Round 8% of the 31.1 million opinions submitted to Tripadvisor in 2024 have been pretend, in line with the corporate’s “Transparency Report 2025.”
That is greater than twice the quantity detected in 2022, in line with firm experiences.
However that does not imply the variety of pretend opinions on the positioning doubled, stated Becky Foley, vice chairman and head of belief and security at Tripadvisor.
Whereas submissions to the web site elevated, Tripadvisor’s insurance policies on pretend opinions additionally modified, she stated, citing the corporate’s extra aggressive stance towards “incentivized opinions.” These happen when corporations provide prospects reductions or freebies in alternate for opinions, or present incentives to workers whose names are talked about in opinions.
“The workers get their mother, finest pal or cousin to submit opinions, mentioning their names,” she stated. “It finally ends up resulting in companies having opinions that are not really worthwhile to our neighborhood.”
What’s a pretend evaluate?
In accordance with Tripadvisor, it’s “any evaluate submitted by somebody who’s knowingly submitting biased or non-firsthand content material, in an effort to govern a property’s fame.”
However there is a greater purpose detections are on the rise, she stated.
“Our system is at all times getting higher,” she stated, citing Tripadvisor’s three-pronged course of that depends on auto-detection, human evaluate and neighborhood suggestions.
Some 7% of submissions in 2024 have been auto-rejected earlier than being posted, the report stated. Auto-detections additionally flagged one other 5% of submissions for human evaluate.
Nevertheless, Tripadvisor’s belief and security group finally moderated greater than 4.2 million opinions, amounting to greater than 13% of all submissions in 2024, in line with the report.
One other 244,000 opinions have been disputed by members on the third stage of evaluate, the report confirmed. Of those, some 72% remained on the positioning, and 28% have been eliminated, it stated.
spot a pretend
Tripadvisor has 4 classes of pretend submissions: boosting, vandalism, member fraud and paid opinions, Foley stated.
There is a false impression that vandalism accounts for many pretend opinions, Foley stated, including that boosting (54%) and member fraud (39%) have lengthy made up the majority of these opinions.
Paid opinions (4.8%) are a smaller however “extra pernicious” class of fakes, she stated.
This contains “evaluate farms,” she stated, including that these authors are sometimes concerned in different varieties of on-line fraud too.
Most paid opinions originate in Asia, Foley stated, although solely 17% of actual submissions got here from the continent final 12 months.
In 2024, greater than one-third of all paid submissions detected by Tripadvisor got here from Indonesia and Vietnam, whereas in 2022 most paid opinions got here from India, experiences present.
A ‘cat and mouse’ recreation
Rooting out pretend opinions is a continuing “cat and mouse” recreation, Foley stated. However Tripadvisor is getting higher at detecting them yearly, she stated.
“We’re the primary to confess that we’re by no means going to achieve absolute perfection,” stated Foley. “We’d not catch [a fake] the primary time, however we’ll catch it will definitely.”
Tripadvisor makes use of its personal expertise, developed over 25 years, to flush out fakes — a system that more and more depends much less on what a evaluate says, and extra on the way it’s posted, she stated. The corporate makes use of synthetic intelligence and behavioral biometrics to seek out patterns, which may detect abnormalities like submission spikes and IP deal with masking makes an attempt.
To catch paid reviewers, Tripadvisor investigators pose as pretend evaluate brokers, Foley stated. When a foul actor posts their first paid evaluate, “we’ve got all the information … lots of of information factors related to it,” which the corporate makes use of to construct patterns to establish different opinions submitted by the creator prior to now.
Violators aren’t pulled off the positioning, however their rankings are penalized for a 12 months, Foley stated. Repeat offenders obtain a crimson badge on their itemizing “that claims this property is attempting to deceive you,” she stated.
AI opinions: not pretend, however not allowed
Most opinions written by AI will not be pretend, Foley stated, calling that “one of many myths I like to bust.”
Most are written by common vacationers that use the expertise to make their writing extra polished, she stated.
Nonetheless, the corporate would not permit them — at the very least not for now.
In 2024, Tripadvisor eliminated greater than 200,000 opinions that it suspected have been written by AI, in line with the corporate’s 2025 transparency report.
“We are going to proceed to observe the developments and patterns. However proper now, we simply don’t need vacationers to come back to TripAdvisor for a sea of sameness.”
