Tools that analyse your writing, not replace it
Most AI writing tools generate prose. These read what you already wrote and show you something you could not see yourself. That is a harder thing to sell and a much easier thing to test — which is the point.
Why this exists
The most common note in manuscript critique is that characters say exactly what they feel. It is also the hardest note to act on, because you know what the line is for — so you cannot hear how flat it lands.
That is a diagnosis problem, not a generation problem. A model that rewrites the line for you solves neither: you learn nothing, and the voice on the page slowly stops being yours.
So these tools read a scene and show you what is underneath it. The rewriting stays yours.
We don't write your prose
These tools read what you already wrote and tell you something about it. Nothing here generates a draft for you, because a draft you didn't write teaches you nothing and drifts toward the model's voice rather than yours.
We publish the losses
A result only counts as measurement if it can come out against us. When one does, it goes on this page rather than into a drawer — there is an example further down.
Honest baselines
The thing to beat is usually a good frontier model with a well-written prompt, not a named competitor chosen because it is easy to beat. If we can't beat the obvious free option, there is no product.
Cited, not affiliated
Where these tools lean on story theory or psychology, that is published work by people who are credited here and have no involvement with us. We claim the engineering, not the ideas.
What we actually ship
Two tools. Not a suite, not a platform — everything else on this site is either research or a bet still being tested.
Inner Monologue Dramaturg
Scores every line of a scene for how on-the-nose it is, and shows what the character is thinking underneath it.
Truby's 7-Steps Validator
Checks a story against the seven steps, and tells you which ones your draft has not earned yet.
The last time we tested this, we lost
In July 2026 we ran a blind test on what was then the flagship feature: an editor that made surgical, word-level changes to keep the author's own sentence intact, instead of rephrasing the whole paragraph the way generic AI tools do. Judges compared both without being told which was which.
They preferred the generic rephrasing in nearly every case.
That was not a bug to fix. Keeping the author's sentence intact was the product thesis, so the test falsified the idea rather than the execution. We stopped building it.
It also taught us something about testing. Neutral readers and AI judges both reward fluent rephrasing, because fluency is the thing they can see — neither of them is judging a voice they have any stake in. So the current bet is judged differently: by authors, on their own manuscripts, where they are the only people who can tell whether it still sounds like them.
See how the current bet is being testedWho builds this
A very small operation — small enough that "the team" would be overstating it. There is no research division here and no cognitive scientists on staff. Where these tools lean on story theory or psychology, that is published work by people who are credited and have nothing to do with us.
What that buys you is on the section above: when a result comes out against the product, there is no roadmap committee to overrule it.
Try the thing that works
The Inner Monologue Dramaturg runs on a scene of your own in about a minute. No account needed to see what it does.
Nothing to buy yet — we'll email when there is.
Writers With AI is not affiliated with or endorsed by John Truby, Linda Seger, Ivana Chubbuck, Dramatica, or any other named methodology or framework. All references are to public domain or widely published aspects of these approaches, with our own engineering on top.