Fighting Fake Reviews
Kindus has already discussed the UK Digital Markets, Competition and Consumers Bill which includes provisions to ensure that sites take reasonable steps to ensure that posted reviews are genuine. While the act will penalise those that do not comply exactly how they are supposed to enforce these rules is less clear.
Sellers such as Amazon gain no benefit from fake reviews, they affect the reliability of the whole review engine and create more cases of having to collect the returned goods. Amazon already claims to have taken down over 23,000 groups sourcing fake reviews in 2022. Other ‘independent’ on-line review sites might rely on paid reviews as a key income source.
In June 2023 Amazon announced that it would be using AI as part of its regime to crack down on fake reviews. Factors that are considered include the relationship of the author with the purchaser and seller together with the buying habits of the purchaser. On the other hand AI text generators could be a source of more believable false reviews. Also a business model has evolved of ‘fake’ reviews from genuine users which if well-written could be hard to pick up on.
Fake review supplyers have been coordinated through Facebook accounts. An anonymous user claimed to have received goods for free in exchange for writing complimentary reviews. The user worked with 10 or more review brokers and received a list of about 100 possible goods from each. This gave a reasonable chance of having some item that the user might want and be able to give meaningful (if false) feedback on. The user would be re-paid the cost of the item; it is unclear if any additional payments took place. There is little risk as even if the item cost is not paid by the agent the goods can still be returned for a refund. The work involved in creating a short review would be insignificant.
The model works best for relatively cheap goods or services where the product cost is easily outweighed by the publicity benefit. It would be less attractive for a more expensive example such as a holiday letting. The model is similar to sending out review samples with the additional caveat that any review will be positive. Costs to the supplier are claimed to be in the region of 100 reviews for $180. Ideally this would be combined with writing off a genuine purchase to ensure that the buyer and reviewer can be linked. Review agent marketing stresses that each poor review needs a large number of good reviews to balance it out further pushing the reliance on fake reviews.
Some of the review suppliers act openly in a grey area as product testing sites. Amazon has taken legal action against these entities. A customer review of one such entity which Amazon did act against indicates that they do pay out to the reviewer but that many of the supplied goods are of low quality and that the final ‘rebate’ may not include postage or PayPal fees so the final sum received does not fully cover the costs of the reviewer. It is hard to see the attraction of this model as the reviewer may be getting a deep discount but for relatively low quality goods that they probably have no need for.
The genuine buyer should always be wary of on-line review systems although a majority of low ratings is probably a bad sign. On Amazon there is a clue to fakes from clicking on the name of the reviewer. The host country, quantity and type of items reviewed might flag up warning signs. A genuine review is more likely to be longer than a fake and incorporate new rather than stock images. The paid review writer would have minimised the time and effort in creating the content to maximise their return although AI text generation may change that trend.
