Everyone is a publisher. Nobody is an editor.

By Ananya RaoData editor

Ananya Rao runs the data desk at The Ledger and has spent nine years arguing that most charts are too small.

11 August 2026Aug 11, 2026, 8:00 AM6 min read

A stack of folded newspapers photographed from the sideUnsplash

The internet solved distribution so completely that we stopped noticing it had ever been a problem. Anyone can publish to everyone, instantly, at zero marginal cost. This was, and remains, an enormous good. It is also only half of what a publishing house did.

The other half was editing, and editing did not get cheaper. It could not, because it is a person reading carefully and then having an uncomfortable conversation. There is no economy of scale in an uncomfortable conversation.

The three jobs nobody replaced

An editor did three distinct things, and we tend to collapse them into the first. Copy editing catches the errors. Line editing makes the sentences work. Structural editing decides whether the piece should exist — whether the argument holds, whether the reporting supports it, whether anyone needs this.

Software has been thrown at all three. Spellcheck handles the first, and handles it well. Language models make a decent run at the second. The third has resisted entirely, and it is the one that matters most, because it is the only one that can produce the answer no.

A platform can tell you your post did badly. It cannot tell you, before you publish, that it is not worth publishing.

What an unedited internet looks like

It looks like this: enormous volume, competent prose, and a persistent low-grade sense that most of what you read did not need to be written. Not false, exactly. Just unnecessary, and confidently so. The style is fine because the style was the part machines could help with.

The institutions that still employ editors have noticed that this is now a differentiator, which is a strange thing for basic quality control to become. Being obviously edited reads as a signal, the way a hardback does. That is a sad way to arrive at a business model, but it is a business model.

The optimistic reading is that editing was never really institutional. It was always just a second person who cared and was permitted to be blunt. That arrangement does not require a building. It requires a norm, and norms can be rebuilt faster than industries.

Is editing really disappearing, or just moving?

Both, unevenly. Line editing has largely survived inside institutions that still employ subeditors. Structural editing — the work of deciding whether a piece should exist in this shape at all — has thinned out almost everywhere, because it is invisible when done well and expensive to staff.

Can a language model do the editor’s job?

It can do the copy pass credibly and the structural pass poorly. A model can tell you a sentence is overlong. It cannot tell you the piece is dishonest, because that judgement depends on knowing what the writer left out, and the model only sees what they put in.

What is the cheapest useful substitute?

One named reader who is allowed to say the piece should not run. Not a comment thread, not a review queue — a single person with the standing to kill it. Most of the value of an editor is the credible possibility of no.

Post on X LinkedIn