Analysis finds that the 50 most viral fake news stories on Facebook in 2017 generated more shares, reactions, and comments than those of 2016
Craig Silverman / BuzzFeed : Tweets: @anneapplebaum , @craigsilverman , and @craigsilverman See also Mediagazer Tweets: Anne Applebaum / @anneapplebaum : no, it's not getting better: the 50 most viral fake news stories of 2017 generated more total shares, reactions, and comments than the top 50 hoaxes of 2016 http://www.buzzfeed.com/... via @CraigSilverman Craig Silverman / @craigsilverman : NEW: In spite of Facebook's fact-checking and product efforts, we found that the top viral fake news of 2017 generated more engagement than last year: http://www.buzzfeed.com/... Craig Silverman / @craigsilverman : Total Facebook engagement generated by 50 big fake viral stories we analyzed in 2016: 21.5 million Total engagement for the top 50 of *2017*: 23.5 million Total Engagement for fact checks of the top 2017 viral hoaxes: 127,543 More data and analysis: http://www.buzzfeed.com/... See also Mediagazer
Context & Ripple Effects
A year after Craig Silverman's pre-election analysis found the top 20 fake election stories out-engaged the election coverage of 19 major news outlets on Facebook [[a:877873]], his year-end follow-up shows the problem growing rather than receding: the top 50 hoaxes of 2017 pulled 23.5M total engagements versus 21.5M for 2016's top 50 — even though Facebook had spent the year rolling out fact-checking flags, whose combined reach was a mere 127,543 engagements.
The finding matters because it became the benchmark every later measurement argues with: within a year, follow-up research recorded a 50%+ drop in Facebook engagement for sites identified as fake-news publishers [[a:933446]], while Avaaz's 2019 study both exaggerated false-story prevalence and exposed genuine loopholes in Facebook's enforcement [[a:948516]]. Silverman's 23.5M number is the 'before' picture those corrections are measured against.
First-order effects
- Facebook's flagship remediation — fact-check warning labels — demonstrably failed to suppress distribution at scale, drawing 127,543 engagements against 23.5M for the hoaxes they were meant to counter, so the burden of action shifts back to Facebook's ranking and product teams rather than its third-party fact-check partners.
- Operators of viral hoax pages get a clear market signal: despite a year of platform crackdowns and press scrutiny, top-tier fake news remained a growth business on Facebook, sustaining the incentive to keep publishing.
Second-order effects
- Facebook faces pressure to move beyond post-hoc labeling into algorithmic demotion of repeat offenders — precisely the lever that later studies credit when fake-news-site engagement fell by half or more between the 2016 election and mid-2018.
- Scrutiny spreads to adjacent platforms: the same research thread that tracked Facebook's decline found shares of fake-news content still rising on Twitter, making cross-platform comparison the new standard for judging any single network's cleanup claims.
Third-order effects
- If the pattern holds, platform accountability migrates from content-level fixes (labels, takedowns) to systemic ones — ranking design, prevalence measurement, and independent audits — which is exactly where the later disputes land: Avaaz overstating the problem yet revealing real enforcement gaps, and feed surveys showing most users saw no news at all in their first posts, complicating what 'exposure' even means.
- Engagement totals themselves become the contested political metric — researchers, platforms, and regulators each measuring different things — pushing the industry toward standardized third-party measurement of misinformation reach rather than self-reported numbers.
The trend: Social-platform disinformation control is shifting from visible content labels to invisible ranking interventions, with independently measured engagement — not platform claims — becoming the yardstick regulators and researchers use to judge whether it worked.