If you ask New York Times reporters to spend less time on Twitter, will they? (Spoiler: yes)

It’s a question that’s crossed the minds of countless news executives over the past two decades: How can I get my reporters to spend less time on Twitter?

For years, Twitter offered journalists an enthralling mix of community, excitement, and eyeballs. It was a buzzing global newsroom, where you could watch big stories break in real time. It was an after-work bar where you could swap stories with colleagues. It was where you could see your readers talking about your work — and everyone else’s. And it was where you could be a human being in your own voice, not just a cog in the publisher’s machine, the one that feeds paragraphs to the copy desk. It was addictive enough that it seemed some reporters were writing more for the retweets than for the audience.

So what happens when a news organization tries to pull its staffers away — not to get them off Twitter entirely, but to cut back? That’s the question behind an interesting paper just published in the journal Journalism Studies. The authors are Shuning Lu of North Dakota State University and Longhan Wei and Hai Liang of the Chinese University of Hong Kong. (I’ve previously enjoyed some of Lu’s other research, which includes work on online discourse, news avoidance, and political efficacy.)

The title is “Social Media Policies as Social Control in the Newsroom: A Case Study of the New York Times on Twitter”; here’s the abstract, all emphases mine:

Baquet’s announcement created what researchers love most, a natural experiment. Did Times staffers’ tweeting look different after April 7 than it did before? And if so, how were those changes distributed within the newsroom? Did, say, political reporters change their behavior more than sports columnists or art critics? Were mid-level managers more likely to conform than low-level reporters? Did journalists in faraway foreign bureaus change as much as those who see their bosses in the Times’ midtown newsroom daily?¹

To find out, Lu et al. gathered 185,969 tweets from 549 Times newsroom staffers over the first six months of 2022 — roughly half before the policy change and half after it. They then used automated text classifiers to rate each tweet as professional or personal, as well as fact-based or opinionated. (I’m sure someone could quibble with any one of those classifications, but the methodology seems sound to me. The authors paid real humans to evaluate a random sample of 1,000 tweets — with five evaluators per tweet! — then used those results to train an AI model that examined all 185,969; more details in the paper.) Overall, their analysis rated 61% of all tweets as professional versus 39% personal; among the professional tweets, 75% were classified as fact-based, 25% as opinionated.²

So what did they find? Let’s start with the big-picture result: Times staffers listened to their bosses. There was a clear before-and-after for their frequency of tweeting. (A huge relief for anyone who’s ever read a newsroom management book.)

The number of staffers tweeting on a given day fell by 14% right after the policy change, and that lower level still persisted months later. (The study singles out two reporters who cut back: magazine reporter Nikole Hannah-Jones, whose daily tweet count dropped from 33 before the memo to 15 afterward, and climate reporter Hiroko Tabuchi, who went from 19 to 5.) For those still tweeting, their output was increasingly professional and fact-based, decreasingly personal and opinionated.

But what about the distribution of that impact within the newsroom? About 30% of the staffers in the sample were managers as opposed to reporters, columnists, and so on. About 26% covered politics, and 11% were stationed overseas. But the authors found no statistically significant difference in how any of those groups changed their behavior compared to their byline-having, politics-ignoring, or ZIP Code-using peers.

They did find some behavioral differences among those groups — ones you might expect. Critics and columnists posted more opinionated tweets, for example, while those covering politics posted more fact-based ones. But they didn’t find that the policy change itself had any differential impact among the various groups. That surprised the authors:

Having mid-2022 data in an early-2025 paper is admirable speed by the standards of a lot of academic research. But it’s remarkable how little the Twitter of three years ago resembles the social media environment most reporters work in today. Baquet’s memo went out to the newsroom on April 7, 2022. Just three days earlier, a regulatory filing revealed that Elon Musk had acquired a minority stake in Twitter. Seven days after the memo, Musk offered to acquire the entire company for $43 billion, a deal that after much drama closed on October 27.³ A few other things have happened since.

“Twitter” is now only the brand identity I use to passive-aggressively refer to something called “X.” The traffic it sends to news publishers has dried up (or, more accurately, been dried up). While Twitter is still arguably the most important platform for journalism — it’s still where most politics and sports news breaks first, for example — large segments of the news media have decamped for Bluesky and other platforms.

That decentralization — along with Musk’s continued animus against the press and the platform’s overall decline — has made Twitter far less important to the industry and its workers than it was three years ago. If a newsroom boss made a similar announcement in 2025 — hey, let’s cool it with the tweeting, everybody, it’s a distraction, it warps our journalism, and it’s a big reputational risk — they’d probably face even less grumbling than Baquet did. “Why I’m leaving X” pieces are to 2025 what “Why I’m leaving New York” pieces were to the mid-2010s.

Lu, Wei, and Liang had one other finding. For each of those 185,969 tweets, they recorded the number of retweets, quote tweets, and likes it received. Would a world of fewer tweets — and even fewer hot takes — mean Twitter users lost interest in what Times staffers were writing? After all, one line of argument has long held that humanizing reporters on Twitter strengthened their bond with the audience, building their own personal brands in a way that benefits the publisher too.

The answer: It didn’t make any difference. Their analysis found the average Times staffer tweet got basically the same audience engagement pre- and post-memo. “This is encouraging,” they write. “Shifting to professional and fact-based reporting on social media not only enhances journalistic norms and maintains the organizational reputation but also retains user engagement, which corroborates research that finds a positive relationship between perceived journalistic accountability and news engagement among news users.”

Ten or fifteen years ago or so, I’d occasionally get asked a question by news company executives: Do we really need to have our own iPhone app? My standard advice then⁴ is that an iPhone app can be a great way to serve your outlet’s most engaged superfans — the ones most attached to your work, the ones willing to give it a spot on their phone’s homescreen. If doing a better job retaining those superfans is something you need help with, then sure, a mobile app can help. But for most publishers, a dedicated iPhone app is a lot of product distraction for relatively little reward. It’ll only be used by a very small fraction of your audience. And even if that fraction is a particularly valuable one, your product investment is likely better spent on your website, your newsletter, and other more direct points of audience contact.

In 2025, from my perspective, social media is becoming more and more like those iPhone apps. The people who follow you on Twitter (or Bluesky, or wherever) are some of your biggest fans. Seeding your news in front of them can play a big part in it reaching larger audiences. A lot of positive outcomes can come from that kind of direct connection. But overall, journalists’ investments in serving social media have probably reached the point of diminishing returns. Individual circumstances will obviously vary, but I suspect most reporters would benefit from taking a couple more steps back.

Illustration via DALL-E 3.

1. Well, this was 2022, so an awful lot of the boss-seeing was probably still on Zoom.
2. Here are a few examples from the paper’s supplemental materials. A professional tweet: “Five key passages from the Alito abortion draft. Nonpaywall link.”

   A personal tweet: “I haven’t posted a photo of my dog here in a while…”

   A fact-based tweet: “‘I don’t know what’s going to happen in the future, but I’m going to keep fighting to try to keep going.’ Rafael Nadal after winning his 14th French Open.”

   And an opinionated tweet: “I love pintos. And I’ve become very fond of some of the other Rancho Gordo beans!” Not the spiciest of hot takes, but you get the picture.

3. Since Musk didn’t gain actual control of the platform until roughly four months after the tweet window being studied here, his wild policy shifts shouldn’t be a confounder in the analysis. And the authors did several tests to see if his general…Muskiness might have influenced the results before he actually took possession and found no evidence.
4. This advice is basically unchanged today.