Regular Yelp users may want to exercise caution next time they make a restaurant choice based on a handful of online reviews.

A recent study by Harvard Business School assistant professor Michael Luca and Boston University assistant professor Giorgos Zervas has found that small independent companies are the most likely type of business to solicit faked positive reviews on Yelp, the popular review website.

According to Luca, some businesses hoping to boost their Yelp ratings hire companies to post large numbers of fake reviews meant to look real. Others ask family members and friends to leave positive reviews for their company and negative reviews for competitors.

The Business School study found that restaurants with poor reputations stemming from low ratings or negative reviews were more likely to partake in review fraud.

In previous research, Luca discovered that a one-star increase on Yelp, where businesses can earn up to five stars, leads to an increase in revenue of five to nine percent.

“Online reviews are shockingly important for a small business that is trying to thrive. Five percent is enough of a difference in sales to make a business stay open or closed,” said Luca, adding that businesses are in an “arms race” to solicit the best reviews.

According to Zervas, Yelp’s own algorithms find that one out of five reviews submitted to the site are fake.

While Yelp attempts to relocate as many suspicious reviews as possible from the main site to a special filtered page, Zervas cautioned that “[Yelp] may misclassify them.” He said that this misclassification could lead to real reviews being filters while faked reviews end up published.

While the Business School study looked only at Yelp restaurant reviews, both researchers recognized that review fraud exists across businesses and online rating platforms.

“The high number of fake reviews on places like Yelp and TripAdvisor brings the credibility of the system into question,” said Luca.

Zervas advised consumers to evaluate online reviews from a more critical standpoint.

“I don’t look at the average rating. I think it’s misleading. Another major component of a business is how many reviews it has,” said Zervas. “Often I’m more confident in the quality of a business that has 3.5 stars and 100 reviews versus a company with 4 stars and 10 reviews.”

According to Luca, while many computer scientists are currently attempting to develop algorithms that might be used to more accurately weed out fake reviews, eradicating the problem entirely will take more than just advances in technology.

“The incentives are strong and the consequences are not serious enough,” Zervas said of review fraud. “I don’t see fake reviews disappearing anytime soon.”