explore@-mention for context
Notion Agent pulls from @-mentioned pages and connected apps. Point it at authoritative sources instead of pasting long text.
Example
“Summarize @Q3 Launch Plan for executives. Three bullets: shipped, in progress, blocked.”
Notion AI
AI assistant in Notion for agents, enterprise search, meeting notes, and workspace writing.
WritingProductivityAgents
How it works / How to use
AI assistant in Notion for agents, enterprise search, meeting notes, and workspace writing. Notion AI lives in your pages, databases, and sidebar—no tool switching. Use inline Edit with AI on selected text, AI blocks on pages, or Notion Agent for multi-step tasks with workspace context. Full AI features are included on Business and Enterprise; other plans get limited trial usage.
- Open Notion and navigate to the page or database you want to work in.
- For quick edits, highlight text and choose Edit with AI (summarize, translate, change tone, fix grammar).
- For broader tasks, open Notion Agent from the sidebar or bottom corner and @-mention pages or people for context.
- Review AI output against your source facts, then accept, edit, or ask a follow-up in the same thread.
Writing and summarization in your workspace
Highlight text for inline Edit with AI, or open Notion Agent and ask with @-mentions to pull context from related pages. State tone, length, and what must not be invented.
What to provide
- The page, notes, or bullets you are working from
- Audience and purpose (team update, executive summary, customer doc)
- Output shape: paragraph, bullets, table, or specific word count
Details that improve the result
- Notion Agent uses context from your workspace and connected apps when enabled
- Try “Help me write” on a blank page with details about the content type
- Ask to fix grammar, expand, shorten, or change tone on existing drafts
Example prompt
Highlight messy meeting bullets, choose Edit with AI → Improve writing, then ask Notion Agent: “Turn this into a 150-word team update. Keep dates and owner names exactly. Tone: direct.” @-mention the project page for extra context.
If the first output is not good
- “Rewrite the opening to sound more confident.”
- “Summarize key takeaways as five bullets.”
- “Make this suitable for an executive—shorter and less technical.”
Common mistakes
- Asking Agent to summarize without @-mentioning the authoritative page
- Leaving out tone and length so the first draft needs heavy trimming
- Treating AI summaries as verified facts without checking sources
AI blocks on pages
Type /AI Block on a page, choose Summary, Key points, or Custom, and write a prompt. AI blocks generate content from the current page context.
What to provide
- The page context AI should use (current page or @-mentioned pages)
- The custom output you want (summary, key points, translation, draft section)
- Format rules: bullets, table, word limit
Details that improve the result
- /AI Block supports custom prompts, page summaries, and key points
- Useful for standing summaries that update when you regenerate
- Combine with inline Edit with AI to refine generated blocks
Example prompt
On a long spec page, type /AI Block → Custom: “List open decisions and owners in a two-column table. Only include items explicitly stated on this page.”
If the first output is not good
- Regenerate the block after you edit the source sections
- Change the prompt to “three bullets for non-technical readers”
- Move block output into the doc and refine inline
Common mistakes
- Expecting the block to read pages you did not @-mention or include
- Using Summary when you needed extraction (use Custom with explicit columns)
- Forgetting to regenerate after major page edits
Database AI autofill properties
Add an AI autofill property to your database, pick AI summary, AI key info, or AI custom autofill, write the prompt, and let Notion generate text per row.
What to provide
- A database with page content or properties to draw from
- The property type: AI summary, AI key info, or AI custom autofill
- A specific prompt with desired length and format
Details that improve the result
- AI summary auto-summarizes each page; custom autofill follows your prompt
- Be specific about length (“one short sentence”, “three words”, bullet format)
- Properties update when you regenerate or when page content changes
Example prompt
On a research database, add AI custom autofill: “Summarize this page in one sentence for a weekly standup. Name the decision status only if stated.”
If the first output is not good
- Switch from AI summary to custom autofill for stricter format control
- Adjust prompt to “simple language for non-technical teammates”
- Sort or filter rows after autofill to spot empty or vague entries
Common mistakes
- Vague autofill prompts that produce inconsistent column lengths
- Expecting autofill to infer data not present on the page
- Not reviewing generated properties before sharing the database
AI Meeting Notes
Start AI Meeting Notes in Notion. It transcribes the conversation, summarizes key points, and keeps notes searchable with Notion AI.
What to provide
- A meeting to transcribe (any client, per Notion help)
- The workspace page where notes should live
- What you need back: summary, action items, decisions
Details that improve the result
- Meeting notes work with any client and stay in your workspace
- Summaries and action items are generated from the transcript
- Search meeting content later with Notion AI
Example prompt
Run AI Meeting Notes on a sprint planning call. After the meeting, ask Notion Agent: “List action items with owners from today’s meeting notes page.”
If the first output is not good
- Edit the auto-summary to fix names or dates
- Ask Agent to “turn action items into a task database”
- Highlight a transcript section and summarize for stakeholders
Common mistakes
- Assuming perfect speaker attribution without review
- Skipping edit pass on action items before assigning
- Not linking meeting notes to the relevant project page
Research Mode for detailed reports
Open Notion Agent, enable Research Mode or ask for a detailed report, bound the scope, and request separation of verified facts from open questions.
What to provide
- The research question and scope
- Workspace pages or connected sources Agent should use
- Deliverable format: report sections, citations style, length
Details that improve the result
- Research Mode generates detailed reports and summaries
- Enterprise Search and Connectors extend context to Slack and Google Drive on eligible plans
- Ask for “what still needs verification” in the output
Example prompt
Notion Agent, Research Mode: “Compare our Q3 roadmap page @Roadmap with @Competitor notes. Output: table of gaps, risks, and open questions. Do not invent launch dates.”
If the first output is not good
- “Expand only the open questions column.”
- “Rewrite for a leadership audience, one page max.”
- “Add a section listing sources used from the workspace.”
Common mistakes
- Unbounded research questions with no output structure
- Treating generated reports as verified without checking @-mentioned pages
- Expecting live web data without confirming connector or web access settings
Custom Agents for team automation
On Business or Enterprise, build a Custom Agent with a trigger or schedule. One builder maintains it; the team benefits. Monitor credit usage for agent runs.
What to provide
- The recurring task (Slack Q&A, project updates, task routing)
- Trigger or schedule and inputs Agent should read
- Output destination and what requires human approval
Details that improve the result
- Custom Agents automate recurring work on schedules or triggers
- Personal Notion Agent is on-demand; team Custom Agents are a Business/Enterprise add-on
- Custom Agents consume Notion credits per run
Example prompt
Custom Agent: every Friday 4pm, summarize @Weekly Updates database rows tagged “shipping”, post bullets to #eng-updates in Slack. Do not post if fewer than two updates exist.
If the first output is not good
- Tighten the prompt after reviewing the first scheduled run
- Add a human-approval step before external posts
- Reduce scope to one database view instead of the whole workspace
Common mistakes
- Automating vague goals (“keep everyone updated”)
- Granting external actions without reviewing the first runs
- Building on a plan without Custom Agent access or credits
How to prompt
Notion AI works best with workspace context. @-mention pages, people, or databases; state audience, format, and length; and separate facts (from your pages) from instructions (what to do with them).
Context: @Project brief and @Meeting notes. Task: draft a one-page status update. Output: bullets for progress, blockers, next steps. Constraints: keep dates verbatim; do not invent metrics.
Use inline edits for local changes
Highlight → Edit with AI for grammar, tone, length, translation, or summary of a section without opening Agent.
Example
Highlight a paragraph → “Make shorter” or “Change tone → Professional”.
Define database autofill precisely
AI custom properties need explicit length and format. “One sentence” beats “summarize”.
Example
AI custom autofill: “One short sentence on decision status. Use TBD if unstated.”
Start blank pages with Help me write
On an empty page, ask Notion AI to Help me write with content type, audience, and details—cover letters, specs, job descriptions.
Example
Help me write a job description. Role: senior designer. Team: 8-person product. Include responsibilities and requirements; no salary numbers.
Iterate in Agent threads
Keep follow-ups in the same Notion Agent chat. Reference prior output and change one dimension at a time.
Example
“Keep the structure. Rewrite section 2 for a non-technical audience.”
Best output tips
Work inside your workspace
Notion AI lives on pages, in databases, and in the Agent sidebar—no export to an external chat tool for most tasks.
Choose inline vs Agent
Edit with AI on highlights for local fixes. Notion Agent for multi-step work, cross-page search, and tasks that create or edit pages.
@-mention authoritative pages
Agent uses @-mentions and connected apps (when enabled) as context. Point at the canonical doc instead of duplicating long paste.
Use /AI Block for reusable outputs
Type /AI Block for summaries, key points, or custom prompts tied to the current page. Regenerate after substantive edits.
Configure database AI properties
AI summary, AI key info, and AI custom autofill generate per-row text. Custom prompts should specify length, format, and empty-value rules.
Capture meetings with AI Meeting Notes
Transcription and summaries stay in Notion and remain searchable. Edit action items before assigning.
Use Research Mode for depth
Research Mode produces detailed reports. Bound scope, name output sections, and ask for open questions separately from stated facts.
Know your plan limits
Notion AI is included on Business and Enterprise. Free and Plus workspaces receive limited trial usage—check Notion AI settings.
Build focused Custom Agents
One outcome per agent: a schedule, clear inputs, defined output, and human review before irreversible external actions.
Track Custom Agent credits
Custom Agents consume Notion credits. Review usage after the first scheduled runs before expanding scope.
Separate facts from instructions
Paste or @-mention source material, then state the task. Say which facts must appear verbatim.
Iterate tone with plain language
“Make more professional”, “friendlier”, or “shorter” on a selection beats restarting from a blank prompt.
Create databases with Agent
Notion AI can create databases with views and properties, populate rows, and help write formulas—review schema before sharing.
Enterprise Search extends context
On eligible plans, Enterprise Search and Connectors let Agent find answers across Slack and Google Drive—not only Notion pages.
Verify before you share
AI can misread or smooth over missing data. Check dates, owners, and metrics against source pages before publishing.
- @-mention pages and databases so Notion AI uses the right context.
- Highlight text for quick Edit with AI; open Agent for multi-step tasks.
- State audience, length, and tone before asking for drafts.
- Use /AI Block for standing summaries and custom page outputs.
- Add AI autofill properties with explicit length and format rules.
- Review meeting note summaries and action items before sharing.
- Business and Enterprise include full Notion AI; other plans have limited trial usage.
- Custom Agents consume credits—start with one focused automation.
- Verify numbers, dates, and names against source pages after AI edits.
Try this AI
Product Details
Pricing, features, limits and latest updates
Product Details
Pricing, features, limits and latest updates
Notion AI
AI assistant in Notion for agents, enterprise search, meeting notes, and workspace writing.
Free / Paid · Free
Pricing Plans
Free
Free
Plus
$10/ Monthly
Business
$20/ Monthly
Enterprise
Custom
Custom Agents (per 1,000 credits)
$10/ Usage
Key Features
Chat
Notion Agent, inline AI, AI blocks, Enterprise Search, and Research Mode; full AI is on Business and Enterprise plans.
Automation
Custom Agents run on schedules or triggers and consume Notion credits.
Agents
Personal Notion Agent is on-demand; team Custom Agents are a Business and Enterprise add-on.
Limits
- Full Notion AI: Full Notion AI is included on Business and Enterprise; Free and Plus receive a limited complimentary trial allowance (Notion Help Center).
- Image generation (beta): 10 per 24h. Image generation beta allowance per user on eligible plans (Notion Help Center).
Ideas / Prompt experiences
Share a prompt that worked for you. Username and email are shown with your submission. External links are not allowed.
Example prompt
Summarize meeting notes in workspace
Prompt
Summarize this page into: 1) Decisions 2) Owners 3) Deadlines 4) Open questions Use only what appears on the page. Mark anything unclear as 'needs confirmation'.
Short explanation
Notion AI works best when the source material is on the page and the output structure is named up front.
Example prompt
Research Mode scoped question
Prompt
Research Mode: What changed in our competitor's pricing in the last 12 months? Use connected workspace sources first, then the web. Return citations and a 'verify manually' list.
Short explanation
Research Mode is for deeper multi-source questions; Notion advises double-checking accuracy on important outputs.
Share your experience
Required fields are marked. Variation, result, and explanation are optional.
