The Failing New York Times
What analyzing nearly half a million headlines reveals about how the paper softened its coverage of Trump.
A friend of mine texted me recently after listening to an episode of The Daily, the New York Times flagship podcast. “What happened to The Daily?” she wrote. “I listened to it for the first time in a while and it was oddly pro-Trump.” I responded that the NYT had been “sanewashing” Trump for a while now, as if it were fact. It’s a prevailing theory you can read on left-wing blogs and hear on TikTok. But is it actually true? Having been a NYT subscriber and avid podcast listener for years, I’ve noticed what felt like a softening toward Trump in his second term. That got me thinking: what does the data actually show?
So I decided to look for myself.
The New York Times reaches roughly 11 million digital subscribers and sets the agenda for downstream coverage across American media. When it names something false, that framing travels. When it softens that framing, the softening travels too. The editorial choices the Times makes about how to describe a sitting president can shape what tens of millions of readers understand to be true.
I signed up for a New York Times Archive API and collected 492,101 articles published between January 2017 and May 2026, spanning Trump’s first term, the Biden presidency, and Trump’s second term. I sorted every article into one of four categories: Opinion, The President, Politics (U.S. political coverage not directly about the president), and Unrelated. I defined three political eras with hard boundaries at each inauguration date: Trump 1 (January 2017 to January 2021), Candidate Trump (January 2021 to January 2025, covering his time out of office, his criminal trials, and his 2024 campaign), and Trump 2 (January 2025 to present).
I then built predefined lists of accountability language and normalizing language, and measured how often each appeared in Trump headlines across all three eras. Accountability language includes words like “falsely,” “illegal,” “unconstitutional,” and “fraud.” Normalizing language includes words like “crackdown,” “policy,” “approach,” and “initiative.” Fact-checks were identified by explicit “Fact Check” headline labeling and equivalent constructions. The findings are reproducible by anyone with a spreadsheet and the methodology is the same as used by academic researchers when analyzing the press.
What I found
The short answer is yes, something has measurably changed but the data tells a more specific and more interesting story than the popular “sanewashing” framing suggests.
The clearest single data point is the disappearance of one word: “falsely.” In Trump’s first term, the New York Times used the word “falsely” in 39 Trump news headlines. It appeared consistently across four years, an average of nearly one headline per month naming something Trump said as false. In Trump’s second term, it has appeared in exactly one news headline in 17 months. The word that defined the paper’s first-term fact-checking posture has effectively vanished.
We see the same story in a different number with the volume of fact-check pieces. In Trump 1, the paper published an average of 3.1 explicit fact-checks of Trump claims per month. In Trump 2, that number dropped to 1.4 per month, a 55 percent decline, even as coverage volume reached record highs. Trump 2 generates 464 Trump headlines per month compared to 291 in Trump 1. The paper is covering him more than ever while fact-checking him less.
In 2016, then-executive editor Dean Baquet gave a series of interviews, explaining that the paper had made a deliberate editorial decision to call Trump’s false statements false, rather than hiding behind “claims” or “says.” The word appeared in headlines because editors consciously chose to use it. His successor, Joseph Kahn, became executive editor in June 2022 and began repositioning the paper toward neutrality, giving multiple interviews explaining the shift. “Falsely” did not fade gradually. It collapsed at a specific moment: November 2024, the month Trump won the election.
The broader pattern confirms what that single word suggests. Across the full dataset, I measured an accountability-to-normalizing ratio for each era: how often direct accountability language appeared relative to softening or procedural language. In Trump 1, the ratio was 0.24. During the Candidate Trump years, when he was out of power, facing criminal indictments, and campaigning for a return to the White House, it rose to 0.50. In Trump 2, it fell to 0.07; less than a third of the first-term level.
Interestingly, the number that explains the collapse is not the accountability count. Accountability language in Trump headlines ran at 7.4 words per month in Trump 1 and 6.1 words per month in Trump 2, essentially unchanged. What changed was the normalizing language: procedural words like “policy,” “approach,” “crackdown,” “initiative,” and “order” appeared at 30.9 words per month in Trump 1 and 83.6 words per month in Trump 2. The paper is not saying fewer critical things it’s burying them in nearly three times as much procedural framing.
Looking specifically at the topics of the data, one pattern stands out: presidential fitness coverage. Biden’s cognitive health became a defining story of his presidency, covered with clinical specificity. Trump, only two years younger than his predecessor, has generated his own wave of fitness-related moments, including multiple documented instances of falling asleep during public events and Cabinet meetings. There is a sharp divergence in how the two were covered.
During Biden’s presidency, the word “memory” appeared in six news headlines directly questioning his recall and “Vigor” appeared three times. The Special Counsel report alone, a single document raising questions about his memory, generated 83 separate news headlines. In the two weeks following the June 2024 presidential debate, the paper published more than 40 headlines about Biden’s fitness and debate performance.
Get Maple Herriot’s stories in your inbox
Join Medium for free to get updates from this writer.
In Trump’s second term, “memory” has appeared in Trump news headlines zero times. “Vigor” zero times. Behaviours that generated public discussion about cognitive fitness, including the rambling speeches, mid-sentence topic changes, falling asleep, and documented verbal slips, generated 122 headlines routed through late-night comedy coverage over 17 months, an average of more than seven per month. The paper has not stoped covering these behaviours but it has reclassified them as entertainment.
This rules out the simplest alternative explanation: that the paper became more cautious toward all presidents over time. It did not. The paper that covered a Democratic president’s cognitive fitness as a serious governing question covers a Republican president’s similar behaviours as material for comedians. The softening is specific to Trump.
It’s the same with foreign policy. In Trump 1, the paper’s Opinion section ran headlines like “President Trump Fails NATO,” “Trump Tries to Wreck America’s Alliances,” “Trump’s Bullying Tactics at the NATO Meeting,” and “Trump Betrays an Ally at the Turkish Border.” These are Opinion headlines, editorial judgments stated as plain facts. In Trump 2’s Opinion section, covering conduct that is arguably more severe, including active realignment toward Russia during a war, abandonment of Ukraine, and open threats to leave NATO, the framing shifted: “Europe’s New Reality: Trump May Not Quit NATO, but He’s Already Undercutting It,” “Trump’s Pivot Toward Putin’s Russia Upends Generations of U.S. Policy,” “Trump Bullied Allies. Now They’re Hard to Find.” As the scope of the conduct drew more severe assessments from critics, the language used to describe it softened.
This language shift is not explained by newsroom turnover either. The journalists who wrote the sharpest first-term accountability headlines are still at the paper and still covering Trump in his second term. Five of the most prolific Trump 1 accountability bylines, reporters who collectively wrote over two thousand Trump 1 articles, have so far written 392 Trump articles in Term 2 combined. Are these once vocal critics of Trump the victims of the policy decisions from the higher ups? I don’t know. The data can’t say, it can only show the pattern.
Fear of Reprisal?
The obvious explanation is fear. In September 2025, Trump filed a $15 billion defamation lawsuit against the New York Times, refiled in October 2025 after a federal judge dismissed the original complaint; the case remains pending. Meanwhile Trump continues targeting media organizations openly and repeatedly. The paper operates in a legal and political environment that makes aggressive accountability journalism genuinely risky. That explanation is not wrong, but it is not complete.
We can see the language shift did not begin with the lawsuit. It began in November 2024, the month Trump won the election, ten months before any legal action. In that month, accountability language in Trump headlines hit its lowest rate of the year. The paper recalibrated at the moment Trump won power, not when he threatened it legally. Fear of a specific lawsuit is a partial explanation for what happened after September 2025. It does not explain what happened in November 2024.
Less than two weeks before the 2024 election, the paper’s own journalists held a nearly 90-minute internal meeting pressing Kahn directly on “sanewashing.” Semafor’s Max Tani obtained a recording of the October 24, 2024 meeting and reported on it in November of 2024. Leadership defended the neutrality approach. The journalists who would go on to write zero accountability-framed Trump headlines in Term 2 had already told their editors their concerns. The language shift reflects documented institutional choices, and those choices preceded any external legal pressure.
It’s the Money, Always the Money
The New York Times spent a decade building something it did not have in 2017: a subscriber base that extends far beyond news. In September 2014, it launched NYT Cooking. In October 2016, one month before Trump’s first election, it acquired Wirecutter for a reported $30 million. It launched The Daily podcast in 2017, put Cooking behind a paywall that same year, acquired Wordle in January 2022, and acquired The Athletic for $550 million later that year.
By the time Trump returned to power in January 2025, the subscriber base looked nothing like it did in 2017. Roughly 32 percent of the paper’s 10.8 million digital subscribers at end of 2024 subscribed only to non-news products: Games, Cooking, Wirecutter, Audio, or The Athletic. They had never paid for the news site at all. Bundle subscribers, people who came in through games or cooking and were upsold into news, accounted for nearly half the total base. The CEO told investors explicitly that the company was building “levers” to buffer against news cycle turbulence. The lever was a lifestyle empire.
The “Trump bump” subscription surge of 2017 through 2020 helped fund that empire. Adversarial coverage drove subscriptions when Trump was new, the resistance was energized, and the paper’s primarily left-leaning subscriber base wanted confrontation. But the steepest subscriber growth came later, during the Biden years, driven by lifestyle products, not news outrage. The paper that benefited from the Trump coverage in 2017 is a structurally different institution in 2025.
In May 2025, at the Sir Harry Evans Investigative Journalism Summit, Kahn publicly acknowledged that the paper had been publishing a new Trump story on average every half hour. The analysis shows explicit fact-checks dropped 55 percent over the same period. The paper is generating Trump content at industrial volume and fact-checking it at a fraction of the rate it once did.
This is the most important distinction the data surfaces. The shift in language began at the election, before any legal threat, before any specific act of government pressure. It began when the business calculus changed. A second Trump term meant a different risk environment for a company with 12 million subscribers, a sports network, a games platform, and a recipe app. The word “falsely” did not disappear because the Times stopped believing Trump lies. It disappeared because such precision is now a liability.
Conclusion
The New York Times did not get scared of Donald Trump. It got optimized around him. The paper covers Trump at record volume in his second term; The alarm is genuine, the anger in much of the coverage is real. What is gone is the precision: the specific words that name conduct as false, illegal, or unconstitutional, the fact-checks that called lies by their name, the headlines that stated editorial judgments as plain facts rather than hiding them inside analytical frames.
The normalization and “sanewashing” occurred not because the paper stopped criticizing Trump, but because it flooded the zone with procedural coverage that buried whatever precision remained.
The timeline points squarely to November 2024, at the election result, not at any subsequent legal threat. The shift is institution-wide and consistent across beats and bylines. The record shows the paper covered a Democratic president’s cognitive fitness as a governing crisis and a Republican president’s similar behaviours through late-night comedy.
A free press can’t be a neutral observer of power because it’s one of the few institutions positioned to name abuses of power as they happen, while they can still be stopped. That function does not depend on which party holds office. Imagine a future president far more dangerous than Trump; journalism is supposed to be the first line of defence for a democracy and based on the evidence the paper of record will bend further, not stand firmer. A newspaper that hedges its language at the moment power is most concentrated is not protecting its readers, it’s protecting its subscriber count.
The Times spent a century building the credibility to call power what it is. Now it’s using that credibility to protect a recipe app.
