May brought a meaningful wave of updates, and the headline is a major expansion of the AI model picker in Notion. GPT-5.5 is now live, alongside a wider lineup of frontier models, and the picker is finally a real lever for matching the right model to the right task.
Layer in a round of refinements to dashboards, custom agents, AI Meeting Notes, and the connector ecosystem, and there’s plenty here worth setting up this week.
Here is exactly what’s new this month, what it means, and what to do first.
Core Concepts:
- GPT-5.5, Claude Opus 4.7, and DeepSeek V4 are all live in Notion AI’s model picker
- The model lineup expanded with new frontier options across reasoning, drafting, and cost-efficient task types
- Dashboards and Custom Agents picked up refinements since their March launch
- AI Meeting Notes custom instructions are clearly paying off for teams that use them
- Connector coverage has continued to expand, making cross-tool AI search more reliable
Who does this apply to: Marketers, agency teams, project managers, solopreneurs, and anyone running a Notion-centric workflow. Most updates are on Plus plans and above, with a few features reserved for Business and Enterprise.
GPT-5.5 Is Live in the Model Picker
OpenAI’s latest model just landed inside Notion AI. GPT-5.5 is the biggest leap in the GPT-5 family since launch — sharper reasoning, longer context handling, and more reliable structured output. For longer briefs, multi-doc summaries, and any task where you’ve been wrestling with hallucinations, it’s the new default worth setting.
To switch: open any AI panel, click the model picker, and select GPT-5.5.
The Model Lineup Just Got Wider
Beyond GPT-5.5, the model picker now gives you a real spread of frontier models. The point is specialization — pick the right model for the right job instead of forcing one model to do everything.
Three models stand out in the current lineup:
- GPT-5.5 (OpenAI) — the new default for long-form drafting, multi-doc summaries, and structured output. Sharpest all-around model in the picker, with the most reliable handling of long context.
- Claude Opus 4.7 (Anthropic) — the strongest pick for nuanced writing, voice-sensitive copy, and careful analysis. Opus 4.7 is noticeably better than GPT-5.5 at matching tone (especially brand voice in marketing copy) and at multi-step reasoning over messy source material. Use it for editorial work, client-facing copy, and any task where judgment matters more than speed.
- DeepSeek V4 — the cost-efficient workhorse. Strong on technical tasks, coding, structured extraction, and bulk operations where you’d rather spend tokens than time. Punches above its weight on reasoning given the price point. Use it for high-volume tasks like extracting action items from dozens of meetings, formatting data, or running first-pass research.
A few rules I’ve been running with:
- Editorial and brand-voice work (blog posts, client comms, email copy): Opus 4.7
- Strategy docs, multi-doc summaries, complex briefs: GPT-5.5
- High-volume, structured, or technical tasks: DeepSeek V4
- Quick summaries and one-liners: any of the faster models still hold their own
If you’ve been defaulting to one model across the board, this is the moment to start matching tasks to models. The quality lift is real.
Dashboards & Custom Agents: Two Months In
The two features that defined 3.4 have both picked up small but meaningful refinements since launch. Dashboards are smoother to build, with better defaults on KPI cards and chart layouts. The Notion Agent does a noticeably better job spinning one up from a natural-language prompt now. If you tried this in March and bounced off it, give it another swing.
Custom Agents have benefited from the broader model picker too. Assigning the right model to the right agent has a real impact on output quality. If you set up custom agents earlier this year and they’re not pulling their weight, the fix might just be switching their underlying model.
Currently available on Business and Enterprise plans. Learn more about dashboards from Notion’s Help Center.
AI Meeting Notes Custom Instructions Are Paying Off
The custom instructions feature for AI Meeting Notes that landed in 3.4 has had time to bake. The pattern is clear: teams that wrote tight, opinionated instructions (preferred section order, length caps, voice rules) are getting dramatically better summaries than teams running the default.
If you haven’t customized your instructions yet, this is a quick win. Two or three short rules can change the quality of every meeting note your team generates from now on. Go to the footer of any AI Meeting Notes block and select Instructions to get started.
Connector Coverage Keeps Growing
Notion AI’s cross-tool search has gotten quietly stronger. Salesforce, the connector that launched with 3.4, has been the standout for sales and client work. Cross-tool search inside Notion AI is now reliable enough to be a real first stop for “where did we talk about X” questions spanning Slack, Drive, Gmail, and your CRM.
For agency and client-work setups, this is where Notion is pulling ahead of single-tool search experiences. Set up under Settings → Notion AI → AI connectors.
The MCP guide on this site covers the framework behind how cross-tool integrations work under the hood.
Custom AI Skills: Library-Building Is Where the Value Lives
Custom AI Skills shipped with 3.4, and the teams getting the most out of them are the ones treating skills like a library, not one-off prompts. Summarization templates, brand-voice drafting prompts, action-item extractors, content rewrite frameworks — each saved as a reusable skill the whole team can trigger in one click.
If you’re still copy-pasting prompts into AI panels, spend an hour this week converting your top five into skills. Available on Business and Enterprise plans.
A Few Smaller Things Worth Knowing
- Performance improvements across Notion AI panels and long-document loading
- Presentation mode refinements — smoother transitions and better embed support inside slides
- Tabs block continues to gain layout flexibility
- Markdown Content API is seeing real adoption from third-party tools, which means cleaner integrations with the rest of your stack
The Bigger Picture
What’s interesting about Notion’s May cycle isn’t a single headline feature. It’s that the platform is getting deeper rather than wider. The biggest moves are around model choice, agent quality, and connector reliability. Notion is making it easier to do more of your real work inside a single workspace, with the right AI model assigned to each part of the job.
For marketers, the practical move this month is simple: open the model picker, get familiar with the lineup, and start matching tasks to models. That alone changes the quality of everything Notion AI produces for you.
The Claude Cowork walkthrough pairs well if you’re building a more delegated AI workflow across desktop and Notion.
About Jason Pollak
Jason Pollak is a marketing strategist with over 10 years of experience building campaigns for entertainment brands, artists, and businesses across music, film, television, eCommerce, and B2B SaaS. As Director of Marketing at Young Money Entertainment, he grew Lil Wayne’s Facebook following from 10 million to 50 million and managed over 60 million followers across the roster. He also served as Paid Media Director at Horizon Media, launching major TV shows for History Channel, A&E, WWE, and Lifetime, and led film marketing for Utopia Distribution, generating over $10 million in revenue on a $200K media spend. Jason specializes in paid media, organic social strategy, email automation, SEO, content development, and AI-driven marketing systems. He holds a BA in English Literature from Binghamton University and a Masters in Media Studies from Brooklyn College. Learn more at jasonpollakmarketing.com.