Google’s old idiom that every picture tells a story causes concern. With all the world’s graphic engineering devices at its disposal, the tech giant is taking a deeper, under-the-cover look at the credibility of the visual media that appears on Google Images to make sure every photo tells the same story.
It takes almost two years to get the message that consumers are skeptical of online photos, but Google immediately implements fact checks in its image-posting process so people are aware of any issues and can make more informed decisions.
Look for mark “Reality Check”
Going forward, when someone searches Google Images, they may see a “Fact Check” label under the picture results. Tapping on that label gives the user a quick overview of what Google found in its fact-check, both for specific images and articles with an allegedly fake image.
Google claims it also checks facts on daily search and in Google News, but the company depends on reports from ClaimReview, an open-source tool that Google also uses for YouTube and one that web publishers use to identify fact-check material to search engines.
“Photos and videos are an incredible way to help people understand the world. But the power of visual media has its drawbacks — particularly when there are concerns about an image’s origin, credibility or meaning, “commented Harris Cohen, Google’s Search Group Product Manager.
Fact-checking is rising
Google isn’t the first — and certainly won’t be the last — to employ fact-checking. Weeding out fake news, videos and photos has become a must for every Big Tech member who wants to stay in good standing with their user base.
There’s no company as proactive as Facebook, though. The social media platform bumped its fact-checking after the 2016 election and went on a tirade earlier this year to bust coronavirus myths.
YouTube CEO Susan Wojcicki noted in a recent Washington Post Live interview that the company yanks videos that violate its policies, including hate speech, inciting violence, or any type of manipulated media that could cause disinformation.
Wojcicki said it doesn’t matter if those doctored videos are from a politician or anyone else, but she said the new rules of YouTube keep some of the videos available if they are presented in context, by news or for educational purposes.
False reviews are a growing concern for customers and the places they depend on to make buying decisions. We know from numerous studies that 90% of internet users check online reviews before purchasing.
The size of the issue of fake reviews isn’t precisely known, but it’s “in the millions” across platforms, according to Curtis Boyd, whose company Objection.co helps companies recognize and delete false online reviews.
Fake Review Types
There are five main sources or styles of false reviews Boyd sees:
- Vendors offering both positive and negative online reviews.
- Directly or indirectly, business owners create false feedback (through fake profiles or vendors).
- Current employees write positive reviews for an employer.
- Ex-employees write negative retaliatory reviews for termination or lay-off.
- Customers misleading or exaggerating a negative experience to get a refund or other gain (e.g. discount).
- Review clusters (e.g., friends and family) compose positive or critical reviews within a short period.
The most common kind of inauthentic review is from a business owner, using a fake profile. After that, according to Boyd, it’s reviews generated by fake review vendors, typically operating offshore in China, India, Bangladesh or the Philippines.
Yelp, TripAdvisor More Effective at Moderating Boyd explained that Yelp and TripAdvisor generally are much better at moderating content and minimizing fake reviews, compared to Google or Amazon, but they still make it through. “There’s no review website that doesn’t have fake reviews,” Boyd explains. “The problem is everywhere and there are varying levels of sophistication.” “There are tens of thousands of [Google] Local Guide profiles owned by review farms; they’ve done a good job of bypassing Google’s detection systems.” A large percentage of these profiles are not real people and they generate huge volumes of fraudulent reviews. The screenshot above reflects a fake profile (Local Guide Level 4) which is part of a “review pod,” according to Boyd.
Sophisticated Cat and Mouse Game Objection.co uses a range of AI and machine-learning techniques to identify fake reviews, including looking at the geographic location of reviewers, review patterns and time frames, relationships between review profiles, review syntax, character count and lexical diversity among other methods. Boyd also buys fake review content and profiles en masse from vendors — who willingly sell out their content and writers. “These are mostly small companies in developing countries and they just want the money.” Boyd has thus been able to buy 280,000 fraudulent profiles and millions of fake reviews and compared them with real profiles and reviews, to train his systems on patterns and data relationships.
We also discussed the cost of fake reviews and Boyd quickly ticked off some prices:
- $50 for fake Yelp review “that sticks.”
- $5 for a fake review on Google.
- $10-15 for TripAdvisor.
Boyd observes that, “When [business owners] start purchasing fake reviews they can get addicted.” It’s often true, in these cases, that “more than 50 percent of their reviews are fake.” Amazon’s ‘Very Bad’ Problem Boyd says the problem is particularly bad on Amazon. “Amazon as a whole is suffering from a fake review profile problem that’s very bad.” He points out, for example, that about 90 percent of Bluetooth-enabled products in the electronics category on Amazon are fake. “Books is another bad one,” with self-published authors generating fake reviews for themselves.
A partial list of other Amazon product categories filled with inauthentic feedback include:
- Weight-loss products.
- Home’s health.
- Kitchen products (for off-brand products).
Many third parties found additional Amazon categories with fake reviews. Boyd further describes a disturbing situation on Amazon where legitimate brands are losing position and market share because of fake reviews. “Some products are losing their bestseller status to fake-review products.” While Amazon has periodically taken action against fake reviews, Boyd speculates the company is unmotivated to become too aggressive, because reviews (real or fake) sell products and Amazon continues to benefit directly and indirectly from these reviews.
Consumers Just Starting to Wake Up Consumers are becoming increasingly aware of the fake reviews problem, though they don’t understand its full extent. They ‘re unhappy about potentially being deceived and are looking to publishers to institute tougher standards and penalties against review fraud.
According to Jebbit’s Consumer Trust Data Index, out of 100 companies Amazon is the brand most trusted by U.S. consumers. If they were fully aware of the site’s fake review issue, Amazon’s reputation could suffer accordingly.
Also, if Google, Amazon and other sites do not crack down more aggressively on review fraud consumers may slowly return to more traditional methods of finding business and product recommendations: word of mouth and expert reviews.