Media companies that rely on so-called “clickbait” headlines or similar tricks to try and get users to click on their posts might want to rethink that strategy. Facebook said on Thursday that it has tweaked its all-powerful news-feed filtering algorithm to reduce the prominence of this kind of content because it says readers want to see more “genuine” posts.
And how does Facebook define what constitutes clickbait vs. a genuine post? In a post on the site describing the new algorithm changes, data scientist Alex Peysakhovich and user-experience researcher Kristin Hendrix reveal that a team at the social network looked at thousands of headlines and came up with a few rules.
“We categorized tens of thousands of headlines as clickbait by considering two key points,” the blog post says. “(1) if the headline withholds information required to understand what the content of the article is; and (2) if the headline exaggerates the article to create misleading expectations for the reader.”
As an example, the Facebook staffers note that the headline “You’ll Never Believe Who Tripped and Fell on the Red Carpet…” withholds information that is required to understand the article. In addition, headlines like “Apples Are Actually Bad For You?!” misleads the reader, the post explains, because apples are only bad for you if you eat too many every day.
Get Data Sheet, Fortune’s technology newsletter.
The Facebook post makes this war on clickbait seem fairly straightforward. But the fact is that this is the fourth or fifth algorithm tweak in the past two years that was aimed at reducing clickbait, which makes it obvious just how difficult it is to stamp out misleading posts, while still keeping readers interested.
Earlier this year, the social network announced an algorithm change that measured how long users spent with specific stories in order to determine the kinds of links that produced value vs. those that didn’t. It also reduced the number of multiple posts from the same site that showed up in feeds.
Last year, there was another algorithm update that looked at how long users spent reading articles inside their news feed, even if they didn’t “like” or comment on them. (As a platform, Facebook prefers that users read in the news feed rather than leaving to visit other sites.)
Facebook’s latest change is also very similar to one that the network introduced in 2014, which it said at the time was specifically aimed at click-bait headlines.
The problem for Facebook is that, as several of its algorithm-change posts have admitted, clickbait headlines tend to get lots of clicks. But in surveys, 80% of readers say they would rather see “genuine” posts, and not those with trick headlines. So should Facebook watch and alter its behavior based on what people do, or based on what they say they want?
Facebook knows how often you use it. Watch:
For media companies, meanwhile, Facebook’s algorithm changes always carry the risk of seeing their traffic from the social network decline rapidly—as it has for many mainstream sites in the wake of recent changes that were designed to promote posts from users about their own lives, rather than links to news stories. Some have seen their traffic decline by as much as 30%.
In a very real sense, Facebook is fighting a monster that it helped to create, as many media watchers pointed out when a senior product manager complained about clickbait in a blog post in 2014. Media companies have become so dependent on Facebook traffic that they inevitably try to game the algorithm in some way, regardless of what the changes are actually designed to accomplish.