What this is: the two different ways you create and edit content, why they feel different, and how to choose. If you only read one conceptual page, read this one.
Both are powered by the same AI, but they're built for different jobs.
| Chat | The pipeline | |
|---|---|---|
| Feels like | Talking to a team of specialists | Filling out and approving a structured document, stage by stage |
| Pace | Fast, exploratory, turn-by-turn | Deliberate, sequenced, approval-gated |
| Control | You steer with each message | The stages themselves enforce structure |
| Best for | Setup, research, planning, quick edits, one-offs | Producing one polished piece with quality checks at each step |
| Where | Chats in the sidebar | The content editor (opens when you work on a specific piece) |
A new chat opens on What would you like to create? with five starter chips, each of which just fills the message box so you can edit before sending:
You never have to use them. Typing your own request works exactly the same way.
When you type a message, the platform reads your intent and hands it to the right specialist(s): a general assistant, a writer, an SEO specialist, or a researcher. They can work together, use tools (web search, keyword data, your brand context), propose something, and hand it back. You react, they revise. It's a loop.
Chat can do things the pipeline can't:
Chat's "gates" are soft: an agent that has done a chunk of work will stop and check with you before carrying on: a checkpoint card showing what it plans to do next, with a few options to pick from (or you can type a refinement instead). The save offer is the same idea at the end of a piece: save it to your library, keep iterating, or discard.
Once you answer, the checkpoint stays in the transcript. It shows the question that was asked, the options that were offered with your pick highlighted, and your answer. Scroll back at any time (or reload the thread days later) and the full back-and-forth is still there, so you can always see what the agent proposed and what you decided.
On an article draft, that card also offers Polish for engagement. It's optional and it never runs on its own. Click it and the writer makes one more pass over the prose only: stronger hooks, a more conversational voice, benefit-led framing, and generic phrasing cut. It won't add sections, change your headings, or touch your citations, and the piece never gets longer. The result lands as a new version of the same draft tab, with the save offer shown again, so you can compare, keep iterating, or save.
The pipeline takes one piece of content through fixed stages:
Brief → Outline → (Research) → Written → Extras → Published
Each stage is a hard gate. You review it, edit it, and click Approve to move on. The earlier stages constrain the later ones: the approved brief shapes the outline, the approved outline shapes the written draft. That's what "grounded" means: the draft can't wander off, because it's built on decisions you already locked in. Before a piece can publish, the pipeline also checks things like internal links are in place.
Why use it: when you want a single article (or whitepaper, report, etc.) to be right (vetted for angle, structure, SEO, and brand at each step) rather than generated in one shot. You always know exactly what's been approved.
Two pipeline-only concepts:
The two connect in one direction. When chat produces a draft and you hit Save to library on its save offer, the platform creates a content project (a pipeline piece) populated at the furthest stage chat completed (usually a full written draft). You then open it in the editor to review, approve stage-by-stage, and publish.
Saving is instant. The assistant stores the structured draft as it writes, so Save to library writes it straight to your library. It doesn't hand the request back to the assistant, so it can't stall, reword your draft, or drop sections on the way in. What you reviewed in the preview pane is what gets saved. The button shows Saving… for the moment it takes, then the piece appears in your library and the assistant drops a link to it in the chat, so you can open it straight from the thread whenever you come back to it.
Two things follow from that:
Drafts created before this change still save through the assistant, which takes a few seconds longer. You don't have to do anything differently; the button behaves the same way.
Research crosses the bridge too. Save a research run and you get a project with a brief written from the findings, plus the findings themselves on its Research tab, so exploratory work in chat becomes the grounding for a pipeline piece instead of being stranded in the thread. Approving the Outline later won't overwrite it. → Research & deep research
So a very common flow is:
Explore and draft in chat → Save to library → refine and ship in the pipeline.
Start a 14-day free trial, or book a walkthrough.