Fake comments about travelers have been a problem for travelers since the early days of the Internet. But now that sketchy businesses can use artificial intelligence to make fake reviews faster than Taco Bell, things are getting worse quickly.
Companies analyzing consumer reviews transparent company estimates 3% of all business units’ comments It is generated by AI in 2024. This may not sound like much, but the company reports that fake AI reviews have been growing 80% One month since June 2023.
This means fake AI reviews almost double every 30 days. Doing math, you will find that a meal that was reviewed in a great restaurant in Barcelona was actually fed Chatgpt’s imagination shortly after.
I’ve worked in the field of generative AI over the past year. I learned a lot about how its digital brain works and how to discover its Telltale Tics and tell. I also looked at the work of studying fake reviews and knowing how to smell fraudulent Linger.
Red flags for AI-generated comments
Here is a guide to guess whether the review you are reading might be a product of artificial intelligence or a guide written by real, cute flawed people.
Empty Carnival: This is the simplest AI telling. Generative AI is essentially a complex, automated completion machine that draws on a lot of data to guess what the next word in a sentence might be. Instructing AI to praise a restaurant or hotel, it will immediately radiate the most common clichés it finds in other positive reviews of that restaurant and all restaurants. The food is “perfect”, the service is “on site”, the overall experience “OMG is the best ever!” – all of which have no specific details to support anything. Meaningless enthusiasm (or for that matter, because sometimes bad reviews of competitors’ competitors) is AI’s biggest giveaway.
Big cliché: Pangram LaboratoryAI text detectors, identified in the Transparency Company report, the following specific “AI tell”: When reviewers say things like “the first thing that shocked me,” “changing game changers” and “make their promises”, the warning ringtone should go away.
Phrases people usually don’t use: I found that the last paragraph from “Summary” was a dead giveaway. Who wrote this? Seventh graders write papers? There may not be anyone reviewing hotels with noisy swimming pools in Kissimmee. (Terms like “really” and “more” are also great, like who is.) In my own legitimate work with AI (creation documents for organizational development), I have to specifically guide it to avoid listings of those rare cowhide words and phrases.
tl;DR: If the review is read too long, a machine may write it. Most human reviewers take up to a few minutes to comment. The machine can issue 700 words in a few seconds. Although there are some real review amateurs, they are exceptions and you can easily identify them. This brings us to the next AI giveaway…
Rookie Commenter: AI reviews are often “written” by profiles created by AI, and these profiles have few other reviews to earn their credit. Sometimes, fake accounts only post that comment. Fortunately, most review platforms contain information about the contributions made by reviewers. A good bet is that “Amy9437” (a numeric suffix like this is also a good AI signal), he only has this 5-star curbside motel orgy that is unlikely to be her name, which is the spawning of the dark machine. But if the reviewer has written about 12 restaurants and 10 hotels in the past two years, you are more likely to believe her comments.
A+ Ingredients: In comparing the outputs of various AI models, I found that the characteristics of the prose they produce were significantly different. But one thing is consistent: they all magically write grammatically correct sentences that are arranged in coherent paragraphs, and sometimes they add bullet points. Most people are not. Spelling and grammatical errors may not be played in high school English classes, but online, they tell you a real person who got a C+ from Mrs. Herbison and is telling you about that underground brewery tour in Cincinnati.
Realistic details: The AI didn’t know that Tasha, behind the front desk, was helping the reviewer’s dog stayed for 5 minutes while going to the rooftop bar to talk to his wife, and it was unlikely that AI would constitute such a story when the task was writing hotel reviews. When you see a similar specific vivid event, you may be more confident that it is real.
The future of fake comments generated by AI
So, hopefully things will improve, don’t you have to be so alert in the future?
In August 2024, the Federal Trade Commission finally confirmed Making making and selling fake reviews illegalincluding fake AI comments. It even sued A popular creator of fake AI recommendations.
It is not clear whether the Trump administration will be enthusiastic about consumer protection.
In the private sector, an international industrial group is called Trusted Comment AllianceTravel companies including TripAdvisor, Booking.com and Expedia Group (including Hotels.com, Travelocity, Orbitz, VRBO and other brands) were established in 2023, which included sharing best practices and regulatory approaches to combat review fraud, but no one claimed to completely clean their Junk Impless reviews.
As we continue to pay close attention to the development of AI, there are two opposing forces worth a look.
On the one hand, including Google and TripAdvisor There are well-funded, technically sophisticated efforts aimed at retaining a variety of fake comments, including those written by AI, on their website. Of course, these programs use AI to police forgery of comments.
On the other hand, you have a cunning, resilient global fraud network that studies the defense capabilities of the platform and is always developing new ways to sneak around.
This established a global AI arms race similar to cybersecurity, with black hats and white hats in constant battles.
at the same time? You can try inserting comments Pangram Labs detection tools See if it smells like a mouse.
There will always be an old-fashioned recommendation engine: word of mouth. Real-life family and friends you’ve been to may still be your best source for traveling Intel.
Editor’s Note: You can also come to Frommer’s. We do not use AI to create travel information. Since 1957, we have been written by humans.
Craig Stoltzformer travel editor Washington Postspent a year working with the U.S. federal government’s generative AI.