Valve has started publicly testing sorting reviews in the Steam store based on how well they help players make purchase decisions about games.

Historically we’ve sorted reviews by the number of ‘helpful’ votes given to each review by other players. However, we’ve seen that many players use reviews for sharing jokes, memes, ascii art and other content that might not be the most helpful for a potential purchaser. 

Its solution to the problem is the new “helpfulness system,” which introduces an enabled-by-default “most helpful” toggle that skips past highly upvoted ASCII art or memes to show reviews Valve deems more informative.

Valve says it uses a combination of machine learning algorithms, user reports, and the assistance of the Steam moderation team to categorize reviews and consider both positive and negative comments. The platform’s already started the process but says it will “likely take quite a while” to evaluate the more than 140 million existing reviews as well as newly published ones.

Users have the option of returning to the old review categorization system if they choose. They just need to head to the “Display” drop-down located right above user reviews and unselect the “Use new helpfulness system” checkbox.

The news comes just a day after Valve also announced it will no longer allow Steam store descriptions to include links starting in “early” September. Steam says it’s trying to prevent its store pages from becoming advertisements for other Steam store pages, as well as prevent confusion over “prologue” games.