Over the past six months, First Draft has been involved in projects monitoring information disorder circulating in the lead up to two elections — the Brazilian presidential election and the U.S. midterms. Misinformation is still a serious problem, but it’s not the 100 percent fabricated stories that got so much press in 2016. Most of what we saw in both countries was content based on a kernel of truth. It was genuine but recycled content; it was imagery taken out of context; it was the use of statistics in ways that could be misread easily by the audience. Rather than text articles, much of this content was shared as standalone visual posts on Facebook, Twitter, Instagram, and WhatsApp.

This type of content took advantage of deep partisan divisions and was designed to reinforce positions and denigrate the other side. In both countries, we saw content clustered around four key themes: election integrity, immigration, hate (in the form of anti-Semitism, misogyny, and racism), and conspiracy theories about global networks of power. This content used dog whistles, logical fallacies, and false equivalency — all tried-and-tested methods of persuasion. This is the misinformation we should be worried about in 2019.

2018 was the year of hyperbolic headlines about “deepfakes,” technology designed to take someone’s face and to use artificial intelligence to portray them as saying or doing anything. Maybe I’m being naive, but this isn’t what I’m worried about at all. Academics and technologists agree that we’re roughly four years away from the level of sophistication that could do real harm, and there is currently an arms race afoot to produce tools to effectively detect this type of content.

Instead, I’m very worried about the drip, drip, drip of these divisive hyperpartisan memes on society. I’m particularly worried because most of this content is being shared in closed or ephemeral spaces, like Facebook or WhatsApp groups, SnapChat, or Instagram Stories. As we spend more time in these types of spaces online, inhabited by our closest friends and family, I believe we’re even more susceptible to these emotive, disproportionately visual messages. (Goes without saying we need much more research on this issue so we have a greater understanding of the impact of messages that travel between trusted connections.)

There is little the platforms could or should do with this type of legal content. Much of it cannot be fact-checked in any formal sense, and many would argue that this type of content is “politics as normal.” But we’ve never had “politics as normal” playing out on the small screens we stare at all day.

Maybe I’m being too technologically determinist. Maybe this type of content won’t have a disproportionate impact. But the problem is I don’t think we have the ten years we need in order to wait for the longitudinal studies to be carried out.

My prediction is 2019 will be the year when misinformation becomes harder to track as it moves out of sight, into more closed and ephemeral spaces. An already challenging situation is about to get much worse.