | 10 | 8,239 | 20.5% | 28.9 | 20.0% |
| Vietnam | 10 | 5,812 | 30.0% | 43 | 60.0% |
| Thailand | 9 | 5,095 | 26.0% | 25.9 | 22.2% |
| Netherlands | 8 | 2,541 | 32.7% | 48.1 | 62.5% |
| Japan | 6 | 2,647 | 25.6% | 38.3 | 66.7% |
| Albania | 5 | 2,581 | 58.1% | 58.2 | 100.0% |
| Brazil | 5 | 3,705 | 35.2% | 46.8 | 60.0% |
| Pakistan | 5 | 4,040 | 51.6% | 54.8 | 100.0% |
By category
The 20 largest groups with at least 5 places.
| Category | Places | Reviews | Flagged | Avg score | Not Clean |
|---|---|---|---|---|---|
| Restaurant | 64 | 63,981 | 23.4% | 29.5 | 25.0% |
| Hamburger restaurant | 29 | 28,994 | 8.6% | 12.7 | 6.9% |
| American restaurant | 24 | 23,997 | 13.0% | 17.1 | 8.3% |
| Italian restaurant | 15 | 14,996 | 17.2% | 22.4 | 6.7% |
| Diner | 14 | 13,995 | 17.1% | 20.4 | 14.3% |
| Fast food restaurant | 14 | 13,996 | 8.9% | 13.7 | 7.1% |
| Pizza restaurant | 13 | 13,004 | 5.3% | 7.2 | 0.0% |
| Steak house | 11 | 10,998 | 23.6% | 26.3 | 18.2% |
| Hotel | 10 | 9,997 | 12.6% | 16.1 | 10.0% |
| Breakfast restaurant | 9 | 8,998 | 23.4% | 39.1 | 55.6% |
| Cuban restaurant | 9 | 8,999 | 15.6% | 19.2 | 11.1% |
| Mexican restaurant | 9 | 8,999 | 21.8% | 25.8 | 33.3% |
| Peruvian restaurant | 9 | 8,998 | 21.7% | 29.6 | 11.1% |
| Bakery | 8 | 7,999 | 7.4% | 11.1 | 0.0% |
| Brazilian restaurant | 8 | 7,998 | 30.3% | 38.4 | 50.0% |
| Shopping mall | 7 | 7,000 | 6.0% | 7.1 | 0.0% |
| Book store | 5 | 5,000 | 5.7% | 10.4 | 0.0% |
| Coffee shop | 5 | 4,998 | 16.8% | 19.8 | 20.0% |
| Ramen restaurant | 5 | 4,999 | 11.7% | 14.6 | 0.0% |
| Seafood restaurant | 5 | 5,000 | 13.7% | 19.2 | 0.0% |
By year the review was posted
Years with at least 1,000 reviews. Recent years are over-represented because each crawl reads the most recent reviews first. A rising flagged share is partly an artifact: the low-activity signal looks at the reviewer's account today, and the authors of recent reviews have had less time to post others.
| Year | Reviews | Flagged | Flagged share |
|---|---|---|---|
| 2017 | 2,748 | 372 | 13.5% |
| 2018 | 5,191 | 658 | 12.7% |
| 2019 | 10,270 | 1,360 | 13.2% |
| 2020 | 8,598 | 1,340 | 15.6% |
| 2021 | 13,995 | 1,861 | 13.3% |
| 2022 | 24,362 | 3,416 | 14.0% |
| 2023 | 51,854 | 7,692 | 14.8% |
| 2024 | 79,779 | 12,740 | 16.0% |
| 2025 | 197,669 | 38,619 | 19.5% |
| 2026 | 313,560 | 82,476 | 26.3% |
How to read these numbers
- Not a random sample. The places were submitted by visitors or added by us to cover popular venues, so they lean toward the United States, restaurants and places with many reviews. A share here describes this dataset, not all businesses on Google Maps.
- "Flagged" means at least one manipulation signal, not "fake". Most flags are low-activity accounts, and most first-time reviewers are genuine. The flag counts come from each place's latest analysis.
- A verdict is a statistical estimate, not a finding of fact. It does not say who posted a review or why. We have not measured a false-positive rate; quote that limitation with any figure.
- Recent reviews only. Each crawl reads up to about 1,000 reviews per place, most recent first, and cannot see reviews Google already removed.
The signals, weights and thresholds are documented in full on the methodology page.
How to cite
Free to reuse with attribution (CC BY 4.0). Live figures change; give the date you read them, or cite a dated report.
Revdict, “Revdict fake review dataset”, accessed 22 November 2026, https://revdict.com/data