Voice
Set your brand voice
The single most important step. A 90-second investment that prevents every future draft from sounding like ChatGPT.
What we need from you
Three to five posts written in the voice you want MarquIQ to write in. Your own tweets, your own LinkedIn posts, a founder letter, a Reddit comment that got traction — anything between 40 and 400 words per sample.
More is not better
Five great samples beats fifty mediocre ones. The engine picks up rhythm from the best exemplars — dilute it with filler and the output drifts toward average.
What the engine extracts
- Sentence length distribution (are you staccato or flowing?)
- How you open posts (question, stat, personal anecdote, bold claim?)
- Punctuation tells (do you use em-dashes, bullets, line breaks?)
- Word-level idiosyncrasies (slang, hedges, technical depth)
- Closing patterns (call-to-action, reflection, link, nothing?)
Per-platform voice
You can seed different voices per platform. Your LinkedIn voice is probably more measured than your X voice. Paste three LinkedIn-style samples in the LinkedIn tab, three punchier samples in the X tab.
How to refresh
Voice drifts. Every quarter, replace one or two samples with something recent. The engine re-tunes in minutes — no retraining, no waiting.
Pro move
When you write a post that lands (high engagement, comments you like), drop it into your voice seeds. The engine learns from winners, not just averages.
Rules we never break
- No em-dashes (configurable, but the default is no — it is the single biggest AI tell).
- No invented stats or customer quotes. Ever.
- No claims about features your product does not have.
- No hashtag spam on LinkedIn or X. One or two max, contextual only.