The Consensus Edit: why your sharpest line never ships

The scene I'm in this week

Last Tuesday I sat in on a homepage copy review for a fintech platform in the $5M-$75M range. Nine people on the Zoom. The document on screen was called 'Homepage Messaging v9 FINAL FINAL' and it carried 214 comments in the margin.

The headline under review: 'Empowering teams with intelligent, end-to-end financial workflows.' Nobody in the room liked it. Nobody objected to it either. That's what version nine sounds like.

I did the thing I always do on these calls. I opened the version history and scrolled back to March. Version one. The founder had typed: 'Your controllers are still closing the books in spreadsheets. That's why your auditors don't trust your numbers.' Rough, specific, a little rude. Two sentences that name a buyer, a villain, and a consequence.

Then I watched six months of edits eat it. Product added 'intelligent' because the claim felt too narrow. Sales wanted integrations mentioned because a deal had stalled on integrations once. Legal softened 'don't trust your numbers' to 'may lack confidence.' A new VP swapped in 'empower' because her last company used it. Every single edit was reasonable. The sum of nine reasonable people was a headline that says nothing to anyone.

The problem on that call wasn't writing talent. Version one proves the talent was there. The problem is the process standing between the true sentence and the live site.

Naming what's actually broken

I call it the Consensus Edit: the review loop where copy circulates until everyone with an opinion has removed the part that worried them, and ships the moment nobody objects. Watch what the approval bar quietly becomes. The question stops being 'is this true and sharp?' and becomes 'is everyone comfortable?' Those are different tests, and they select for opposite copy.

The mechanism is boring, which is why it survives. Every reviewer subtracts risk from their own corner. Product qualifies the claim until it's precise and inert. Sales widens the audience so no deal is ever excluded. Legal hedges the outcome until it promises nothing. Leadership floats everything up one level of abstraction, because vision. No villain in sight, just nine people doing their jobs, and comfort compounding with every pass.

Here's the part that makes it a PitchKitchen problem and a story problem: the Consensus Edit is Solution-Centric Marketing running your approvals. Every seat in that review belongs to someone defending a piece of the solution. The buyer, the person with the bleeding problem version one actually named, holds no seat. I've started calling that the empty chair. Scroll your own comment thread and look for one note written on the buyer's behalf. You'll usually find zero. This is just truth.

There's a companion failure downstream, and I've written about it in How do we know if our messaging actually works, or just sounds good in the room?. That post is about testing what ships. This one is about the machine that decides what ships. Fix the test and keep the machine, and the machine will keep handing you mush to test.

Why this is worse now than ever

AI brought the cost of copy to zero, and the Consensus Edit got a power tool. The committee now has infinite variants to circulate. Each stakeholder pastes the draft into ChatGPT and asks for a 'more professional' pass before commenting. Understand what that does. A language model is an averaging machine trained on everyone. A review committee is an averaging machine staffed by everyone. Chain them together and you ship the average of an average.

The second shift matters more. Your buyers now brief themselves through AI before they ever reach your site, and an engine can only repeat something specific. When a buyer asks ChatGPT what you do, the model works from whatever your public language commits to. Consensus copy commits to nothing, so the engine either mumbles a category description or reaches for the competitor whose site actually says something. I unpacked what 'saying something' takes in How do we differentiate when everyone claims the same benefits?.

And the buyers those engines send are the best ones you'll ever get. Blazly's May 2026 analysis found AI-referred visitors convert at 14.2 percent, against 2.8 percent for Google organic traffic. Machine-briefed buyers arrive pre-sold. They get sent to companies the machine can quote. A headline built to survive nine reviewers gives it nothing to quote.

The diagnostic: is your copy surviving review, or dying in it?

Run these three on your own team this week. No consultant required.

1. 1The Redline Test. Open the version history on your last homepage or deck rewrite. Count the edits that made a claim more specific against the edits that made a claim safer. Healthy teams add specificity in review. If your history runs ten-to-one toward softening, you're not editing copy. You're running the Consensus Edit.
2. 2The One-Owner Test. Ask out loud: who owns the headline? One name is a pass. 'Marketing' is a fail. A meeting is a fail. A sentence with nine owners has none, and copy with no owner defaults to whatever bothers the fewest reviewers.
3. 3The First-Draft Test. Put draft one next to the live page. If the first draft names a sharper buyer, villain, or outcome than what shipped, your process subtracts truth, and everything you route through it will come out grayer than it went in.

What I see across 200+ B2B companies

Here's the pattern I can't unsee after 200+ B2B messaging engagements: the winning line already exists, and it's usually in the deleted history. When I do the archaeology on a company's message, the sentence that ends up anchoring the new homepage traces back to a founder's first draft, a sales call transcript, or an email written at 11pm far more often than to anything produced inside a review cycle. On those engagements my job looks less like writing and more like recovering deleted files.

“The founder who lived the pain builds different than the founder who researched it. The positioning has a clarity no amount of market research produces.”

... a founder thread on Indie Hackers, May 2026

That clarity is single-author. Ten smart people who each half-believe a claim will negotiate it down to something all of them can stand behind, which turns out to mean something none of them would say out loud to a customer. Conviction gets authored by one person who was in the room when the problem happened. Review can check that person's sentence for accuracy. Review can't write it.

It also explains why the message dissolves again every time the team changes. New leader, new review cycle, another pass through the grinder. I wrote about the reset version of this in Why does our messaging start over every time we hire a new marketing leader?. The Consensus Edit is the same disease running between hires: the message erodes a comment at a time instead of all at once.

A real example

A clinical software company, low twenties in revenue, PE-backed. Details blurred, dynamics real. When I arrived, the homepage said 'AI-powered clinical workflow optimization for modern health systems.' Eight words of AI-Parmesan, approved by eleven people. In our first working session the founder said, unprompted: 'Nurses chart for an hour after every shift ends. That hour is why they quit.' I asked when he'd last said that publicly. He laughed. It had been the deck's opening line three years earlier. A board advisor had called it 'too negative.'

We didn't invent a story. We extracted the one he already had, documented it, and changed what review meant. The narrative got decided once, with the founder in the room, and written into the framework. From then on, reviewers checked claims for accuracy against the document. Comfort stopped being a blocking objection. The nurse sentence went back up, essentially as he'd said it.

Within a quarter the tells showed up where they always show up. Discovery calls started with buyers repeating the line back: 'we're the ones with the hour problem.' Reps stopped rewriting the first three slides of the deck. And the CEO stopped fielding the 'so what do you actually do' question at conferences. None of that came from new words. It came from old words, un-deleted, and a process that could no longer kill them.

What this means for you

If your homepage reads like the minutes of a meeting, buyers can hear it, and so can the engines briefing them. The fix costs less than you think, because the truth you need usually already exists in your own document history. Three moves this week:

1. 1Recover one deleted sentence. Open the version history on your homepage doc, find the sharpest true line that got edited out, and put it back for 30 days. Watch what happens on sales calls when your site finally says something.
2. 2Convert reviewers into fact-checkers. One owner decides the message, and it should be the person who lived the problem, usually the founder. Everyone else gets exactly one kind of veto: 'that's untrue.' 'That makes me uncomfortable' stops being a blocking comment.
3. 3Have the argument once, upstream, in a document built for it. Asset reviews turn into philosophy debates when the philosophy was never decided. Decide it once and reviews shrink from taste votes into conformance checks.

That third move is the real one, and it's the work I do all day. The Magnetic Messaging Framework (MMF) is that upstream decision document: a strategic narrative system built around four anchors ... category design, villain framing, an old-way / new-way contrast, and a promised-land outcome ... decided with the founder in the room and written down so the argument never reopens every time someone touches a page. Once the framework exists, your homepage, your deck, your outbound, and every AI tool you point at your brand all write from decided truth instead of open season. The committee loses the power to re-litigate conviction, because conviction is documented.

That's PitchKitchen. I'm Greg Rosner, founder of PitchKitchen and author of Story Craft for Disruptors, and I build Magnetic Messaging Frameworks for founder-led B2B companies in the $5M-$75M range whose sales are stalling because the message isn't doing the work. The whole practice rests on one bet: your sharpest line already exists. My job is getting it past your own review process alive.