AI Agent
Learn on-brand from past work
Agents study your approved, team-rewritten and published work before writing, match its voice, and reuse your brand images. Off until you turn it on.
Turn it on in Project settings → Agent features → Learn on-brand from past work. From then on, the best examples of your own work guide what agents write.
What counts as a good example (strongest first)
- Approved: items that passed the approval chain.
- Rewritten by your team: what someone kept after heavily rewriting an agent’s draft (from Learn from feedback).
- Published: social posts that went out (most engaging first) and articles or landing pages you published.
Where it is used
- Agents (desktop, Reaudit assistant, Claude or ChatGPT over MCP) read the examples before writing and are told to match voice, structure and length without copying sentences or reusing facts that do not fit.
- Content Studio articles and Broadcast’s AI post text include the top three examples in their instructions.
- Your brand images (images marked as brand assets in the design library) are offered to agents to reuse before they generate new ones.
- See what agents learn from in the What agents learn from panel on the Agent features page, filtered by channel.
Examples only come from the same project, and only for people on that project. Nothing is read or added while the feature is off.