Tours and Activities Should Brace for Spike in AI-Generated Fake Reviews

Generative artificial intelligence like ChatGPT threatens the authenticity and trustworthiness of online reviews.  

Online travel agencies with a vested interest in their communities’ opinions and recommendations should note its potential to send fake reviews through the roof. This is according to travel industry consultant Peter Syme, who warns that the review model could be less valuable if AI review verification isn’t brought to book in some way.  

Review optimization, which sees companies paid to falsify reviews on behalf of a business to boost ratings and rankings, is a strong case in point.

Tripadvisor spokesperson Brian Hoyt said the impact of ChatGPT is on the company’s fake review radar. 

The company joined other leading review platform stakeholders to discuss the issues of review fraud in San Francisco towards the end of last year. Participants included the likes of Amazon, Expedia, Trustpilot and Yelp. This as 2022 closed off with explosive interest in generative AI that saw the OpenAI app race to a million users in under five days.

Hoyt said the intention was to share best practices to shut down fraudulent behavior, but he could not disclose any high-level outcomes of the meeting or why the group of stakeholders had not looked to meet sooner to address the issue of review fraud. 

“We’re always trying to stay one step ahead of the fraudsters. ChatGPT is a new technology that we’re looking at to ensure the efficacy of reviews on our platform,” said Hoyt. 

Review texts generated by ChatGPT carry a disclaimer, but they are unlikely to be labeled as such because it’s not a requirement on any platform. 

This is especially problematic for travel companies who rank experiences with algorithms based on real review sentiment.

Staggering Scale of Review Fraud

And this is something Tripadvisor as the biggest platform in user-generated reviews, with over 1 billion reviews, knows all too well — having fallen victim to review auditing chinks in its amour. 

While the company has yet to release its 2023 Review Transparency Report, expected later this year, its 2021 Review Transparency Report shows that at least 1 million reviews of some 26 million reviews submitted to the website during that year were fake.  

“About 67 percent of the fake reviews submitted to our platform never see the light of day. The other 33 percent were taken off the site throughout the year.” added Hoyt.

And while Tripadvisor’s own investigations resulted in a precedent-setting jail sentence in Italy, not all rules are created equal. Penalties can vary, with deliberately misleading marketing only subject to a fine in some jurisdictions.

External Vetting  

TUI Musement said its online reviews are managed and vetted through different technology partners, including Trustpilot, which provides certified reviews through a very strict vetting process. 

Reviews can only be done if a purchase has been completed. The company also monitors third-party products.

In-App Control

Similarly, GetYourGuide said its app ensures review credibility as only customers who have booked can leave a review. 

The company also use granular data from its reviews to gain “deeper insights into a specific activity. AI tools also undertake sentiment analysis to identify emerging trends influencing its search ranking algorithm.

Harnessing Sentiment Analysis

Broad-based sentiment, however, remains one bugbear for the founder of content channel management software company Magpie, Christian Watts. 

He acknowledged that while AI brings the risk of widespread fraud, it also holds the potential to improve services and positioning if the quality of reviews is well managed. 

After all, businesses nor users can sift through 30,000 reviews of one attraction. But a generative AI tool can. 

As an example, Magpie has built a product optimisation tool using a ChatGPT interface that he believes can be expanded to harness reviews.

“These reviews can and must be mined for more authentic content and deeper sentiment analysis to improve and shape content offerings,” said Watts. 

But first and foremost, the reviews have to be real.