AI editorial feedback is written commentary from a language model on how a piece of work could be stronger, rather than a decision about whether it is good enough. On Kind Channel it assesses five criteria — clarity, originality, social impact, engagement and feasibility — shows its reasoning, and never blocks a submission.
Diagram of the 5 areas this page covers: The distinction that matters: editor, not gatekeeper; The five things it assesses; How to read a low score without over-correcting; Why it arrives before publication rather than after; What it cannot judge.
Two very different things get called "AI review". The first is a filter: the model scores your work and something is allowed through or not. The second is an editor: the model reads your work and tells you what it noticed, and you decide what to do about it.
Kind Channel uses the second. No submission is rejected by the AI. It produces notes, you may revise and resubmit as often as you want, and human community votes — not the model — determine which ideas get scheduled.
Each criterion is scored and, more usefully, explained in plain language so you can see the reasoning rather than just a number.
What each criterion means
| Criterion | The question it answers |
|---|---|
| Clarity | Can a reader tell what this idea actually is? |
| Originality | Is this a fresh angle, or a familiar one restated? |
| Social impact | Does it address a real human or community need? |
| Engagement | Would an audience want to watch or take part? |
| Feasibility | Could this realistically be produced? |
A score on its own is close to useless. The reasoning underneath it is the part worth reading, and the common failure is treating a low number as an instruction to delete whatever caused it. Most low scores are pointing at something missing rather than something wrong, and the difference decides whether your revision helps.
Clarity notes rarely mean the sentences are badly built. Far more often the proposal has described a subject area rather than a specific thing — "a piece about loneliness in older people" names a topic, not a programme. The fix is to add the specifics that are absent: who is in it, where it happens, what somebody watching would actually see. Rewriting the same sentences more elegantly will produce the same note again.
An originality note is not a verdict that a subject has been exhausted. It usually means the angle in the proposal is the one most people would reach for first. The subject can stay exactly as it is; what needs to change is which part of it you are pointing at. A familiar subject approached through one specific person, place or afternoon reads as original in a way the general version never does.
Feasibility notes tend to appear when a proposal describes an outcome without describing the labour. Saying a piece will bring a community together is a hope. Saying you have already spoken to the three people who would appear in it is a plan. Adding what access you genuinely have, and what you have already arranged, shifts this note far more reliably than scaling the ambition down does.
Almost every other platform gives you feedback after you publish, as metrics and comments. By then the work is finished, every decision is already made, and the information arrives too late to act on. You learn that something did not land without learning why, which is the least useful form feedback can take.
Editorial feedback before submission inverts that timing. The notes arrive while the idea is still cheap to change — before filming, before anyone has given up an evening, before you have asked somebody to appear in something you have not thought through. That timing is most of the value on offer here, and it rewards being used properly: resubmission costs nothing, and the second set of notes on a genuinely revised idea is usually sharper than the first.
A language model has no way to know whether your account of your own life is true, whether a community actually wants the thing you are proposing, or whether you personally can deliver it. It is reading text, not verifying claims.
It is also working from patterns in what has been written before, which means genuinely unusual ideas can read as unclear to it. A low clarity note on an unconventional idea sometimes means the idea needs explaining rather than changing — that judgement is yours, which is exactly why it does not get a veto.
On Kind Channel, typically around twenty seconds from submission to written notes across all five criteria.
Not on Kind Channel. The model never blocks a submission. It returns notes, and you choose whether to revise. Community voting decides what gets produced.
It carries the biases of the text it was trained on, which is why the reasoning is shown rather than hidden behind a score — a visible argument can be disagreed with. Community voting acts as a second, human check.
No, and it is a trap worth naming. Optimising for the notes produces ideas that score well and interest nobody. The feedback is most useful for finding genuine ambiguity in your writing, not as a target to hit.