Take a verbose draft, get back a tightened version — same meaning, fewer words.
Paste a draft that runs too long. We tighten it while preserving every specific name, number, deadline, and URL verbatim. Pick how aggressive: mild (~20%), moderate (~35%), aggressive (~50%).
The single biggest predictor of reply rate on cold outreach isn't tone, personalisation, or A/B-tested subject lines — it's length. Below 100 words consistently outperforms above 200 across every dataset we've seen (Lavender, Outreach, Mailshake's public studies, our own data). The reason is simple: recipients triage their inbox by skim, and a six-line email gets read fully where a twelve-line email gets archived.
The model is constrained to preserve the parts that change meaning: specific names, dates, numbers, and URLs are kept verbatim. What gets cut is the filler — "I hope this finds you well", repeated context, soft caveats, redundant pleasantries. The tone is preserved (friendly stays friendly, formal stays formal) but the words-per-point ratio shrinks.
A subtle gain at the deliverability layer: shorter emails ship cleaner through filters. Spam classifiers weight text-density and structure; image-heavy and copy-heavy emails both score worse than tight, prose-only short emails. Shortening isn't just a UX win, it lifts inbox placement.
No — the system prompt explicitly forbids changing instructions, actions, or factual claims. If the model can't shorten without losing meaning, it returns close to the original length.
Yes — those are preserved verbatim by the prompt. If something specific gets mangled in practice, it's a prompt-tuning bug; tell us.
Use the AI Email Writer for a fresh draft. The shortener only goes one way.
Not in the free version. Paste both into a diff tool if you need a side-by-side; or just trust the prompt and re-read the output.
Tight prose is half the win; the other half is sender reputation. SieveGuard's Domain Health add-on monitors SPF/DKIM/DMARC on every sending domain so your shorter, sharper emails actually land.