How it works
Sample extraction, LLM attributes, confidence (coming soon).
How Brand DNA is computed
Coming soon: Brand DNA writing-style analysis and AI drafting are not available yet. This page describes how they will work.
- MySocials collects samples of your writing from your connected accounts.
- An LLM extracts style attributes — tone, sentence length, humor, formality, signature phrases — stored as structured JSON with per-attribute confidence scores.
- Deterministic stats (readability, length distributions) are computed separately and never by the model.
- The profile refreshes as new samples accumulate; each build records which model produced it and how many samples it saw.
Confidence
Thin sample sets produce lower confidence. The dashboard shows how much data the profile rests on — a profile built from 50 posts is steadier than one from 3.