Train the bot
Practice conversation
What just happened
How to train it well
Drafts written for real fans
Your bot wrote these for actual conversations. Nothing has been sent. Send it, fix it, or bin it — every fix is the best training data there is.
Import your old conversations
The fastest way to make it sound like you. Paste real exchanges and it picks up from them straight away.
Where to get them
Open any conversation on the platform you use, copy a stretch of it, and paste it here with Fan: and Me: at the start of each line.
Fifty exchanges is plenty to start. It only needs enough to hear how you talk.
Your accounts
Connect once and everything runs from here — one post to all of them, every DM in one inbox. You can disconnect any of them at any time.
Connected
Available to connect
Conversations
Thread
Pick a conversation.
Simulate an inbound message
This drives the real reply pipeline — every safety gate, the spend guard, disclosure, and escalation. The trace below shows what each step decided.
Shadow Mode queue — drafts awaiting you
Every edit here becomes a (draft, your version) preference pair. This is the training pipeline, and it costs you nothing because you were answering these anyway.
Voice dials
Two dials are not free. Explicitness is capped by the destination platform's policy. Sales pressure is capped in real time by the spend guard — set it to 100 and it will still be forced to 0 for a fan in distress.
Autonomy
Boundaries
Yours
Platform — cannot be removed
Training
Platform reference
What each platform allows, and what we deliberately don't support.
Version history
Personas get worse. This is the undo.
Shot list — what to shoot, ranked by what fans will pay for
Everything else in ThotBot reduces your costs. This raises your revenue: content is pre-sold before it's produced, so nothing is shot speculatively.
| Shoot this | Fans | Paid | Quoted | Potential |
|---|
Exposure dial → weekly plan
Pre-shoot outreach
This month
Ledger — every charge, with its evidence
Open any fee and see which messages earned it. Disputes are resolved by data, not argument.
| Gross | Rate | Our fee | You keep | How it was calculated | Evidence |
|---|
Open handoffs — the bot paused and wants you
Safety events
| Category | Action | Severity | Detail | When |
|---|
Complaints — 7-day card-network SLA
Where one post can go
Rate a post once and the composer routes it. Anything a platform doesn't permit is refused here, before dispatch — a rejection at their end can cost you the account.
Mainstream platforms
Ordinary social media management. These prohibit explicit content and the CRM enforces that for you.
| Platform | Can do | Max rating | Autonomy ceiling | Ban risk to you |
|---|
Adult-permitting platforms
These allow rated content, usually with labeling. Ban risk is risk to your account — that matters more than risk to ours.
| Platform | Can do | Max rating | Autonomy ceiling | Ban risk to you |
|---|
Deliberately not supported
Promising these would be a promise we can't keep, to a customer who'd notice in week one.