Online reviews must always strike a balance between protecting the anonymity, credibility, and independence of the reviews. But researchers have found that many online review scenarios incentivize misbehavior by companies to game the system. Prominent ones, such as Yelp, have had to invest in software and monitoring to alert users to manipulators.

It’s not always easy to prove when reviews fraudulently favor a company — a single bad review might just be a bitter ex-employee or competitor, for example. But the Wall Street Journal used long-term ratings data from 8,500 companies with at least 100 reviews on Glassdoor to identify spikes in ratings that, in some instances, may have resulted from company pressure on employees to write 5-star reviews.

One enterprising former fake review writer even got a fictional restaurant named for his backyard shed to top London’s TripAdvisor rankings.

Users have discussed such manipulation openly in the past, but the new investigation provides some compelling larger-scale data. Companies it identified with unusual review spikes include SpaceX, Slack, LinkedIn, Anthem, and Clorox. The spikes coincided the most with the month leading up to when Glassdoor calculates its annual rankings.

Many of the companies identified said that they do encourage employees to give feedback. Others, such as one SpaceX recruiter who listed Glassdoor rankings-boosting campaigns on her LinkedIn profile, didn’t reply to the WSJ’s request for comment.

Glassdoor, which sold last year to Recruit Holdings for $1.2 billion, rejects up to 10% of reviews for failing to meet its community guidelines, a spokesperson said. Some 16% of Yelp reviews as filtered as fraudulent, researchers reported in 2015.

What’s the solution? Perhaps more transparency about the weaknesses of online reviews: a study that appeared last month concluded that reviews sites could gain user trust by retaining and flagging fraudulent reviews, rather than censoring them.