“How does the approve and reject loop make my posts better?”
How does the approve and reject loop make my posts better?
Short answer
Approve what sounds like you, reject what does not, and always reject for a reason you would repeat. Threadgehog compares your approvals and rejections, rewrites its rules for the next batch, and sends up to 8 of your recent approved posts along as examples. Read the rules in Settings > Learning and edit any that look wrong.
Steps
- 1
Review the first batch honestly
Setup writes 30 drafts. Approve only the ones you would post. The first week teaches it the most.
- 2
Reject for one reason at a time
Too salesy, too long, wrong topic, a word you hate. A repeatable reason makes a clear signal.
- 3
Press Learn now
After a batch, open Settings > Learning and press Learn now. It reads approvals, rejections and post numbers and rewrites the rules.
- 4
Read the rules in plain language
Edit anything wrong. Rules are written as sentences you can read and change.
- 5
Keep a weekly review
Auto-approved posts are left out of learning, so reviewing the rest keeps the signal clean.
What the loop does
Every draft ends in one of two places. Approve moves it to the Queue. Reject moves it to Rejected. Those two piles are the whole training set, along with your post numbers when a scheduler reports them.
Threadgehog uses them in two ways. When it writes new ideas and the daily batch, up to 8 of your most recent approved posts go along as examples. And with Threadgehog AI it compares what you approved with what you rejected and rewrites the rules for the next batch: every 3 days on Solo, every day on Pro, up to 3 times a day on Unlimited. Learn now does it on demand. The general version is on the page about AI that learns from approved posts. This page is the practice.
Think of it as a small feedback cycle you run every week. You judge a batch, Threadgehog rewrites its rules, the next batch follows them, and you judge again. The cycle only helps if your judgments are steady, which is why the rest of this page is about how to judge, not how to configure.
Nothing in the loop needs a spreadsheet. The Rejected tab keeps the drafts you turned down, so you can scan it when you want to see what pattern you are teaching.
A four-week example
Week 1: you sell handmade candles and the first 30 drafts arrive. You approve 11. You reject 19, and when you look at the pile, 12 of the rejects share one problem: they open with a question. You press Learn now.
Week 2: the rules now say something like 'avoid opening with a question'. The next batch has fewer question openers. You approve 17 of 30. The rejects now cluster around posts that mention the shop in the first line.
Week 3: you edit one rule by hand to say the shop name only appears in posts that carry the link. Approval climbs to 21 of 30.
Week 4: your own posts have started drawing replies. You paste your three best into Example posts and press Rebuild the brand voice. These numbers are an illustration of the pattern, not a result we measured.
Reject for reasons you would repeat
The loop follows your choices. If you reject a draft for being too salesy on Monday and approve a salesy one on Tuesday, the rules blur. Pick a standard and hold it for a month.
A quick way to stay consistent: keep a sticky note with the three things you reject most often. When a draft hits one, reject it. When a draft hits none, approve it, even if it is not your favorite.
When to edit instead of reject
Press E to fix a draft you like with one flaw. Editing does not tell Threadgehog the draft was bad, so use it for a detail like a number or a name, and use Reject when the whole angle is wrong.
Space sets a draft aside to come back to later, and Command-Z undoes your last choice if your thumb slips. More shortcuts are in reviewing the queue in five minutes.
Where auto-approve fits
Auto-approve waits until Threadgehog agrees with your choices often enough. Project settings shows how close it is, from Just started to Ready for auto-approve. After that, drafts it is sure you would approve skip Pending and go to the Queue.
Auto-approved posts are left out of the learning. If you let auto-approve run for a month and never look at Pending, the rules stop improving. Keep one weekly sitting for the rest, as in posting daily without writing every post.
What to look at: the approval rate
The simplest number to watch is how many drafts you approve out of each batch. In the example above it went from 11 of 30 to 21 of 30 in three weeks. Yours will differ. What matters is the direction. If the rate rises after Learn now, the rules are working. If it stays flat, your reasons for rejecting are probably inconsistent or your example posts do not match the voice you want.
Write the count on a sticky note after each batch. It takes ten seconds and tells you more than any setting. If the rate drops, look at what changed: new example posts, a new topic, or a different mood on your part.
Settings > Learning also has a What's working view that compares proven drafts and new ones against your typical post. With Buffer, Publer or Postiz connected, and Metricool for X, post numbers feed it.
Three fixes when the loop stalls
First, make the reason concrete. Replace 'felt off' with 'too salesy', 'too long' or 'wrong topic', and only reject for those. Second, read the rules in Settings > Learning and delete any that contradict each other. Rules are plain sentences you can edit.
Third, refresh the examples. Borrowed posts get stale. When your own posts draw replies, paste them and press Rebuild the brand voice. That resets the style base. The loop then has a better starting point and your approvals have more to build on.
Resist changing all three at once. Do one fix, write a batch of 10 to 15, and look at the approval count before touching the next.
Questions
How many decisions does it need?
Threadgehog does not publish a number. Auto-approve becomes available once it agrees with your choices often enough, and Project settings shows how close it is.
Where do I see what it learned?
Settings > Learning shows the rules it wrote, in plain language, and a What's working view that compares proven and new drafts with your typical post.
Does learning cost drafts?
No. Only writing a new draft counts toward your monthly drafts. Editing, rewriting for X and learning from your approvals are free.
Can I reset the rules?
You can edit any rule by hand, and Rebuild the brand voice rebuilds from the example posts you paste.
Does learning use my results?
With Buffer, Publer, Postiz or Metricool for X, Threadgehog reads post numbers and builds most new drafts on what your top posts share.
Is it the same on every plan?
Every plan has every feature. How often the rules are rewritten automatically depends on the plan: every 3 days on Solo, every day on Pro, up to 3 times a day on Unlimited, with Threadgehog AI.
Can I try it?
The Mac app is on a waitlist; the 14-day trial starts when it opens. Join on the pricing page.
Sources
- Threadgehog home (checked 2026-10-17)
- Threadgehog FAQ (checked 2026-10-17)