Despite Facebook's new fact-checking initiatives, it's ridiculously easy to get major traction with fake news on Facebook.
How easy? Well, I ran an experiment to find out, and what I discovered is pretty shocking.
Armed only with $50 and an hour of spare time, I created a fake news site with a fake news story:
As well as a fake Facebook Page and an ad to promote it.
My fake news company was called Citizen News Network and I used a backwards CNN logo for branding. It's not like I was even trying to be super convincing, you guys. I intentionally chose an old, tired rumour that's already widely known as fake news to promote. I think Facebook should have been able to catch this with any manual review or even fake news filters.
So I shared my fake news blog post on Facebook and boosted my post for $50, and here's what happened:
4,645 people reached
200 website clicks
Almost 300 engagements for $50 on a brand new fake news website, on a brand new fake news Facebook Page.
And if you can believe it, Facebook's ad platform gave this campaign a score of 7 out of 10. What does that mean? Simply, Facebook rewards engaging ads with more reach, putting them in front of more people the algorithm determines may be interested in that ad.
This creates a snowball effect of free exposure, because:
When a user engages with the ad, Facebook then displays that engagement in the Newsfeeds of some of their friends, at no additional cost to the advertiser.
Anyone new who engages generates even more post visibility and engagement (and around and around we go).
Facebook claims that 10 million people were reached with these fake Russian ads. However, the snowball effect of free organic exposure generated by the way Facebook pushes ad engagement into Friends' Newsfeeds leads us to believe that it's more likely that 20 to 100 million people were reached.
So how could the Russians do such damage with only $100k budget? We ran a second analysis based on the facts that during the 2017 election:
Facebook ad costs were about half as much as they are today.
News about the candidates was way hotter. Had I ran this ad a year ago, it would have gotten much higher engagement rates and more visibility at a lower cost. How high would the engagement rates have been?
We used average ad fatigue data to model and extrapolate what a likely engagement rate would have been a year ago, then computed a) the cost to reach, and b) the cost to engage with a number of people equal to the margin of victory in the swing states of Wisconsin, Michigan and Pennsylvania.
As a result, we believe Russians were buying Facebook ad impressions for approximately $3.48 per thousand impressions, or 1.6 cents per user engagement during the 2016 election.
Fake news via Facebook Ads is not a US partisan issue as it can come in any form. It seeks to exploit and undermine divisions within society and is a threat to global democracy.
You can see detailed information on our experiment here, with images. Here are the full set of results: