When I realized the power of online journalism in the early aughts, I saw transparency as key to its promise. I’d watched Gawker X-ray New York’s media scene, and seen bloggers tear apart mainstream reporting on the 2004 presidential campaign. I found that I could drive the political conversation simply by telling my readers what I knew in plain English, when I knew it. At Politico in 2007, we adopted Gawker’s ethos that many of old-school journalists’ most interesting stories were the ones they told one another in bars, rather than the ones they printed, and applied it to American politics. We immediately hooked political junkies on a steady stream of scoops that assumed readers were on a first-name basis with Hillary and Barack, and that they didn’t need us to provide much context or analysis.

At its best, this ethos bypassed the patronizing, gatekeeping practices that often led great American institutions to mislead the country on vital public subjects. At its worst, it encouraged journalists to publish things that their predecessors had good reason to pass over, such as leaked sex tapes.

And then there were the hard cases, the explosive facts and documents that journalists had long worried citizens would take out of context if they were revealed in full. I found, and still find, that concern ludicrous in this digital age. But the trajectory of the document known as “the dossier” has disabused me of my Panglossian assumption that the new transparency is a simple blessing.

I first got wind of the dossier in December 2016, when I was the editor in chief of BuzzFeed News. One of our reporters, Ken Bensinger, received an unusual invitation to a small gathering at a hilltop mansion in Sonoma County, north of San Francisco. He’d been invited by an acquaintance, Glenn Simpson, a onetime journalist who had become a kind of private investigator and co-founded the opposition research firm Fusion GPS. Ken got lost and showed up late, finding a boisterous, all‑male affair: plenty of booze, hunks of meat on the grill, some weed being smoked outside. Simpson drew him into a conversation about a mutual acquaintance, a former British spy named Christopher Steele. Simpson then told Ken something he didn’t know: Steele had been working the case of the president-elect, Donald Trump, and he’d assembled evidence that Trump had close ties to the Kremlin—including claims that Michael Cohen, one of his lawyers, had held secret meetings with Russian officials in Prague, and that the Kremlin had a lurid video of Trump cavorting with prostitutes in the Ritz-Carlton Moscow that would come to be known as the “pee tape.”

Ken told Simpson’s story to our investigations editor, Mark Schoofs, who told me about it. Simpson wouldn’t give Ken the document, and neither would Steele. It was merely high-grade Washington gossip, irresistible chatter.

I heard about the report again over lunch in Brooklyn, when a peculiar character in Hillary Clinton’s orbit passed through town. David Brock had been an anti-Clinton journalist in the 1990s. Now he was Hillary’s fiercest ally, a genius at raising money for Democratic groups. He showed up at a café a couple of days before Christmas wearing a coat with a lavish fur collar, and stashed full shopping bags beside the table. Brock was consumed with the mission of stopping Trump, manic; he was headed, it turned out, for a heart attack that landed him in the hospital. He wanted to spread the word about a dossier of allegations involving Trump’s ties to Russia. Brock didn’t have the document, he said. But he knew The Washington Post did, and so did The New York Times. Politicians had it too, he told me, and spies; as far as I could figure out, so did everyone, except the reading public. And me.

That, I believed, made it exactly the sort of thing you should publish. The dossier would be a great story, a journalistic and traffic sensation.

We were hardly the first journalists to get the document—but we may have been the first to get it without promising to keep it secret.

Simpson, whose firm was working for the Democratic National Committee, had months earlier summoned the leading lights of Washington journalism to the Tabard Inn, a tatty hotel off Dupont Circle. There, Steele calmly shared his shocking suggestion that Trump had been compromised by the Russian government. The journalists came from The New York Times, The New Yorker, ABC News, CNN. BuzzFeed didn’t get an invite.

To Simpson’s frustration, the reporters couldn’t confirm the dossier’s allegations. And because they had promised Simpson that they wouldn’t write about the dossier itself, its author, or its path through the American government, they couldn’t report on these things either, even as they became equally interesting stories.

On December 29, David Kramer invited Ken to his office at the McCain Institute. He then did something careful Washington insiders do: He left Ken alone in the room with the document for 20 minutes, without, in Ken’s view, giving clear instructions about whether he could make a copy. Ken took a picture of every page. (Kramer later denied that he’d allowed Ken to copy it, though I believed the denial was a fig leaf. Kramer eventually clarified that denial to say that he had wanted Ken to take a paper copy with him, rather than take pictures of the document.) I printed out the 35-page document and pored over it, looking for details that we could confirm, or refute. Then I hid my copy in the back of a closet. We scrambled—as other news outlets had done—sending reporters to check out the details; one went to 61 Prague hotels to ask whether anyone had seen Michael Cohen.

On January 10, CNN’s Jake Tapper announced a big scoop: “CNN has learned that the nation’s top intelligence officials provided information to President-elect Donald Trump and to President Barack Obama last week about claims of Russian efforts to compromise President-elect Trump.” The briefing, CNN reported, was “based on memos compiled by a former British intelligence operative whose past work U.S. intelligence officials consider credible.” The memos included, the network said—ominously and hazily—“allegations that Russian operatives claim to have compromising personal and financial information about Mr. Trump.”

The dossier was in circulation, affecting the course of American politics. Now that CNN had effectively waved it in the air, surely someone, soon, would let regular people in on the secret? I knew what I thought we should do, but I asked Mark; our executive editor, Shani Hilton; and Miriam Elder, the former Guardian Moscow correspondent editing our international coverage, if we should publish it. They all agreed that the document itself was news.

We stood around Mark’s laptop as he started typing. Ken, on speakerphone, warned that we could get sued; I too-curtly told him that I wasn’t asking him for legal advice. Then we turned to writing. “A dossier making explosive—but unverified—allegations” had been in wide circulation, we wrote. The allegations were “specific, unverified, and potentially unverifiable.” Miriam had noticed a couple of odd, minor false notes in the discussion of Russian specifics. She took a turn at the laptop. “It is not just unconfirmed: It includes some clear errors,” we said. I sent a copy of our story to our in-house lawyer.

By 6:20 p.m., about an hour after Tapper’s segment concluded, we had 350 careful words explaining what we knew. In the best traditions of the internet, we published that short introduction alongside a PDF of the full document.

Then I went to stand in the middle of the newsroom and watch the traffic flow.

For the next hour, my eyes flicked between a big screen where I watched the dossier go viral, and my phone, where I watched it dominate Twitter. The tweet that came up the most included a screenshotted excerpt from the dossier describing a “perverted” scene at the Ritz-Carlton Moscow, where Trump had allegedly hired “a number of prostitutes to perform a ‘golden showers’ (urination) show in front of him. The hotel was known to be under FSB control with microphones and concealed cameras in all the main rooms to record anything they wanted to.” That excerpt was shared and shared again. Our caveats didn’t always accompany it.

The news organizations that had accepted Simpson’s invitation to the Tabard Inn were furious. This was, I believe, in part because they had been boxed out of covering the real story by their agreements with a source, but also because they genuinely thought that what we’d done—floating inflammatory, salacious, and unverified claims about the president-elect of the United States—was wildly irresponsible.

Jake Tapper sent me a furious email that evening saying that publishing the document “makes the story less serious and credible,” which was probably true—but if keeping a document secret makes it more credible, you might have a problem. Tapper also said he wished we had at least waited until morning to give his news the attention it deserved: “Collegiality wise it was you stepping on my dick,” he wrote.

I’d expected that backlash, and at first welcomed it. I thought we were on the right side of the decade-old conflict between the transparent new internet and a legacy media whose power came in part from the information they withheld. And, of course, I loved the traffic. This was a huge revelation, a secret unveiled. What made me uncomfortable was the gratitude.

My phone lit up with text messages from Democrats thanking us for publishing the dossier and revealing Trump to be as depraved as they had always believed him to be. Hillary Clinton had never mastered social media; her supporters had never developed the dense networks of memes and conspiracy theories that powered the Trump movement. But now liberals, forming a nascent “resistance,” were starting to build their own powerful narratives on social media that were sometimes more resonant than factual. The notion of a single, vast conspiracy seemed to answer their desperate question of how Trump could have been elected. Russia clearly had helped. WikiLeaks’s hack-and-dump operation was a crucial factor among many in a very close election. You didn’t need to believe all the details in the dossier to know those things.

But perhaps I should have thought a little more about WikiLeaks. A couple of weeks before the 2016 election, I’d attended a Trump rally in Edison, New Jersey, and on my way in, I’d encountered a supporter chanting, “WikiLeaks! WikiLeaks!” I asked him which specific documents he thought painted Hillary Clinton in such a bad light. He didn’t exactly know. I realized that I was looking at social media in real life, a man shouting information cast as a symbol of what he already believed about Clintonian corruption, not as anything meant to convey new knowledge.

Something similar happened with the dossier. We had embedded it as a PDF, which meant that it could travel context-free, without our article’s careful disclaimers, and that’s exactly what happened. I watched uneasily as educated Democrats who abhorred Trump supporters’ crude rants about child sex rings in Washington pizza joints were led by the dossier into similar patterns of thought. They read screenshots of Steele’s report; they connected the dots. They retweeted threads about how the plane of a Russian oligarch—previously unknown to them, now sinister—had made a mysterious stop in North Carolina.

We’d been careful, I found myself having to remind people, to say we didn’t know whether everything in the dossier was true when we published it. I defended the decision in public, in a New York Times op-ed and in a deposition, after a Russian man whom Steele had suggested was tied to the Democratic National Committee hack sued us.

Months after we released the dossier, the media executive Ben Sherwood came by my office. We’d met years earlier when he’d been at Disney, which had been trying to buy BuzzFeed. We had turned Disney down, but it had been a hard decision. I told him that running BuzzFeed had gotten more difficult, with the complexities of management and the realities of digital advertising bearing down on me.

So what did I think now? he asked. Didn’t I wish we’d done the Disney deal? “Would we have been able to publish the dossier?” I asked. “Not in a million years,” he told me. Then I told him I was glad we hadn’t taken the money.

That was an easy position to hold in 2017. It seemed reasonable to argue that publishing the dossier had been, on balance, good for the country. It had blown wide open a Russia investigation and forced voters to ask just why Trump seemed so friendly with Vladimir Putin. But although the biggest-picture claim—that the Russian government had worked to help Trump—was clearly true, the release of Special Counsel Robert Mueller’s investigation in April 2019 did not support Steele’s report. Indeed, it knocked down crucial elements of the dossier, including Cohen’s supposed visit to Prague. Internet sleuths—followed by a federal prosecutor—had poked holes in Steele’s sourcing, suggesting that he’d overstated the quality of his information.

And there had always been a more mundane version of the Trump-Russia story. Trump was the sort of destabilizing right-wing figure that Putin had covertly supported across Europe. Trump’s value to Putin was related not to a secret deal, but to the overt damage he could do to America. And Trump, BuzzFeed News’s Anthony Cormier and Jason Leopold discovered, had a more mundane interest in Russia as well: He had drawn up plans to build the biggest apartment building in Europe on the banks of the Moskva River. The Trump Organization planned to offer the $50 million penthouse to Putin as a sweetener.

That real-estate project wasn’t mentioned anywhere in the dossier. Yet it seemed to explain the same pattern of behavior, without the lurid sexual allegations or hints of devious espionage.

And publishing the dossier wasn’t, in the end, a dagger to Trump’s heart. If anything, it muddied the less sensational revelations of his business dealings and his campaign manager’s ties to Russia. An FBI agent who investigated Trump, Peter Strzok, later said the dossier “framed the debate” in a way that ultimately helped Trump: “Here’s what’s alleged to have happened, and if it happened, boy, it’s horrible—we’ve got a traitor in the White House. But if it isn’t true, well, then everything is fine.”

It was, the reporter Barry Meier wrote, “a media clusterfuck of epic proportions.” The dossier’s overreaching allegation of an immense and perverse conspiracy would, he predicted, “ultimately benefit Donald Trump.”

Six years after publication, I accept that conclusion. And yet I remain defensive of our decision. I find it easiest to explain not in the grandiose terms of journalism, but in the more direct language of respect for your reader. Don’t you, the reader, think you’re smart enough to see a document like that and understand that it is influential but unverified without losing your mind? Would you rather people like me had protected you from seeing it?

Imagine the alternative, a world in which the American public knows that there is a secret document making murky allegations that the president-elect has been compromised, a document that is being investigated by the FBI, that the president-elect and the outgoing president have been briefed on, and that everyone who is anyone has seen—but that they can’t. This would, if anything, produce darker speculation. It might have made the allegations seem more credible than they were.

We faced a difficult series of lawsuits, but we won them all, in part because we’d maintained our journalistic distance. We argued, successfully, that we were not making these claims ourselves; we were making the “fair report” of what amounted to a government document. We’d published the dossier while holding it at arm’s length, noting that we hadn’t been able to verify or knock down its claims—even if we had inadvertently launched a million conspiracy theories in the process.

And that’s the part of the dossier’s strange trajectory that remains most disturbing to me. The way the document became a social-media totem for the anti-Trump resistance rebutted my confidence that people could be trusted with a complex, contradictory set of information, and that journalists should simply print what they had and revel, guilt-free, in the traffic. We seemed to be in an impossible, even dangerous, situation: The public had lost trust in institutions while simultaneously demanding that those same institutions filter the swirl of claims that surround democracy’s biggest decisions.

I have no pat conclusion. If I had to do it again, I would publish the dossier—we couldn’t suppress it, not once CNN had discussed it and its implications on air. But I would hold more tightly to the document, so that no one could read it without reading what we knew about it—that we weren’t sure it was true, and in fact we had noticed errors in it. Releasing a document that could be shared without context—and this is as true of the WikiLeaks material as it is of the dossier—created partisan symbols, not crowdsourced analysis.

In technical terms, that means I wouldn’t simply publish it as a PDF, destined to float free from our earnest caveats. At best, we could have published the document as screenshots attached to the context we had and the context we would learn. Perhaps in some small way, this would have limited its transformation from a set of claims into a banner of the “resistance.” But I’m not under the illusion that journalists could have contained its wildfire spread, any more than I think we could have concealed it.

I’m now leading a news organization, Semafor, that is also rooted in transparency. But I no longer think transparency means that journalists can be simple conduits for facts, obscuring our own points of view, leaving our audiences to figure it out. The best we can do, I think, is to lay our cards on the table in separate piles: Here are the facts, and here’s what we think they mean—and to retain some humility about the difference between the two.

This essay is adapted from the forthcoming Traffic: Genius, Rivalry, and Delusion in the Billion-Dollar Race to Go Viral.