AI won't replace journalism — it forces it to be better

Every few months I am asked, usually by someone who has never set foot in a newsroom, whether artificial intelligence is going to put journalists out of work. The honest answer is that it already has — just not the journalists most people are picturing. The first casualties are not the investigative reporters or the foreign correspondents. They are the content mills: the offshore rewrite farms, the algorithmic aggregators, the websites that exist only to repackage someone else's reporting into search-engine bait. A machine does that work now, faster and cheaper, and frankly without much loss to the world. I cannot bring myself to mourn it.

What that collapse does, though, is far more interesting than the panic suggests. By making the cheap layer of the news essentially free, AI strips away everything that was never really journalism in the first place and leaves, in sharp relief, the part that always was. When summarising a press release costs nothing, summarising a press release is worth nothing. The premium shifts, abruptly and permanently, to the things a model cannot do — and once you list them honestly, that list turns out to be the whole point of the craft.

What the machine can't reach

A model cannot be in the room. It cannot sit across a table from a frightened source and earn, over months, the trust required to be told the thing no one is supposed to say. It cannot stand in the rubble and describe what the official statement omits. It cannot read a face, weigh a silence, notice that the number in the third paragraph of the report contradicts the one in the appendix. It cannot be held accountable — cannot stake a reputation, cannot be sued, cannot stand behind a sentence and say, on the record, this is true and I am the one who checked. Everything that distinguishes journalism from text generation lives in exactly those gaps.

This is why I find the doom-mongering not just wrong but backwards. For two decades the economics of the web pushed newsrooms in precisely the wrong direction — toward volume, toward speed, toward the cheap and the churned, because that was what the click economy rewarded. AI does not deepen that trap; it detonates it. You cannot win a race to the bottom against a machine that has no bottom. The only viable response is to climb in the opposite direction, toward the work that is expensive precisely because it requires a human being to do it — the reporting, the presence, the judgment, the willingness to be wrong in public and answer for it.

Lean into the difference

None of this means newsrooms should pretend the tools do not exist. Used well, they are a genuine gift to the parts of the job that were always drudgery — transcribing hours of interviews, searching ten thousand leaked documents for the one that matters, translating a source's testimony, building the chart that makes an investigation legible. I want my reporters spending less time on the mechanical and more on the irreplaceable. The trap is not the tool. The trap is using the tool to make more of the cheap thing, rather than to make room for the costly one.

So no, I am not afraid of the machine, and I think a newsroom that is afraid has misdiagnosed its own value. The reckoning AI forces is uncomfortable but clarifying: it asks every outlet to prove it was ever doing journalism at all, or merely manufacturing words. The ones that were only making words will lose, and should. The ones that were doing the real thing — being there, asking the question, standing behind the answer — have just watched their competition for attention thin out and the value of what they alone can do rise. That is not the end of journalism. It is, if we are honest enough to seize it, the thing that finally forces journalism to be worth the name.