FlowLabs

Prufrock

Does this sound like you?

AI detection tools ask the wrong question. "Was this written by AI?" doesn't matter. What matters is whether this sounds like it came from you.

Your clients know your voice. Your partners know how you write. The person reading your deliverable can feel whether the words carry your experience or whether they were assembled from a training set that learned from everyone and sounds like no one.

That gap between "correct" and "yours" is where trust breaks.


What Prufrock does

Prufrock ingests your actual writing — emails, docs, transcripts, published work — and extracts a voice profile. Sentence patterns, vocabulary fingerprint, paragraph architecture, regional markers, professional tribe language. The patterns you don't even know you have.

Then it scores any new document against that profile. Not "is this AI?" but "is this you?" Line by line, with a fidelity score and specific flags where the voice drifts.


How it works

You send me your writing. I extract your voice profile, score your AI output against it, and hand you back everything you need to make your AI sound like you.

Voice extraction. Send 15–20 documents you've written. Emails, deliverables, blog posts — anything in your natural voice. I extract your profile and deliver it back with a full breakdown of your patterns.

Fidelity report. Send me your AI-assisted output. I score it against your profile, flag every line that drifts, and tell you exactly what to fix. You get the report and the instructions to feed your AI.

Recalibration. Your voice evolves. New clients, new domains, new patterns. Come back when your AI starts drifting again. Same process, updated profile.


Built from personal friction

I run multiple AI systems. One of them wrote "don't rock the ship" instead of "don't rock the boat." Two review systems missed it. I caught it because something in my chest said wrong.

That six-word error opened a door into forensic linguistics — indexicality, shibboleths, idiolect, formulaic sequences. Turns out linguists already had names for what I was sensing. I built the tool that morning.

Read the full story


Built for

Consultants and advisors whose clients know what they sound like. Law firms where voice consistency is a credibility signal. Solo practitioners who use AI to write but send under their own name. Executive teams whose comms get scrutinized by boards and regulators.

Anyone who uses AI to write and can't afford to sound like it.


If you're sending AI-assisted writing under your own name, I want to hear from you.

Get in touch  ·  Follow on Substack