Audience voice intelligence: field notes

Why generative AI mistunes for the audiences you're trying to move — and what to do about it.

The authenticity gap: why persuadable voters tune out generic AI copy

AI made the national-headquarters voice cheaper to produce than the native voice. The persuadable voter hears the difference and swipes. A persuasion ceiling dressed up as a platform problem.

AI & Marketing

Why your AI ad copy sounds like an NPR host

Read your last three AI ad variants out loud. If you hear Ira Glass, you have work to do — and it's not a prompt-engineering problem.

AI & Marketing

Why preference-label RLHF has a systematic audience-voice bias

A structural argument: stacking a second LLM to 'fix' bias is bias laundering at higher compute cost. The only durable fix is a signal whose provenance is independent of the rater pool.

Strategy

Thrift stores, vinyl records, and why single-signal AI gets audiences wrong

Three phrases, one market, three confident-but-wrong AI reads. The most valuable output of a multi-signal system isn't the consensus — it's the disagreement.

See your own copy scored against your own audience.

15 minutes. We score a sample of your current copy against a pre-built audience profile, rewrite it in the audience's voice, and show you the delta. If it's not a fit, you keep the benchmark score.

Book a 15-minute demo →