At least 78,000 customer reviews across 14 popular tech products have been removed from Amazon in the past three years, an investigation by Which? has found.
After finding thousands of potentially fake or suspicious reviews in our analysis earlier this month, Which? worked with online shopping analysts ReviewMeta to build a clearer picture of the issue. We discovered:
- The popular tech product categories that may be more risky to shop in
- The true scale of a proliferation of ‘unknown’ brands on Amazon
- A massive rise in the number of unverified and potentially fake reviews in March this year
Tech products with the most removed reviews on Amazon
We analysed data collected by ReviewMeta to study activity on Amazon across a range of popular tech products over three years – and many of the findings matched those in our recent investigation. The graph below shows the percentage of reviews removed from each category in that period.
In our previous investigation, headphones were the most heavily affected product across the categories analysed. We uncovered more than 10,000 unverified reviews across just 24 pairs of headphones that appeared on the first page of a search when sorted by average customer review on Amazon UK – all of which were from ‘unknown’ brands.
We also found a high number of smartwatches, activity trackers, compact cameras and dash cams from unknown brands, with perfect five-star ratings and a number of unverified reviews – many of which showed the same suspicious activity.
Review removal could indicate that product categories are being targeted by fake reviews to artificially inflate customer scores and rankings – potentially placing products with a lot of five star reviews higher up in search results when sorted by average customer score, or within Amazon’s featured reviews.
The rise of ‘unknown’ brands at Amazon
For the purpose of our investigation, we classed an unknown brand as one that our tech experts hadn’t heard of. Our research showed that Amazon is awash with unfamiliar faces on the first page of popular tech product categories.
We looked at 2,663 brands that are selling on Amazon across 14 popular tech product categories, and found just 256 familiar brands. Almost 9 in 10 were unknown.
Should you avoid buying unknown brands?
Buying from brands you haven’t heard of isn’t always risky, but you’d be wise to take extra care. Further analysis revealed that the number of reviews removed from Amazon across unknown and known brands differs significantly. Less than 5% of reviews that were removed were left on products from known brands, while 18% of removed reviews came from unknown brands.
Not only that, but the number of verified reviews was far lower for unknown brands – 77.3% compared to 89.5% for familiar brands.
These figures suggest that lesser-known brands may be more likely to rely on ‘fake’ or incentivised reviews to stand out to customers. Our research would certainly indicate that this is one way some of these brands are becoming top-rated products.
Are fake reviews on the rise?
According to data from ReviewMeta, the percentage of unverified reviews on products has increased significantly. In the first quarter of 2018, just 6% of reviews were unverified, but for the same period in 2019 this had risen to a whopping 31% – leaving just 69% of reviews from verified customers.
Shockingly, in March 2019 the number of unverified reviews on Amazon rose by nearly 300% compared with the previous month. And the average star rating of unverified reviews in March? An almost perfect 4.95 out of 5.
With so many positive unverified reviews, it’s not a stretch to assume that this activity bears all the hallmarks of a ‘push’ for fake reviews to artificially boost the ratings of products.
What did Amazon say?
Amazon issued the following statement in response to our recent investigation into fake reviews:
Amazon invests significant resources to protect the integrity of reviews in our store because we know customers value the insights and experiences shared by fellow shoppers. Even one inauthentic review is one too many. We have clear participation guidelines for both reviewers and selling partners and we suspend, ban, and take legal action on those who violate our policies.
We use a combination of teams of investigators and automated technology to prevent and detect inauthentic reviews at scale, and to take action against the bad actors behind the abuse. We estimate more than 90% of inauthentic reviews are computer generated, and we use machine learning technology to analyse all incoming and existing reviews 24/7 and block or remove inauthentic reviews. Our team investigates suspect reviews, works with social media sites to stop inauthentic reviews at the source, pursues legal action to stop offenders from planning reviews abuse, and feeds new information into our automated systems so it continues to improve and become more effective in catching abuse.
We work hard to enrich the shopping experience for our customers [and selling partners] with authentic reviews written by real customers. Customers can help by reporting any requests they get to manipulate reviews to customer service.
It seems clear that Amazon is taking measures to clamp down on fake reviews. In fact, since we reported our findings to Amazon, many of the products or suspect reviews that we found in our investigation have been removed.
Unfortunately, more have immediately taken their place.
Natalie Hitchins, Which? Head of Home Products and Services, said:
Our research shows that while thousands of potentially fake reviews are being removed from Amazon each month, this isn’t nearly enough to address a real problem for online shoppers that seems to be getting worse, not better.
To avoid being tricked into buying a product that you might otherwise have avoided, watch out for fake reviews and search for independent and trusted sources when looking to make a purchase.
Where do fake reviews come from?
The scale of the problem with fake reviews suggests that it’s big business, but where do suspicious reviews actually originate?
In our 2018 investigation into fake reviews we went undercover, and found Facebook Groups with tens of thousands of members designed to generate incentivised positive reviews for Amazon product purchases. Members of these groups are instructed to purchase the product from Amazon and review it, before they are refunded or in some cases, given additional compensation. This means that the reviews that appear on Amazon are ‘verified purchases’.
Facebook told us that facilitating or encouraging the trade of fake user reviews is not permitted on the social media platform.
But in a recent follow up to the investigation, we found that four of the five groups we originally found were still active – three of which included the word ‘Amazon’ in the group name. Between them, they had nearly 70,000 members. We also uncovered a newer Facebook group that appears to be a reviews group, that at the time of writing has 15,180 members.
We contacted Facebook with our recent findings, and it investigated and removed the four groups that we flagged as incentivised review groups. It told us:
We don’t allow people on Facebook to facilitate or encourage the trade of fake user reviews. The groups brought to our attention have now been removed for violating our policies. We urge people to continue to use our reporting tools to flag content they think break our rules so that we can investigate and take swift action.
When we asked Which? members to tell us about their experiences with fake reviews, several reported that they’d been approached via email by review ‘clubs’.
One of these groups offered ‘VIP membership’ to product testing for free product giveaways, or discounted prices. Another member told us they’d been contacted out of the blue with an invite to a free product review program.
These review clubs seem to work in a similar way to the Facebook groups – they can generate verified reviews by asking people to purchase products from Amazon or other online retailers, and then refund the cost once they’ve received proof of a positive review.
Amazon told us that it estimates more than 90% of inauthentic reviews are computer generated. It uses machine learning technology to analyse all incoming and existing reviews 24/7 and block or remove inauthentic reviews – but as we’ve discovered, this isn’t always effective.
Automated reviews can be sophisticated and difficult to spot, but sometimes they’re obvious. In our previous investigation we found a product with over 400 unverified reviews, all five-star, and all left on the same day. It was also easy to see repetition and duplication among the reviews.
For more tips, read our guide on how to spot a fake review.
Help us fight fake reviews
With so many products available to buy online, customer reviews are an important way to help narrow down your options, and make a wise purchase of a product that will last. Fake reviews muddy the waters, and mean you may not be getting what you expected.
If you’ve come across online customer reviews you think are suspicious and may be fake, have been approached to write an incentivised positive review for a product – perhaps with the offer of a refund or free gift – or are asked to join a review club, you can help by sharing your experience.
Get in touch and tell us your story – email the Which? Fake Reviews team at firstname.lastname@example.org.
Additional reporting: Josh Robbins