Consumers Trust Amazon Reviews More Than Ever. But Should They?

It's around 7PM, and I'm conducting a focus group with consumers - one of many that I've committed to between the end of 2023 and the beginning of 2024. As usual, the conversation goes off on fun tangents that reflect the personalities in the session, but now we've reached that predictable point where I'm asking consumers how they learn about products and brands. I brace myself, hoping to hear something new. But it's always the same two answers first.

"Google," one of them says. "It gets me wherever I need to go."

"Amazon reviews," says the next. "They sell everything. I figure if there's a problem, someone will mention it there."

The first answer has been a stock answer in product and brand research for quite some time, because pretty much everyone uses Google (or their preferred search engine that behaves just like it) for any question that they have. When I push, consumers often articulate how they like to be in the driver's seat, locating information for themselves and making evaluations based on their research. Granted, their research doesn't often seem too deep (and a surprising number of people consider reading what a product company or retailer has to say about a product sufficient for their purposes), but it provides them with a sense of doing their due diligence - ensuring that the product or brand doesn't have any major issues that should be obvious to anyone.

But the second answer concerns me much more, because I've been a customer of Amazon's since the early days when the website just sold books, and I've watched the site's product reviews grow from a fun new feature to an essential service to an opportunity for sellers to recover dissatisfied customers to the nearly useless state they're in today. The "everything store" has become a place where nearly every product seems to be well-rated and where searching for a product by category rather than a specific brand name can lead you down a deep well of nonsensical brands you've never heard of and product listings and reviews that are most definitely being generated by AI platforms.

That last point became quite clear this month when users started seeing listings where the actual product name for things like tables, kitchen chairs, sofas and paperback books included "I'm sorry I but I cannot fulfill that request it goes against OpenAI use policy." The problem was widespread enough that it made headlines around the internet before Amazon stamped it out, but it pointed to a major issue with the site - automated programs using generative AI are creating listings at such an alarming rate that human beings (both on the side of the sellers creating the listings and at Amazon itself) aren't even bothering to review the product names anymore.

Amazon's long had a problem with unreliable and outright fraudulent product listings (including ebooks that look like actual books, but which were written by generative AI), but the check to keep the balance has always been that customer reviews are in place to warn other buyers if a product is a scam and to provide not only a star rating for the product itself, but also detailed written content explaining the issues. Amazon's crackdown on various seller schemes to skirt the system has always been a game of whack-a-mole, targeting the worst and most rampant schemes to boost review scores, but 2023 saw a fresh new wave of fake reviews authored by generative AI thanks to the popularity of programs like ChatGPT. Most of these were initially easy to spot because they would include phrases like, "As an AI language model, I don't have a body, but..."

While fraudulent reviews have been a problem for awhile, journalists and industry professionals who have examined fake reviews on Amazon are constantly finding fraudulent reviews that are indicative of AI platforms. For example, a study from the UK optimization app company Circuit used the AI-powered Fakespot platform to detect AI reviews (I know, I know...) and found that 88% of clothes, shoes and jewelry listings had unreliable reviews, as did 53% of electronics listings and around a third of home and kitchen, beauty/personal care, pet supplies and sports & outdoor listings. It's not that hard to write fake reviews; aside from ChatGPT, you can also use free sites like ToolBaz, WriterHand or Tasklab.ai. What's more, there's a thriving market for writing these reviews, sometimes subcontracted by fly-by-night Chinese companies to users recruited on Facebook or other social platforms, but also offered by brokers who have thousands of "product testers" who can post reviews.

This is additionally alarming because Amazon itself has just launched an AI tool that allows users to not only see an auto-generated summary of all of the review content, but also to ask the AI tool questions about the product which are, presumably, going to be sourced from the user-generated content and not actual, reputable third-party reviews from outside websites. If the AI tool is consulting fake or fraudulent listing details and reviews to generate its summaries, it's likely to become a system that can be gamed by the same people who are clogging Amazon's search results with products from nonsensical brands like Mebakuk, JUCFHY or CAKKA. And while it's cute that the tool can respond in the form of a haiku or write a bedtime story or create a summary of the reviews in Master Yoda's voice, there is terrible potential for the tool to be exploited by sellers to provide dangerous misinformation about products.

Which takes me back to the consumers in my focus groups, who are not only incredibly trusting of the reviews they see on Amazon as providing objective crowdsourced information, but who also feel that this level of search is as far as they need to go. When I've probed to find out whether they are going to more authoritative reviews sites like Wirecutter or Tom's Guide or even stodgy standards like Consumer Reports, most are not and, what's more, don't feel the need to research their purchases that deeply unless they're planning on spending a lot of money. And even the most popular form of consumer research - word of mouth from friends and family, either in-person or virtually - is becoming secondary to DIY resources such as Google or Amazon reviews.

And that's not just an assumption on my part. According to 2023 research posted by PowerReviews (a user-generated content company), 99.5% of consumers research products online and 87% do it regularly. 50% of consumers start their search with Amazon reviews, and around nine in ten consumers eventually consult Amazon reviews (94%) or other retail website reviews (91%) in their search. The same research found that nearly half (45%) of consumers won't purchase a product without finding a review associated with it first. By contrast, only 60% of consumers rely on word of mouth for making a purchase decision.

As an insights researcher, I'm intrigued by the marketing possibilities for a world where online reviews are able to play an important role in helping democratize information and guide consumers to the best possible purchases. To marketers, I'd encourage ensuring that anything you're offering or selling has reviews available online; they definitely are going to play a role in the future of your business.

But as someone who's watched fake reviews flourish on Amazon and who's noticed the tremendous rise of cheap products, fake reviews and AI-generated listings over the last year, my advice to any consumer who trusts these reviews without any need for further information is to be careful, particularly with products that have batteries or electrical parts or which might contain dangerous chemicals. Those user reviews you're reading may not be real, and there's never a good way to find out the hard way.