The Unofficial TripAdvisor Obituary RIP

by Robert McGarvey

I have long been something of a fan of TripAdvisor, especially when it comes to booking at an independent hotel. Yes, my confidence in it was rocked when it became evident that the site had deleted posts claiming sexual assaults by hotel employees. And then there were the reports of hoteliers offering bribes to guests who had posted cranky reviews if they took them down.

And probably the biggest problem in recent years has been a growing army of scribblers who for a dollar or three write up favorable reviews of hotels they have never stayed in – never even seen in person. But, frankly, this hasn’t bothered me as much as it has bothered others because I thought – possibly stupidly – that I could spot such reviews. The language was usually stilted, the praises were generic, and – always – they were too favorable.

I have long believed it is normal to write a cranky review of a place that disappointed – while you are waiting, maybe an hour, for room service to bring your morning coffee why not tap out a blistering write up?

When I see positive reviews I have been skeptical – but some have convinced me they are honest. I had persuaded myself I could spot the phonies and read the honest ones.

But there now is a new player in town and it will put all the scribblers out of work because it is AI. The Guardian has the headline: Fake reviews: can we trust what we read online as use of AI explodes?

The Guardian’s thesis: “Artificial intelligence produces plausible verdicts on hotels, restaurants and tech in an instant.”

You doubt it? So did I I. So I went to the free version of ChatGPT – https://chat.openai.com/ – and asked it for a review of Enchantment resort in Sedona, a property I know and have followed for a couple decades. It came back with a long review that included this: “Enchantment Resort offers a truly enchanting experience in the heart of Sedona. The combination of stunning natural beauty, luxurious accommodations, superb dining, and exceptional service creates a memorable and rejuvenating getaway. Whether you’re seeking relaxation, adventure, or a spiritual retreat, Enchantment Resort is an ideal choice.”

The review is written in first person and – notably – it included that the easy access to hiking trails from the resort is a real plus. That is true.

Was there anything false? I don’t think it’s in the “heart” of Sedona – it’s out of town – but, really, the review was otherwise accurate.

Duplicate this experience yourself. Here’s the prompt I used – write a review of enchantment resort in sedona. (With large language model AI tools, they kick into action when given a prompt, kind of an AI variant a Google search command. Good prompting takes practice. If you don’t like what you get back, try another prompt.)

The ChatGPT review worried me. So I went to Google’s Bard – https://bard.google.com/ – and gave it the same prompt.

Bard came back with what I judged a better, even more credible review. Again it’s written in the first person and it even included a list of Cons:

• Expensive
• Some of the amenities, such as the spa and golf course, can be costly
• The resort is located in a secluded area, so it can be a bit of a drive to get to Sedona’s main attractions

Bard concluded: “Overall, I would highly recommend the Enchantment Resort to anyone looking for a luxurious and relaxing vacation in Sedona. It is a bit pricey, but the experience is worth it.”

How will we spot these AI generated reviews? Probably we won’t. The Guardian’s reporter, Patrick Collinson, observed that before we could spot fake reviews by their bad English, today “one sign that a review is fake will be that the sentence structure is a bit too perfect.” How are your Chicago Manual skills these days?

Meantime, TripAdvisor et. al. had been fighting trench wars with bogus posters – Collison said it identified 1.3 million fake reviews in 2022, a year that I’d guess saw fewer reviews than most because who was traveling then?

A scary reality is that Bard and ChatGPT spit out those reviews in seconds.

TripAdvisor isn’t clueless about this. It notes in a recent report, “As with all new technologies, the benefits of generative AI also come with challenges, and we expect to see attempts from businesses and individuals to use tools like ChatGPT to manipulate content on Tripadvisor. Our Trust & Safety team will continue to monitor the use of these tools on the platform and will take all available steps to stay ahead of threats to Tripadvisor’s brand integrity.”

Good luck with that. Arguably, AI tools could be trained to spot AI generated content – a kind of real world Terminator battle – but, for now, I am swearing off looking at “user” reviews of anything, from Amazon products to hotels and restaurants.

The machines are winning this battle.