How Cogent works

Comment sections have been broken for thirty years. Not because people are terrible — because the design rewards volume over quality, and the platforms that own them have no financial reason to fix that. Cogent is built on a different premise: that most people, given a real reason to think before they post, will actually think.

Editor, not bouncer

Every other platform that's tried to fix comment sections has used the same tool: a bouncer. Draw a line, delete what crosses it. The problem is that a bouncer can only catch what's already obviously wrong — slurs, threats, spam. It can't catch the confident paragraph that states a contested claim as settled fact, or the reply that argues against a point the essay never made. And it teaches the person who wrote it nothing.

Cogent uses an editor instead. Before a comment posts publicly, an AI reads the full essay and your comment together and asks the same questions a sharp editor would: Are you responding to what was actually written? Do your facts hold up? Does your argument follow from your evidence? If something's off, you get a private note explaining exactly what and why — with a suggestion for how to fix it. Your original comment stays in the box, unchanged. You revise and resubmit as many times as you need. Once it clears, it posts.

Nothing gets deleted. Nothing gets hidden. The reviewer isn't judging your opinion — it's checking whether the case you're making actually holds together.

What happens when you comment

You write your comment and hit post

Nothing goes public right away. The comment goes to the AI reviewer.

The AI reads the full essay and your comment together

It holds the entire piece in context — every claim, every piece of evidence — alongside what you wrote. It asks whether your comment engages with what was actually said, whether your facts are accurate, and whether your argument holds up logically.

If something needs work, you get a private coaching note

The note explains specifically what's off and suggests how to fix it. Your original comment is still sitting in the box. You can edit in place and resubmit — as many times as you need.

Once it clears, it posts publicly

Every posted comment includes a "How was this reviewed?" link showing the full AI verdict, so readers can see exactly what standard was applied.

What the AI is actually looking for

The reviewer has no stake in which side of an argument you're on. It applies the same standard to every viewpoint.

PostsA sharp disagreement with the essay's central argument, clearly stated

PostsA challenge to the evidence backed by a real counter-source

PostsA blunt, unpopular point that genuinely engages with what was written

Gets coachingAttacking the author's motives instead of their argument

Gets coachingStating a disputed claim as settled fact

Gets coachingArguing against a position the essay never actually took

Gets coachingCiting a source that doesn't say what you're claiming it says

Real identity

You sign in with Google, Facebook, or LinkedIn. Your display name comes from whichever provider you use. Anonymous posting is not available by default — your name sits next to everything you publish.

You can request a name change, and it will be approved. We're not in the business of forcing anyone to expose their legal identity. But we keep a deliberate step in the process: the name you land on stays attached to your comments going forward. You can't be a different person in every thread.

Transparency

Every posted comment carries a "How was this reviewed?" link. Click it to see the full AI verdict — what was checked, what the reviewer found, and how the comment was evaluated. Authors can also see the AI review of their own essays before they publish.

The goal isn't a black box that quietly filters comments. It's a visible, auditable standard that every reader can inspect.

Want the full argument for why we built this? The three essays that explain the problem, the philosophy, and the design are published on Cogent itself. Browse essays and look for the Cogent series.