AI Features
Jotaid uses AI to sharpen your thinking, not to replace it. Every feature below except the AI chat panel runs on your own API key, and none of your data is ever used for training.
INFO
AI is a Pro feature and needs a provider key. Start with AI Setup.
The exception is the AI chat panel at the end of this page: it is an embedded browser, needing neither Pro nor a key.
At a Glance
| Feature | Where to trigger it | What you get |
|---|---|---|
| Titles and tags | Automatic as you write (can be turned off) | A title plus a few tag suggestions |
| Smart Node | Left column + menu → Smart Node | Concepts worth making into nodes |
| Smart Theme | + menu in the Theme view's left column | Groups worth making into themes |
| Theme summary | Theme Summary on a theme note | Insights, open questions and HMWs |
| Enrich Content | Enrich Content on a node's detail view | A first draft of that concept's body |
| Smart Format | ⇧⌘F, or the editor's ⋯ menu | Structured text, wording untouched |
| Explain a relationship | Open a cell in the co-occurrence matrix | Why those two concepts relate |
| Semantic suggestions | The SUGGESTIONS section of a node | Related notes worded differently |
| AI chat panel | ⇧⌘B (free, uses no key) | Ask without leaving your notes |
Automatic Titles and Tags
As you write, the AI reads your content and proposes a descriptive title, plus relevant tags drawn from the vocabulary you already use.
You stay in charge — nothing is imposed, and every suggestion can be edited or ignored. On by default; toggle it in Settings → AI.
Smart Node
When you cannot see which concepts deserve to exist, have the AI read through: Smart Node scans the current project's notes, proposes concepts worth making into nodes, and lists the notes citing each one.
It lives in the left column's + menu. Nothing is ticked by default — only what you tick and apply gets created. The full scan flow, the extra-requirements field and re-scanning are covered in Creating Nodes.
The dot on the +
Jotaid watches for candidate concepts as you write. When enough have accumulated, a small dot appears on the + — that means "there are new suggestions worth a look", not a count of chores.
Smart Theme
Have the AI read the project and propose which themes should exist — grouping notes that keep turning up together, naming each group, describing it in a line and picking a card colour.
It lives in the + menu of the Theme view's left column (on iOS, the + on the THEMES sub-page), mirroring Smart Node in the Node view. It looks at the current project's top-level cards, so it is unavailable inside a theme or while a theme filter is on.
Again nothing is ticked by default — only ticking and applying creates themes and moves cards on the canvas. That is deliberate: ticking by default would let the AI rearrange your canvas for you, and themes should be your own thinking. Unhappy with the result? Add a line of extra requirements and re-scan.
Not the same as theme summary
Smart Theme answers "which themes should I have", facing notes that are not grouped yet. Theme summary answers "what does this theme say", facing a theme that already exists.
Theme Summary
The Theme Summary button on a theme note: the AI reads every note in that theme and writes the summary back into the theme note's body. Good for closing off a line of work.
It returns four parts: what the theme is about, synthesised key findings, 2–4 insights, and 2–3 How Might We questions. Insights state what is happening and why it matters, never solutions — see Design Thinking.
The same button sits on a project note, where it summarises that project's themes instead.
Enrich Content
Enrich Content on a node's detail view: the AI reads the notes citing this node and expands its body into a draft you can build on, rather than a blank page. Useful when a concept exists but you cannot yet write its definition.
Smart Format
⇧⌘F (macOS) or the editor's ⋯ menu (top right of the note on iOS): tidies the current note's Markdown structure in one step — consistent heading levels, tables rescued from code fences, bullets for parallel items — while changing not one word.
Especially worth it on something you typed in a hurry. See Smart Format.
Explain a Relationship
Open a bright cell in the co-occurrence matrix and have the AI say why those two concepts are strongly related — turning a value into a sentence you will remember.
The graph and matrix panels can also read the whole network, a single concept or a single pair; see AI Interpretation.
Semantic Suggestions
The SUGGESTIONS section of a node's detail view: having read your notes, the AI proposes which notes — worded differently — belong linked to this node.
It deliberately skips anything unlinked mentions already catches: that layer is literal, accurate and free, so there is no point paying a model to repeat it. What appears here is what literal matching cannot reach.
AI Chat Panel (Embedded Browser)
⇧⌘B opens a panel beside the editor holding the web versions of Claude, ChatGPT, Gemini and DeepSeek as tabs (plus Kimi in some regions), signed in exactly as they are in your browser. It exists so you can ask something without leaving your notes.
This one is a different animal from everything above:
- It loads those services' websites, using your own logins there — it does not consume your API key
- So it needs no Pro and works on the free tier
- Read an answer worth keeping? Select it and hit Save to note in the toolbar to store it with its source, section and context — see Quick Capture
- Panel details (tabs, resizing) are in macOS Interface
Which Model to Use
Different features ask very different things of a model: titles and tags are fine on something fast and cheap, while theme summary and semantic suggestions benefit noticeably from a stronger one. You can pick a model per feature — see AI Setup.
A truncated or empty result usually means the model is too small for the task — move to a larger one, or a provider with a higher output limit.
Privacy
- Your API key is stored locally in the system Keychain
- Note content is sent to the provider you configured, and only when you trigger a feature
- Nothing is collected, and nothing is used for training
- All of the above covers Jotaid's own AI features; whatever you type into the AI chat panel is governed by the terms of the site you are visiting
Next Steps
- AI Setup — Providers, keys and switches
- Unlock Pro — AI is the second half of a two-step setup
- Semantic Relations — The layer behind semantic suggestions