There is an assumption in AI product design that the right output eliminates the need for editing. If the AI generates good enough content, the user just approves it and moves on. We have found the opposite to be true. Better AI output increases the user's desire to edit, because it gives them a starting point worth refining rather than replacing.
This is why Cleo has a full rich text editor at its core - not a text area, not a markdown preview, but a proper document editing environment with formatting, hierarchy, media embedding, and collaborative features.
The editing hypothesis
When AI generates a blog post draft, the user almost never publishes it unchanged. They adjust the opening. They rearrange sections. They add a specific example the AI could not have known. They soften a claim that is too bold. They strengthen one that is too cautious. This editing process is where AI-generated content becomes the user's content.
If the editing environment is limited - if users have to copy the output to another tool to format it properly - the magic of seamless AI-to-publication is broken. The workflow that should take minutes stretches into the same multi-tool juggle the AI was supposed to eliminate.
Structure, not just text
Marketing content is not flat text. A blog post has headings, subheadings, pull quotes, embedded images, and callout boxes. An email has sections, buttons, dividers, and responsive layout concerns. A social post has character limits, hashtag placement, and media attachments.
The editor understands these structures natively. When the AI generates content, it produces structured output that the editor renders correctly - headings are headings, not bold text pretending to be headings. When the user edits, they work with semantic structure rather than fighting with formatting.
Version awareness
Every significant edit creates a version checkpoint. The user can see how the content evolved from the AI's initial draft through their edits to the final published piece. This version history is not just an undo mechanism - it is a record of the collaboration between human and AI that produced the final output.
Over time, these editing patterns become data. The system learns which types of edits the user consistently makes - which phrasings they prefer, which structures they favour, which claims they tone down. This feeds back into the generation process, producing drafts that require fewer edits over time.
The integration depth
The editor is not a separate tool that happens to be in the same application. It is deeply integrated with the AI conversation. A user can ask the AI to revise a specific section, and the edit happens in-place in the editor. They can highlight text and ask for alternatives. They can request structural changes that the editor applies without losing the user's manual edits.
This integration depth is what makes the AI-plus-editor combination more than the sum of its parts. The AI generates. The human refines. The tool supports both seamlessly.
- Cleo's Team