Major news outlets are doing a better job of rebutting President Donald Trump’s falsehoods on Twitter than they were earlier this year but there is still room for improvement, according to a new study from Media Matters. 

Media critics argue that news outlets err by passing along Trump’s misinformation in headlines and social media posts without fact-checking him. In order to assess the scope of the problem, Media Matters reviewed the roughly 2,000 tweets about Trump comments sent by 32 Twitter feeds controlled by major news outlets between July 14 and August 3. We recorded whether the tweets referenced a remark that was false or misleading according to The Washington Post’s Fact Checker database and, if so, whether the tweet disputed the misinformation. We compared the results to a study we released earlier this year covering the output of the same Twitter feeds from January 26 through February 15.

Journalists typically treat a politician’s statements as intrinsically newsworthy and highlight them without context in headlines and social media posts. But media critics have argued for years that journalists need to update this traditional method to reflect that this president lies constantly (according to The Washington Post, Trump has made more than 13,000 false or misleading statements as president). When news outlets pass along those statements unchallenged, particularly in the headlines and social media posts that research shows are often all their audience reads, they amplify Trump's misinformation and damage the broader information ecosystem.

Earlier this year, Media Matters provided the most extensive review yet of the scope of this issue. Our groundbreaking study reviewed thousands of tweets sent between January 26 and February 15 by 32 Twitter feeds controlled by major news organizations. 

The Twitter feeds we studied were those managed by:

• U.S. wire services: (@AP, @AP_Politics, @Reuters, @ReutersPolitics) 

• Major broadcast, cable, and radio networks: (@ABC, @ABCPolitics, @ABCWorldNews, @ThisWeekABC, @CBSEveningNews, @CBSNews, @FaceTheNation, @NBCNews, @NBCNightlyNews, @NBCPolitics, @CNN, @CNNPolitics, @FoxNews*, @BreakingNews, @MSNBC, @NPR, @nprpolitics) 

• National newspapers: (@nytpolitics, @nytimes @politico, @postpolitics, @washingtonpost, @WSJ, @USAToday, @latimes) 

• Capitol Hill newspapers and digital outlets that cover Congress and the White House: (@axios, @thehill, @politico, and @rollcall)

We coded every tweet that referenced a Trump quote for whether that comment was false or misleading according to The Washington Post’s Fact Checker database and, if so, whether the tweet had disputed the misinformation. We also engaged in an extensive literature review and consulted with an array of journalists and media ethicists to develop a set of best practices for media outlets to avoid passing along falsehoods in their headlines and social media posts. 

This follow-up study covers the output of the same Twitter feeds from July 14 through August 3. That time period incorporates a variety of major events, including former special counsel Robert Mueller’s congressional testimony and Trump’s racist, false attacks on Democratic Reps. Alexandria Ocasio-Cortez (NY), Ilhan Omar (MN), Ayanna Pressley (MA), and Rashida Tlaib (MI) and since-deceased Rep. Elijah Cummings (MD).

Media Matters has tracked numerous discrete instances in which individual news outlets’ Twitter feeds failed to rebut Trump’s misinformation, from his false claims about his environmental record to his lies about the Ukraine abuse of power scandal. But our results show that, on balance, the passive spreading of Trump’s misinformation by news outlets’ Twitter feeds has declined since our first study -- albeit with substantial room for additional growth.

Trump is still making false claims frequently, but the news outlet Twitter feeds we studied are doing a better job of handling them.

News outlets amplified significantly fewer false or misleading Trump claims without disputing them -- 325 in total, or 15 per day, compared to 407 total, or 19 per day, in our first study. That’s a sizable decline of more than 21% on a per-day basis.

This isn’t because Trump is lying less frequently, or because feeds are reporting on his lies less often. In our first study, 30% of tweets about Trump remarks referenced a false comment, and that figure grew slightly to 33% in the second study. 

Instead, outlets are passing on less misinformation because they are doing a better job of debunking the president’s lies. Outlets disputed Trump’s false or misleading claims 50% of the time, up from only 35% in our first study. 

More than two-thirds of the outlets we reviewed did a better job of debunking Trump’s misinformation in the second study than the first. That said, as outlets are still collectively amplifying Trump’s falsehoods half the time, there is still substantial room for additional gains.

When a tweet about a false or misleading Trump comment included a link -- which often indicates that the tweet’s text is the headline of the article found at that link -- the outlet disputed the misinformation 55% of the time, up from 44% in our first study. 

Media outlets performed worse when they sent tweets about Trump claims that featured embedded video, a format often used to report on comments the president has just made. Outlets tweeting embedded video disputed false or misleading Trump comments 38% of the time -- a poor showing that is nonetheless substantially higher than the first study, in which outlets disputed these comments in only 6% of such tweets.

We continue to find that media outlets dispute the misinformation Trump spews at different rates depending on the venue where the president made his comments.

The Twitter feeds we followed again performed the worst when Trump was speaking at a press gaggle or pool spray, disputing his false or misleading claims only 41% of the time. The rate shows improvement from the first study, when outlets disputed Trump only 8% of the time. Our data show that outlets amplified misinformation in 110 tweets about press gaggles or pool sprays, an increase from the first study, when we recorded 61 such tweets. This increase is likely because Trump is doing more press gaggles on the White House lawn in an effort to bolster his message, leading to increased opportunities for outlets to promote his false claims.

Sixty percent of tweets featuring a false or misleading claim Trump made during an interview disputed the misinformation. There were a total of only two undisputed false claims amplified from interviews during the course of the study. In our last study, media outlets disputed false interview claims only 27% of the time.

Fifty-four percent of tweets featuring a false or misleading claim Trump made during a speech disputed the misinformation. There were a total of 18 undisputed false claims from speeches amplified during the course of the study. In our last study, media outlets disputed false interview claims 51% of the time.

Fifty-three percent of tweets featuring a false or misleading claim Trump made in a tweet disputed the misinformation. There were a total of 195 undisputed false claims from Trump tweets amplified during the course of the study. In our last study, media outlets disputed false Trump tweet claims 24% of the time.

Broadly, the Twitter feeds of major news outlets did a better job of preventing the flow of misinformation in the period we studied over the summer than they did earlier this year. Of the 32 accounts we reviewed, 22 showed an improved rate of disputing Trump falsehoods. The increases for improving outlets were often drastic: Six outlets saw their dispute rates increase by more than 40 percentage points, and 12 by more than 20 percentage points.

But many outlets that showed improvement did so from a particularly low floor, and thus they’re still regularly allowing Trump’s falsehoods to reach their audiences.

The Twitter feed of The Hill, which has 3.5 million followers, remains the worst offender we reviewed, producing 48% of the tweets that pushed Trump’s misinformation without disputing it over the course of the second study. It promoted Trump falsehoods unimpeded 155 times -- more than seven times per day. The Hill actually improved the rate at which it disputed false Trump claims, from a miniscule 13% of the time to a still-low 30% of the time. But the feed is the biggest source of passive misinformation we reviewed because it generates many more tweets about Trump’s comments than any other outlet -- eight times the total of the next most-prolific feed. It frequently tweets Trump quotes without context, and often resends the same tweet over and over again, strengthening the misinformation through repetition. 

ABC News’ Twitter feeds -- @ABC, @ABCPolitics, @ABCWorldNews, and @ThisWeekABC -- also stood out. They collectively sent 300 tweets about Trump quotes, disputing falsehoods in 50 of the 97 tweets about false claims. That rate of disputing Trump falsehoods 52% of the time was a sizable increase from the outlet’s 29% rate in our original study. 

• ABC News’ main Twitter feed (14.7 million followers) disputed Trump’s falsehoods 45% of the time -- an improvement from 26% in the first study -- but still sent 12 tweets amplifying Trump’s falsehoods without disputing them. 

• The network’s politics feed (814,600 followers) disputed the president’s misinformation 55% of the time -- up from 36% in the first study -- but still sent 14 tweets promoting false or misleading Trump claims without disputing them..

• The feed for its evening news broadcast, World News Tonight (1.4 million followers), disputed Trump’s misinformation 38% of the time -- a big change from our first study, when it did not dispute a single falsehood -- while still sending 13 tweets promoting Trump’s false or misleading claims without disputing them.

• The feed for its Sunday morning political talk show, This Week (182,000 followers), disputed Trump’s false or misleading claims 65% of the time, up from 36% in the initial study, but still sent eight tweets passing along Trump’s false claims without disputing them.

Likewise, some of CBS News’ feeds improved greatly from the previous study while still amplifying misinformation most of the time. The network’s general news feed (7 million followers) disputed Trump’s misinformation 12% of the time, up from 8% in the first period, while amplifying Trump’s misinformation without dispute 15 times. The feed for CBS Evening News (317,100 followers) disputed Trump’s falsehoods 47% of the time, up from 28%, while passing them along uncorrected nine times.

NBC’s feeds were somewhat more effective in disputing Trump’s misinformation but still show room for improvement. The network’s main news feed (6.9 million followers) disputed misinformation in 13 of 18 tweets for a rate of 72%, up from 48%. Its politics feed (531,400 followers) did so in nine of 12 tweets for a rate of 75%, up from 57%. And the feed for its nightly news show (992,800 followers) did so in three of five tweets, a rate of 60%, up from 11%. MSNBC’s feed (2.8 million followers) included disputes in 22 of 29 tweets about false Trump claims, or 76% of the time, an increase from 45% in the first study.

A handful of outlets managed to entirely avoid promoting Trump’s misinformation over the course of the study, but they did so while tweeting about only a few Trump falsehoods. 

The Associated Press feed debunked the misinformation in all four false claims it tweeted about, while its politics feed did the same in the two relevant cases. The main feed also did not pass along misinformation in our first study.

The feeds for NBC’s and CBS’ Sunday morning political talk shows, Meet the Press and Face the Nation, had perfect records in their handling of six and three tweets about Trump falsehoods, respectively. Both improved from particularly poor records in our initial study -- Face the Nation did not dispute claims in a single one of the 16 tweets it sent about Trump’s misinformation, while Meet the Press did so in only one of six tweets (17%). 

The primary Twitter feeds for The New York Times and Los Angeles Times disputed false Trump claims in eight of nine and seven of eight tweets, respectively. The former outlet saw improvement from a 70% rate in the first study to an 89% rate in the second study, while the latter’s dispute rate was largely unchanged.

Axios’ feed disputed Trump’s misinformation 12 out of 13 times (92%), giving it the highest rate of any feed we studied that tweeted about 10 or more false Trump claims. That’s a big step forward from our first study, in which Axios disputed only 36% of the time. MSNBC’s feed was the next best of those that tweeted about 10 or more false Trump claims, disputing misinformation 76% of the time (22 of 29 tweets). That also represents a big gain, as the feed disputed only 45% of the time in our first report.

While most of the news outlet Twitter feeds we studied improved from our first study to our second, six saw their rates of disputing Trump’s falsehoods decline. Two in particular are worth discussing, as we had highlighted their success in our original review.

The Washington Post’s main feed, which had disputed 89% of false Trump claims (33 of 37 tweets) in the first study, did so only 58% of the time in our second (18 of 31 tweets). Its politics feed also declined, from disputing 67% in the first study to 59% in the second.

And CNN’s main feed, which had disputed 75% of false Trump claims in the first study (nine of 12 tweets), disputed only 60% in the second (18 of 30 tweets).

Media Matters reviewed more than 56,000 tweets sent between midnight EST on July 14 and midnight EST on August 4 from the following Twitter feeds of U.S. wire services; major broadcast, cable, and radio networks; national newspapers; and Capitol Hill newspapers and digital outlets that cover Congress and the White House: @AP, @AP_Politics, @Reuters, @ReutersPolitics, @ABC, @ABCPolitics, @ABCWorldNews, @ThisWeekABC, @CBSEveningNews, @CBSNews, @FaceTheNation, @NBCNews, @NBCNightlyNews, @NBCPolitics, @MeetThePress, @CNN, @CNNPolitics, @FoxNews*, @BreakingNews, @MSNBC, @NPR, @nprpolitics, @nytpolitics, @nytimes @politico, @postpolitics, @washingtonpost, @WSJ, @USAToday, @latimes, @axios, @thehill, and @rollcall.

We chose that time frame both because it involved a period of high-stakes political turmoil in which the information the public received was especially crucial, and because the president made public remarks at that time in a variety of ways, including via political rallies, other speeches, press gaggles and pool sprays, interviews, and hundreds of tweets.

Media Matters identified of that group roughly 2,000 tweets that referenced a comment Trump had made. We then coded those tweets for whether they referenced a remark that’s included in The Washington Post Fact Checker’s database of false or misleading Trump claims. In such cases, we reviewed whether the outlets’ tweets had disputed the Trump claim. We reviewed the text and images embedded in the tweets but did not review embedded videos.

*@FoxNews did not tweet during the period of the study.