Editing AI drafts so they sound like you

Can I edit an AI draft before it posts?

Yes: every AI draft is an ordinary draft you can change freely until you schedule it. Nothing posts until you click Schedule or Post now, so the version that goes out is always one you’ve seen.

Click Edit on a card to open the draft. Changes save as you type, and you can change the text, images, first comment, and the channel’s own settings, like the subreddit on Reddit.

Once a post is scheduled, the editor locks and says it can’t be edited now. To change it, open the post in Posts and click Cancel scheduled post. It goes back to being a draft you can edit and schedule again.

Read it out loud

The quickest test of a draft is whether you’d say it. If a line sounds like something you’d never write, change it or cut it. Small edits, like your own opening line or a phrase you always use, make a big difference.

A common example: if you’d write “rates went up again” and the draft says “interest rates saw another increase”, use your version. Check the opening line most of all; on most platforms it’s the only part people see before deciding to read on.

What does “Possible AI tell” mean?

It means Karrow found something in the draft that tends to make writing read as machine-made, and its own rewrite didn’t remove it. Karrow checks every draft for long dashes, 24 overused words like “delve”, “leverage”, “robust”, or “game-changer”, 14 stock phrases like “Let’s dive in”, “Here’s the thing”, or “Let that sink in”, and links in the body of the post.

It rewrites once to remove them. If any are still there, the draft shows a “Possible AI tell” flag. Hover the flag on a card to see what it found. In the editor, the note lists them so you can fix each one. It’s worth a quick edit before you publish.

The flag is a note, not a block: you can still schedule a flagged draft. It also doesn’t update as you type. It lists what was found when the draft was written, so after you fix something, check the text itself rather than waiting for the note to clear.

Rewrite part of a post with AI

Select some text in the editor and a small menu appears above it. Pick Shorten, Add hook, Punch it up, or Fix grammar, or type your own instruction in the “make it…” box and press Enter.

Karrow rewrites only the part you selected and drops the new text in place, with a brief highlight so you can see what changed. Add hook writes a new opening line and keeps your original text under it. Don’t like the result? Press ⌘Z (Ctrl+Z on Windows) to undo it in one step.

Shorten is given the channel’s character limit to aim for, 280 on X or 3,000 on LinkedIn. While “Rewriting…” shows, the editor is paused so your typing can’t land in the middle of the change.

You can select up to 2,000 characters at a time. Inline edits have their own daily limit, separate from drafts: 60 a day on the free plan and 600 on Pro.

Should I edit or regenerate?

Edit when the idea is right but the wording isn’t; regenerate when the angle itself is wrong. On the board, set how many options you want and click Regenerate for that channel. It replaces that channel’s AI drafts and keeps anything you wrote by hand or already scheduled.

Regenerate also replaces AI drafts you’ve edited but not scheduled, so schedule the ones you want to keep first. Drafts you made with Duplicate to are kept. A Regenerate uses the library’s top formats for that channel, not a saved playbook you picked the first time.

If only one option is off, delete it instead: click the bin on its card and confirm “Delete this option?”. The rest of the channel stays as it is, and the deleted draft tells Karrow what you didn’t want.

Each draft counts towards a daily limit: 20 a day on the free plan, 200 on Pro. A Regenerate at 3 options uses 3.

Does Karrow learn from my edits?

Yes: when you schedule a draft you edited, Karrow learns from what you changed, so future drafts on that platform start closer to how you write. It also takes note of the AI drafts you delete or regenerate away, and the ones you schedule untouched.

This is kept per person, per publication, and per channel, so what you change on LinkedIn doesn’t shape your X drafts, and two people in one workspace each get their own. Karrow updates its notes after every three new signals. Posts you wrote by hand don’t count, because there is no AI version to compare them with.

Something you always fix? Add it as a rule in the publication’s Voice tab instead, so it applies every time. Your rules always win over what Karrow has picked up. To start over, click Reset learned tone on the same tab; your rules stay as they are.