The AI honesty lines

Incumbents lead with AI and agent vocabulary, so we win by being plain-spoken. Mention it plainly, never lead with it, never oversell it.

• "And yes, there's AI under the hood. We've just spent a long time teaching it to sound like you."
• "Inside Maktura every draft is labelled 'Holly, in [name]'s voice', so the person approving it always knows it was drafted by AI. At the levels we recommend, a person approves every post before it goes out."
• If asked which model: "It runs on Anthropic's Claude. The model isn't the point. Anyone can buy a model. What matters is the voice for each person, your rules, the workflow and Holly doing the work." (Who leads on this answer is open: D23.)
• Generic AI content: "LinkedIn has said refining your own words with AI is fine. What it's pushing down is generic output, which per-person voices and checks before posting are there to reduce."
• Does it learn from them: "It learns how each person writes, and you control that: pause voice learning, or mark a post 'don't learn from this'. You can download everything we hold or delete your account from Settings." Anything beyond that (training, sub-processors, data location): "I'll get you that in writing from our legal team."
• Detection: "I won't promise anything about detection. It's built from your writing, so it starts from how you sound, and you see it before it goes."

Rule

Never say: "AI-powered", "agentic", "voice cloning", "undetectable", "beats the algorithm"; anything suggesting readers of a published post can tell it was AI-drafted, or any watermark, fingerprint or disclosure feature (undecided: D9); a mobile app, voice input, an avatar or a hologram.