SEO & Content
Forty assets a month per client with a 22% rejection rate at the editorial gate. The rejection rate is the interesting number, because a system with no rejections is not a quality system.
The problem
Sludge reads fine and converts nothing
Sludge is not bad grammar, models solved that. It is text that takes 400 words to say something worth 90, hedges every claim, and could have been written about any company in the category.
Our gate is tuned for exactly that failure. Generation is cheap, so the constraint is judgement: a human picks the argument, and an editor has authority to reject with a reason that feeds back into the spec.
What you actually get
Every engagement ships these. No line items you cannot point at.
Evidence layer
Client data, transcripts, support tickets and analytics. No claim enters a draft without a source here.
Angle selection
A human picks the argument for every asset. This is the step we have never successfully automated.
Draft pipeline
Model-generated structure and prose against a house style spec, with claim-to-source mapping preserved.
Editorial gate
Claims checked, padding cut, approve or reject with a reason. Rejections route back to the layer that caused them.
Typical outcomes
0 assets
Shipped per client per month
0%
Rejected at the editorial gate
−0%
Cost per published asset
How we run it, week by week
Gate first
We write down what a rejection means and who can issue one, before a single prompt exists.
Evidence and spec
House style, claim sourcing rules and the evidence corpus assembled with your team.
Production ramp
Volume raised only as the gate holds. Quality is the constraint, not capacity.
Performance review
We measure assets that earn a citation or conversion within 60 days, not raw publishing count.
Stack we work in
Platform-agnostic. We work in whatever you already own and tell you honestly when it is the constraint.
Questions buyers ask us
AI is in the draft layer. The argument, the evidence and the approval are human, and 22% of drafts do not survive review. If that ratio drops we treat it as the gate failing, not the pipeline improving.