The dimensions of writing style
Effective tone matching works across at least five dimensions: formality (the gap between 'Please find attached' and 'here it is'); sentence rhythm (length and syntactic complexity); vocabulary breadth (common words vs. precise rarer ones); social signals (greetings, how you address people, closings); and structural conventions (bullets vs. prose, paragraph length, whether you restate the question before answering).
Social signals are high-visibility and easy to get wrong — a system that defaults to 'Hi [First Name],' when you open with a bare first name produces drafts that feel subtly off even when the body is perfect.
How style profiles are built
Profiling starts with a corpus of your actual sent messages. The system analyses each and computes aggregate statistics across those dimensions, stored as your profile — the extracted patterns, not the raw messages.
Profiles can be static or dynamic; dynamic ones are more useful because they adapt as you keep sending and support context conditioning — if the incoming message is formal, the system checks whether your history shows you matching formal inputs with formal replies, and applies that.
The role of the knowledge base
Tone matching handles how you write; the knowledge base handles what you write about. Without one, a tone-matched draft can sound exactly like you but leave placeholders for facts it doesn't know. In echo the knowledge base is a set of text entries you maintain; when a draft is generated the system checks whether the question maps to an entry and populates the draft accordingly.
Accuracy, limits and privacy
Current tone matching is accurate enough to be useful on the first draft but not perfect: strong on formality, length and sign-offs, weaker on subtle rhetorical habits and humour. Expect to edit — the goal is to reduce editing, not eliminate it.
Privacy is a real consideration, because profiling requires reading your sent mail. A trustworthy system is transparent about what it reads, doesn't share your content with other users, doesn't train shared models on your data, and lets you review and delete your profile. Echo connects to one Gmail account via OAuth and uses your data only for your own drafts.
frequently asked
- How accurate is AI tone matching really?
- On broad dimensions — formality, sentence length, sign-off style — accuracy is high enough that most readers can't distinguish the draft from your real writing. On subtle habits it's good but imperfect; most users edit 20–30% of a given draft.
- Can it handle different styles for different recipients?
- Yes, if the training data includes that variation. A system that's seen you write both casual internal email and formal client proposals can learn to match formality to context — verify this in early drafts and correct where it's wrong.
- What happens to my email data used for profiling?
- It varies by vendor and matters a lot. Look for OAuth access rather than password storage, an explicit no-sharing policy on your content, no use of your data to train shared models, and a defined retention limit.
Published June 9, 2026 · Last updated June 16, 2026