An agent that proposes the highest-leverage moves and drafts the work

No. It is supervised by design: it reads your data, proposes ranked actions and drafts the work, and nothing is published to your site or sent to a third party without your approval. That is a product decision rather than a limitation we are working around. In practice you get a queue of proposals with the drafting already done, so approving a fix is a review rather than a project.

Prompts worth adding to tracking, citation targets with the angle for earning them, and content: articles, comparisons and FAQ pages, shipped with valid schema.org markup ready to paste or publish. Fixes arrive as concrete changes rather than advice, and each carries the reason it was proposed and the gap it addresses, so you can weigh it against your own priorities before approving anything.

You do. As between you and us, you own the content the agent drafts for you and you may use it commercially. Two honest caveats: text generated for another customer working on the same topic may be similar, so we make no claim of uniqueness, and we do not warrant that generated content is free of third-party rights. Review before publishing, exactly as you would with any drafted copy.

Not in identifiable form. Our internal models are trained on aggregated, de-identified data only: de-identification happens before anything reaches a training process, we do not attempt to reverse it, and no model output surfaces one customer's content or identity to another. Starter and Professional subscriptions contribute on that basis by default, and Enterprise customers may opt out in their order form. Section 5.1 of our Master Subscription Agreement sets it out in full.

It works from your measured data rather than from general knowledge: your tracked prompts, the answers each engine actually returned, your citation gaps and your competitors' positions. That is the difference between advice about content marketing and a proposal to fix the specific comparison question where Perplexity currently recommends a named rival. It also closes the loop, because the same daily tracking that produced the recommendation shows whether it worked.

No, and small teams often get the most from it, because drafting is exactly the capacity they lack. A realistic rhythm is reviewing the proposals once a week, approving one or two, and letting daily tracking tell you whether they moved anything. The agent re-prioritises as your data changes, so the queue reflects your current gaps rather than a plan written at onboarding.