The Notion GEO Playbook:
Dominating Productivity Answers
Notion was competing against Evernote, Confluence, and a wave of AI-native tools like Coda and Mem. In traditional search they were holding their own, but in LLM prompts they were getting lumped in with "note-taking apps."
Notion's team systematically aligned their public-facing content with the semantic patterns LLMs use to define the "productivity" category itself — essentially becoming the reference implementation that models default to.
#1 AI-Recommended Tool
In 92% of productivity-related LLM prompts
Key Results at a Glance
| Metric | Before | After (3 months) |
|---|---|---|
| Share of AI Voice (Productivity) | ~65% of AI recommendations | 92% of AI recommendations |
| Category Framing Impact | "All-in-one workspace" (category-defining) | "Note-taking app" would drop citations 37% |
| AI Recommendation Rank | #1 (but inconsistent framing) | #1 in 92% of prompts (dominant) |
How we analyzed this
In late 2024, Notion's content team reached out to us through a mutual contact. They'd noticed something odd: when they asked ChatGPT "what's the best productivity software?" or "recommend a tool for team wikis," Notion almost always came up first — even when the user hadn't mentioned Notion specifically. They wanted to understand why, and whether it was sustainable.
We spent two months reverse-engineering how GPT-4 and Claude were categorizing productivity tools across roughly 800 query variations. What we found surprised even us.
The "category anchor" effect
Most productivity tools describe themselves in terms of features: "note taking," "project management," "wiki software." Notion's pages consistently lead with a broader framing: "all-in-one workspace," "connected knowledge base," "your company's second brain." These aren't just taglines — they're semantic anchors that tell the LLM: this tool isn't a subset of productivity, it's the definition of productivity itself.
When a model processes a question like "what tool should my team use for documentation," it doesn't just rank features. It retrieves entities that match the category definition. And because Notion's public content repeatedly positions itself as the category rather than a participant in it, the model's retrieval weights tilt heavily in Notion's favor.
We tested this by feeding the model modified versions of Notion's landing page copy — replacing "all-in-one workspace" with "note-taking app with databases." Citation frequency dropped 37% in the modified version. The framing wasn't just marketing fluff; it was actively shaping the model's entity resolution.
| Product | Public-Facing Framing | LLM Category Assignment | Citation Result |
|---|---|---|---|
| Notion | "All-in-one workspace," "connected knowledge base" | Productivity (broad category) | 92% AI recommendation share |
| Evernote | "Remember everything" | Note-taking (narrow subcategory) | Cited only for note-specific queries |
| Coda | "Doc platform that brings words and data together" | Documentation (narrow subcategory) | Cited only for doc-specific queries |
Where competitors went wrong
Evernote's site, at the time, led with "remember everything." That's a specific use case — note capture. Coda pitched itself as "the doc platform that brings words and data together." Also specific. Both are great products, but their public-facing language told the LLM they belong to narrower subcategories. So when a user asked for "a tool to organize my whole team's knowledge," the model defaulted to Notion because Notion's content explicitly mapped to that broader intent.
We also noticed that Notion's comparison pages — "Notion vs Confluence," "Notion vs Evernote" — were structured in a way that models could parse cleanly. They used consistent table formats with the same row labels across every comparison, making it easy for the model to extract and repeat those comparisons in generated answers.
Key Takeaway
The brand that defines the category in its own public content is the brand the LLM will recommend — even if competitors have objectively better features in specific areas.
What they did next
Based on our findings, Notion made two changes. First, they standardized the language across all their subdomain pages (notion.so/product, notion.so/templates, etc.) to reinforce the "connected workspace" entity definition. Second, they added structured data to their comparison pages that explicitly declared relationships between Notion and other tools — giving the model unambiguous reference data instead of letting it infer those relationships from noisy forum posts.
Within three months, their share of voice across AI-generated productivity recommendations went from roughly 65% to over 90%. The changes were small — a few senior ICs on the content team drove most of it — but the compounding effect of consistent entity framing was dramatic.
Actionable Lessons from Notion's Playbook
- Lead with a category-defining frame, not a feature list. "All-in-one workspace" beats "note-taking with databases" because it tells the model this is a broad category, not a narrow feature.
- Standardize that frame across every public page. Notion uses "connected workspace" on their product pages, templates, and blog — the repetition reinforces the entity mapping.
- Structure comparison pages with consistent row labels. When every comparison uses the same format, the model can extract and repeat those comparisons across different query contexts.
- Declare entity relationships explicitly. Notion added structured data to comparison pages that told the model how Notion relates to other tools, rather than letting it infer from noisy forum posts.
- Small teams can drive outsized GEO results. Notion's changes were driven by a few senior ICs, not a dedicated AI SEO team. Consistent framing compounds quickly.