From title library to publishable articles

1. Import titles: bulk title imports are deduplicated automatically; empty and duplicate titles never enter the task queue.
2. Pick templates and prompts: each batch binds a content template and prompt combination; client, categories, model and pricing are frozen into a task snapshot at submission.
3. Generate in parallel: multi-site tasks run concurrently, with task status, failure points and retry actions visible end to end.
4. Review, then publish: generated content enters the review flow first; only confirmed content moves into the publish queue, so nothing goes live unchecked.

The platform supports Bring Your Own Key: use your own AI provider API keys and pay the model provider directly.

• Model choice: different tasks can use different models, letting you tune the cost/quality mix per task type.
• Clear cost ownership: AI token spend stays with your own provider account, while the platform bills on successful publication — two separate, independently verifiable ledgers.
• Key protection: API keys are encrypted at rest and never echoed in the interface; rotate them anytime. See BYOK & Security.

Batch generation does not mean batch publishing. The content state machine covers draft, pending review, published, unlisted, rejected and archived:

• Generated content enters the review step by default; only human-confirmed content reaches the publish queue.
• Both the title library and topic opportunities carry duplicate detection (unique / possible duplicate / duplicate), so repeated topics are caught at import time.
• Failed publish tasks keep their failure reason; after fixing, retries never double-publish or double-bill.

See Content Quality Control for the full picture.

• Success-only billing: only successfully published articles count against your plan allowance or overage.
• Transparent allowances: successful-article allowances are Starter 150/month, Growth 800/month, Agency 3,000/month. Full pricing is on the homepage pricing table.
• Controlled overage: before exceeding an allowance you get an explicit prompt; overage is off by default and never silently billed.

• What if generated quality is inconsistent? Iterate on template and prompt combinations and gate output with the review step; task snapshots preserve the exact configuration behind every run, so you can identify which combination performs best.
• Can I use different AI models? Yes — under BYOK, different tasks can be configured with models from different providers.
• Do failed generations cost anything? No. Only successfully published articles are billed.