The first time I read a paragraph I’d written, then ran through ChatGPT, then read back the model’s edit — I noticed something missing. The grammar was tighter. The argument was clearer. The voice was gone.
That experience isn’t unusual anymore. It’s the default. Anyone who writes for a living has had it. AI is in every editor, every browser, every drafting tool. The output is competent. It is also, increasingly, indistinguishable from every other writer’s AI-assisted output.
So here’s the question this product exists to answer: when you use AI to draft, edit, or polish your writing, does your voice survive?
Default-mode AI prose is safe. It hedges in the same places. It transitions with the same connectives. It reaches for the same competent vocabulary band. It’s not bad writing; it’s neutral writing. And neutral writing, in volume, is the death of authorial voice across whole professions — academics whose papers all sound the same, marketers whose blog posts blur together, pastors whose sermons could have been generated by any other pastor’s prompt.
Voice isn’t an accessory. It’s the load-bearing element of writing that has a person behind it. Voice is sentence rhythm — the way you swing between an 8-word punch and a 26-word digression, or never do that, or only do it on the third paragraph of an argument. Voice is what you’re willing to hedge on, and what you’ll commit to. Voice is the words you reach for and the ones you don’t trust. Voice is what your reader recognizes when they pick up something you wrote and know it was you before they see the byline.
We don’t think AI is the enemy of that. We use AI in this product. AI is genuinely good at making prose grammatical, eliminating redundancy, smoothing a draft. It’s bad — by default — at keeping the writer in the writing. That’s not a model problem; that’s a prompt problem. Asked correctly, given a profile of your voice, the same model can produce text that sounds like you, with the AI tells removed.
That’s what Inksong does. It takes AI-generated or AI-assisted text and rewrites it through a lens you provide — your tone, your domain, your accumulated style features — so the output reads as a continuation of your writing rather than a replacement of it. AI text in. Your voice out.
We’re aware of the failure mode: a tool that promises to make AI-generated work pass as human can be used badly. We’re not naive about that. But the people who write fast and well already use editors and editors-in-software; what we offer is a domain-aware editor that defaults to honoring your style rather than the model’s. The students cheating on essays were going to cheat anyway. The pastors, researchers, novelists, marketers, and translators doing real work — we built this for them.
Use the tools. Don’t sound like them.
AI text. Your voice.