With the help of an AI application, Amazon wants to make purchasing decisions easier for its customers in the future. But the new app feature may not be as useful as the company claims.
It is all too easy to get lost in the sheer endless expanse of Amazon’s shopping offering. If you want to equip yourself with a new power bank or gaming mouse, you are spoiled for choice between dozens, if not hundreds, of models and brands. The experiences of other buyers can be a real decision-making aid. Amazon now wants to expand this concept of customer reviews with an AI application. And once again the question arises: does the use of AI make things better or much, much worse?
Amazon’s trouble with the fake reviews
Amazon itself looks somewhat glorified at its system of customer reviews. When the company introduced the review function in 1995, the concept of buying a product based on the opinions and experiences of previous buyers was still relatively new. Hard to imagine today, after all, you can now rate almost anything and on all sorts of platforms. Amazon customers also make good use of it. In 2022 alone, they delivered 1.5 billion reviews and ratings worldwide.
At the same time, Amazon would have to admit that the star of online ratings has been falling since ChatGPT and Co. at the latest. Because the star ratings have long since become a lucrative business, which is constantly crumbling the customers’ trust in the reviews. Amazon has been struggling with manipulated or simply fake product reviews for a long time. In 2020 alone, the group blocked 200 million fake reviews.
While some dishonest sellers buy masses of positive reviews, some members of the Vine Club product testing group regularly post reviews without actually testing the products. Reviews written by AIs like ChatGPT have also been increasing since 2022. These are not helpful for Amazon customers either, but they can at least make you smile. CNBC has collected some amusing example reviews, all in a similar vein: “As an AI language model, I may not have a body, but I know the importance of comfortable clothing during pregnancy.”
Also interesting: What does “verified purchase” actually mean on Amazon?
AI analyzes customer reviews
Although ChatGPT and Co. could cause some trouble for Amazon, the group is now relying on AI itself to better support customers in their purchasing decisions, as stated in a statement. To do this, the AI should analyze the reviews of a product and then present a short summary right next to the product details. This extract of opinions is intended to provide customers with a quick overview, but its content is more meaningful than the five stars.
In addition, the AI works out keywords that are mentioned particularly frequently in the reviews, such as “ease of use”. These keywords should then act as a button and lead to the reviews that address this specific aspect. Here, too, the idea is that undecided customers can quickly find the review that, according to the AI, will help them the most. So far, this AI extension is only available in the Amazon app in the US and is still in the testing phase there as well. This means that only a few users can access the function so far.
No AI without criticism
Although Amazon says it only has the best in mind for its customers, some points of criticism immediately catch the eye. First and foremost, as with all algorithms of this type, the AI is only as good as the data it uses – i.e. the customer reviews. And as has been shown, a not inconsiderable part of the reviews are manipulated, falsified or even invented by an AI.
Amazon claims to take action against this, but ultimately these measures remain a black box for customers. If you read individual reviews yourself, you can at least recognize the obviously inauthentic ones. With an AI-generated summary, however, this self-assessment is omitted.
New feature could hurt sellers
In addition, there are concerns that the AI could be influenced by Amazon and then, for example, concentrate primarily on positive aspects in the reviews in the analysis. Finally, as a marketplace, Amazon benefits from the purchases. But there is also the other way around first cases, in which the AI highlights in its summary the “opinions of the majority” that do not match the average star rating. If the AI highlights negative reviews, even though most reviews are positive, this can cause massive damage to the seller. On the other hand, if products are falsely presented too positively, the AI completely undermines the purpose of the customer reviews. So as long as such discrepancies exist, customers and sellers are more at risk of disadvantages from the AI application. Instead of relying on them, customers now have to spend extra time verifying the AI summary.
This also applies to the keywords that Amazon’s AI extracts from customer reviews. Because many of the authentic reviews come from laypeople who – in the spirit of the matter – share their personal experiences. But while one reviewer of a gaming mouse may place great value on aesthetics and lighting effects, the prospective buyer may be more interested in the size or the ergonomic design – which may or may not fit depending on the size of the hand. Here, too, the promise that the AI will filter out the most important aspects is difficult to keep, as many products have very different customer needs. It remains to be seen whether the AI will really be of any help to Amazon’s customers after it has been fine-tuned in the test phase.