Author voice & integrity

Two per-project settings shape how the AI writes for you: your voice, so the output sounds like you, and your integrityrules, so it never crosses an academic line. Both live in your project’s .multilo/ folder.

Author voice

Multilo can learn your writing voice from samples of your own work and store a description in .multilo/voice.md. From then on, suggestions, chat, and agents write in that voice — your sentence rhythm, your register, your habits — instead of generic model prose.

Integrity guardrails

Research-integrity rules live in .multilo/integrity.mdand are enforced across everything the AI writes. They encode the non-negotiables of academic work: support claims with real sources, don’t fabricate citations, and be honest about the limits of what you can verify.

Strict mode

With strict mode on, the AI refuses to fabricatea citation rather than inventing one to fill a gap. If it can’t find a real source, it says so — which is exactly the behavior you want when your name is on the work.

# Research integrity
- Cite only real sources from the project library.
- Never invent authors, journals, or page numbers.
- State clearly when a claim cannot be verified.

Where it applies

Voice and integrity are loaded once and injected everywhere the AI acts: chat, inline suggestions, and every agent, including the Full Draft Writer.