Ship more content —without drifting off your own voice.
Lock your brand voice once, from your real writing. Four Claude Skills draft multi-channel content from a single anchor. Then a deterministic guard scores every draft against the locked profile — ON-VOICE, DRIFTING, or OFF-BRAND by its weakest dimension — with an anti-slop gate that catches machine-wash the dimensions miss.
AI lets you publish ten times more — and sound like everyone else ten times faster.
Mark each draft against your locked profile. The weakest dimension sets the read.
Tap a dimension to cycle 0 → 1 → 2; tap the slop flag to toggle it. The read is the weakest of the six dimensions — one off-brand line makes the whole draft off-brand. The slop flag demotes a draft that reads like a machine even when every dimension looks on-voice. Grades a draft, not a person. Not legal advice.
A zero-dependency engine returns the identical verdict from your marks.
BRAND-VOICE DRIFT GUARD -- read as of 2026-07-04
==============================================================
LinkedIn post — launch 100 ON-VOICE
fix first: Register & tone match
Email — nurture #2 100 DRIFTING [anti-slop gate: demoted]
fix first: Register & tone match
X thread — hot take 83 DRIFTING
fix first: Lexicon & signature terms
Blog intro — repurposed 67 OFF-BRAND
fix first: Claim & fact fidelity
--------------------------------------------------------------
BATCH ROLLUP: VOICE BROKEN
Revise first: Blog intro — repurposedThe nurture email scores 100 and is on-voice on all six dimensions — but its slop flag is set, so the anti-slop gate demotes it to DRIFTING. The weakest-dimension rule alone would have shipped it. The blog intro scores 67 but has an invented claim (fidelity=0), which forces OFF-BRAND. Sounding on-brand and sounding like a machine are two different tests, and this runs both.
Three rules keep the voice honest at volume.
The voice is extracted from your real writing and locked. Every later draft is graded against that anchor — not a generic default.
The read is the lowest of six dimensions, never the average. One off-brand line makes the draft off-brand, because that's how a human reads it.
The Python engine, the Excel workbook, and this page's demo run the identical read and the identical anti-slop gate, byte-for-byte.
A drafting-and-drift working aid — not an autopilot, and not a people-scorer.
- A locked brand-voice profile drawn from your own writing.
- Multi-channel drafting from one anchor, in that voice.
- A reproducible ON-VOICE / DRIFTING / OFF-BRAND read per draft.
- An anti-slop gate that catches machine-wash the dimensions miss.
- Not an autopilot — nothing is scheduled, sent, or published.
- Not a claim generator — an invented claim scores 0 and forces OFF-BRAND.
- Not a score of any person — it grades a draft, an artifact.
- Not an AI detector, and not legal advice.
Grades a draft, not a person. This System is a drafting-and-drift working aid. It scores a draft against a brand-voice profile you locked — an artifact — and does not score, rank, or profile any individual. It schedules, sends, and publishes nothing; it is not an AI-detection tool and not legal advice. You review and ship.
Anyone accountable for sounding like themselves at scale.
- · Founders publishing across five channels a week
- · Content & social leads managing a voice across writers
- · Agencies protecting a client’s voice at volume
- · Solo operators repurposing one anchor many ways
- · Ghostwriters proving fidelity to a locked voice
- · Anyone tired of AI-shaped sameness in their feed
Where the voice work connects.
The honest answers.
The Brand Voice Engine codifies your voice into a spec and checks a draft against it. The Anti-Slop Content System is a style guardrail on the way in. This is the drafting engine with a drift gate built in: you lock a voice profile once, the Skills draft multi-channel content from a single anchor, and a deterministic guard scores every draft against the locked profile — with the anti-slop gate as a second, separate test. It multiplies at volume and catches drift from the anchor.
Because drift is a weakest-link phenomenon. One off-register line, one banned phrase, or one invented claim reads as off-brand to a human no matter how good the other five dimensions are. So the read is the lowest of the six marks — ON-VOICE only if every dimension holds. The 0–100 score is shown for context, but it never sets the verdict.
Machine-wash. A draft can match your register, lexicon, rhythm, stance, claims, and channel — score on-voice on all six — and still read as generic AI: hedge-stacking, empty transitions, “in today's fast-paced world” filler, a conclusion that just restates the intro. Sounding like your brand and sounding like a machine are two different failures. When the slop flag is set, the read is demoted one band. In the shipped sample, a nurture email scoring 100 on every dimension is demoted to DRIFTING for exactly this reason.
No. The drafting Skill keeps every factual claim traceable to your source anchor. If a channel seems to need a claim the anchor didn't make, it flags it for you rather than manufacturing it — and in the drift guard, an invented claim is scored 0 on claim fidelity, which forces the draft OFF-BRAND.
Yes — byte-for-byte. The Python engine, the Excel workbook, and the on-page demo run the identical weakest-dimension read and the identical anti-slop gate. The workbook reproduces every read from your own marks with live formulas, so you can work in Excel, Google Sheets, or Numbers and get the same verdict.
Neither. It grades a draft against a profile you locked — an artifact — and never a person. Nothing is scheduled, sent, or published: the Skills draft and stage, and you review and ship. It's a drafting-and-drift working aid, not legal advice.
Publish at volume.
Still sound like you.
One purchase, lifetime access, 12 months of updates. $99, once.
Grades a draft, not a person. A drafting-and-drift working aid that scores an artifact against a profile you locked. It publishes nothing on its own and is not an AI detector or legal advice.
Sold by RedHub AI LLC · Secured by Stripe · redhub.ai