Autopilot
Autopilot runs the content loop for a blog on a schedule: it finds topics, writes them, checks them, and — if you let it — publishes them. It's configured per blog under Autopilot, with a six-step setup wizard that can activate several blogs in one pass.
Autopilot spends credits like any other AI generation. Everything below exists to make sure it spends them on posts you'd have wanted anyway.
The three levels
Autopilot is one setting with three positions, and you should climb them in order.
1. Topic discovery. Keeps a buffer of researched, ranked topic ideas topped up. Nothing is written. You approve or reject each one, and every suggestion comes with the research that justified it, so you can see why it was proposed. This is the level to start on: it tells you whether Bloggable understands your niche before it spends anything on drafting.
2. Auto-drafting. Writes full drafts from approved topics and leaves them for you. Nothing goes public without you opening it.
3. Full autopilot. Sources, writes, checks, schedules and publishes unattended. It runs hourly.
Off is the fourth position, and pausing the blog also stops autopilot.
Setting the volume
Rather than picking an abstract "level of aggression", you configure output: five planned, three drafted, one published a week (or per month). The pipeline sizes itself to hit that. Cadence — daily, three times a week, weekly, fortnightly, monthly — sets the rhythm.
Before you switch anything on, the wizard shows a live credit forecast for the volume you've asked for. You'll know the monthly cost before it starts, not after.
The guardrails
The publish gate. Nothing publishes unless it passes every threshold you set: minimum word count, minimum AEO score, and the AI-integrity check. These are objective measurements, not the model's opinion of its own work. A draft that fails is held back and flagged for you, not quietly published anyway.
A hard budget cap. Set a monthly credit ceiling per blog and autopilot stops at it. This is separate from the forecast — the forecast is an estimate, the cap is enforced.
Approval. At discovery level, every topic waits for your yes or no. Nothing gets written from a topic you rejected.
It never repeats itself. A topic that has already been published, drafted or queued cannot be planned again — on any path. That covers autopilot, "ask AI for ideas", and Joe. You will not wake up to three articles about the same thing.
Where topics come from
You choose which sources feed it:
- Your own site. Indexed pages from your website and helpdesk steer what gets planned — not just where posts link.
- Competitors. The blogs you listed in the blog's settings. Counter-angles found by the weekly competitor scan become next week's planned posts, interleaved so every monitored competitor gets a hearing rather than the loudest one dominating.
- Keyword seeds. Terms you supply directly.
- Cluster mode organises output around pillar topics instead of scattered one-offs.
You can also give a briefing — free text that constrains everything it plans, e.g. "no listicles, and never write about pricing".
Optional stages
Each adds credits per post, so each is a toggle:
- Featured image — picks and attaches an image automatically.
- Humanise — rewrites the draft to strip AI patterns and apply your voice.
- AI-pattern check — scores how machine-written the draft reads, and flags anything below your threshold for human review.
Publishing settings
Set the publish time and timezone, default tags, the target funnel mix (top / middle / bottom of funnel) and the default byline. The byline can be a team member or a persona; if you don't set one, a persona is assigned automatically so posts aren't published anonymously.
Running on demand
Run now triggers a manual pass. It's rate-limited to roughly once an hour per blog, which keeps a stuck loop from becoming an expensive one.
Watching it
The activity log records every run: what ran, what it produced, what it cost, and why anything was held back. If autopilot fails repeatedly, we email you.
A note on the per-event notification toggles. The wizard offers switches for "planned topic added", "draft ready" and "published". These are not yet wired up — failure emails do send, but those three per-event notifications currently do nothing. Don't rely on them.
A sensible rollout
- Run at discovery for a couple of weeks. Approve and reject honestly.
- Move to auto-drafting. Read the drafts. Adjust the briefing and voice based on what's wrong with them.
- Only then go to full autopilot, with the quality gate set where you'd actually be comfortable, and a budget cap you'd be happy to pay.