Sales and outreach
Make AI-written outreach sound like a person wrote it
Personalized outreach only works when it reads like a person took the time to write it. If your sequences are AI-generated, DraftFilter is the quality gate that catches the messages that sound like a template before they reach a prospect.
The problem: templated email gets deleted
AI makes it cheap to generate a personalized-looking email for every prospect on a list. It also makes it cheap to generate the same stilted opener, the same flattering-but-generic observation, and the same call-to-action phrasing a thousand times. Prospects read a lot of email; writing that sounds machine-generated reads as low effort, and low effort gets deleted.
The fix is not less automation — it is a quality bar. Before a sequence goes out, you want to know which messages, and which sentences inside them, sound like AI instead of like your team.
How it works: scan, heatmap, rewrite
1. Scan a message
Paste a generated email and get a 0-100 AI-likelihood score with a per-sentence breakdown, so a rep can sanity-check a single high-stakes message in seconds.
2. See which lines sound robotic
The heatmap highlights the sentences that read most machine-written — usually the opener, the transition into the pitch, and the sign-off — so you fix the lines that matter instead of rewriting the whole message.
3. Rewrite and rescore
The rewriter reworks the flagged sentences and scores the new version, so you can iterate until the message sounds human and keep the best variant for the sequence.
No black box: read how the detector scores text, including what the score means and where it can be wrong. It is a probabilistic signal for review, not proof of authorship.
At scale: gate whole sequences through the API
Nobody pastes five hundred emails into a web form. On the Pro and Max plans, the same detector is available as a JSON API: submit a batch of drafts as one score job, poll it, and get a per-item is_ai score back. Gate your sending pipeline on that score — messages that come back low go out, messages that come back high get routed through the rewrite API or back to a human.
Idempotency keys make retries safe inside an automated pipeline, and the OpenAPI spec and docs console let you try the endpoints before writing integration code.
curl -sS https://api.draftfilter.com/v1/score-jobs \
-H "Authorization: Bearer $DF_API_KEY" \
-H "Content-Type: application/json" \
--data '{
"idempotencyKey": "outreach-batch-014",
"items": [
{ "id": "email-1", "text": "First outreach draft..." },
{ "id": "email-2", "text": "Second outreach draft..." }
]
}'Score an outreach draft now
Paste one message and see which sentences sound like AI. Free scans, no account required; API access comes with Pro and Max.