Junk news and Russian misinformation flooded Twitter during the 2016 election

Polarizing conspiratorial content and inaccurate “junk news” flooded Twitter during the 2016 presidential race, according to new research from Oxford University, particularly in swing states like New Hampshire, Pennsylvania and North Carolina.

The new findings about the extent of misinformation on the site comes as Twitter prepares to brief House and Senate lawmakers on the extent to which Russian-backed agents may have co-opted social media in a bid to interfere with the U.S. election.

But unlike the recent revelations about the extent of Russia’s use of Facebook to disrupt the U.S. presidential election, the study didn’t conclude if there were specific agents behind the spread of misinformation on Twitter. Instead, the idea was to look at the spectrum of tweets and examine how often people (or bots) shared bad information.

It found about 20 percent of tweets sampled around the U.S. presidential election qualified as “polarizing and conspiracy content,” including links to “junk news,” WikiLeaks or Russian sources, like Sputnik and RT. At the same time, about 20 percent of the tweets also included links to “professional news brands,” without citing specific news sources.

The study, conducted by researchers at Oxford’s Computation Propaganda Project, sought to focus on U.S. political conversations on Twitter at a few major moments: The presidential debates as well as the 10 days prior to the election. They included only tweets from profiles that contained location information, a link and an election-related hashtag, about 1.2 million tweets in all.

Researchers also discovered the greatest concentrations of misinformation happened in swing states — 12 of 16 of those states, seen as crucial in every presidential election, had higher levels of dubious content than the national average in 2016. The southeastern U.S. also registered higher-than-average misinformation.

Twitter, however, cast doubt on the report’s methodology, citing how the authors compiled tweets using a specific Twitter database, in this case its API, which only allows researchers to download about 1 percent of all tweets.

“Furthermore, we believe the peer review process exists to ensure published research is authoritative and empirically sound. In this case, there does not appear to have been any such review,” the spokesperson said in a statement.

The study’s authors acknowledge those limitations in their work. “Full access to user location data would provide better resolution on the distribution of users across states,” they wrote, “but this response rate is sufficient to help us begin to understand the relationship between where users are and what kinds of political content they are being served by Twitter.”

The findings still pose new problems for Twitter, which is set to head to the U.S. Capitol today to brief the House and Senate Intelligence Committees. The two panels are investigating Russian meddling in the 2016 presidential election while Robert Mueller, a former FBI director, spearheads the U.S. government’s own independent inquiry.

Democratic lawmakers on both intel committees already have expressed deep fears that Russian agents used social media sites to spread misinformation, and some are particularly concerned that’s happened on Twitter as the result of bots. Congressional investigators also have taken aim at one of Twitter’s peers, Facebook, where Kremlin-aligned forces purchased about 3,000 ads in order to sow political discontent, sources have said. Both companies, along with Google, are set to testify in public about the matter at congressional hearings this fall.

This article originally appeared on Recode.net.