You want to create a Notion database but are unsure which property types to use or how to structure the columns. Notion AI can analyze a plain English description of your data and generate a complete database schema with property names and types. This article explains how to use Notion AI to turn a text description into a functional database table, including the prompt format, property mapping, and manual adjustments needed after generation.
Key Takeaways: Generating a Database Schema with Notion AI
- AI > Generate with AI > Database from description: The exact menu path to start schema generation from a text prompt.
- Property type mapping: Notion AI maps text descriptions to Select, Multi-select, Date, Number, Email, Phone, URL, and Relation property types automatically.
- Manual property review: After AI creates the database, you must rename properties, adjust types, and add formulas or rollups that AI cannot generate.
What Notion AI Can and Cannot Do for Database Schema
Notion AI uses a large language model to interpret your description and generate a new database page with pre-filled properties. The feature is available on Notion AI add-on plans (Plus, Business, Enterprise with AI). You must have AI credits available in your workspace.
The AI can create property types including Title, Text, Select, Multi-select, Number, Date, Email, Phone, URL, Checkbox, and Relation. It can also generate a few sample rows of data to illustrate the schema. However, Notion AI cannot generate Formula properties, Rollup properties, or database views (Gallery, Board, Timeline). It also cannot create linked databases or set up database templates.
The quality of the output depends heavily on the specificity of your prompt. A vague description like “track tasks” produces a generic schema. A detailed description like “track software development tasks with priority, assignee, sprint, estimated hours, and status” produces a more accurate result.
Steps to Generate a Database Schema From a Description
Follow these steps to create a new database using Notion AI. You need edit permissions on the page where you want the database to appear.
- Open the target page in Notion
Navigate to the page where you want the new database to reside. This can be a top-level page or a subpage inside an existing workspace. - Type a forward slash and select AI
Press the / key on your keyboard to open the command menu. Type “AI” and select Generate with AI from the list. Alternatively, press Ctrl + J (Windows) or Cmd + J (Mac) to open the AI quick prompt. - Choose Database from description
In the AI submenu, click Database from description. A text input field appears at the top of the page. - Write a detailed description of your data
Type a description that covers the purpose of the database, the columns you need, and the type of data each column should hold. Example: “Create a customer relationship management database with columns for company name, contact person, email, phone number, industry, deal value in USD, deal stage (Prospecting, Negotiation, Closed Won, Closed Lost), last contact date, and notes.” - Submit the prompt and wait for the schema
Press Enter or click the Submit arrow. Notion AI processes the request and inserts a new database below your cursor. The database includes property columns and two to three sample rows. - Review and rename properties
Click on each property header to rename it if the AI used a generic name. For example, change “Select” to “Deal Stage” and adjust the options in the Select property. - Adjust property types if needed
Some property types may be incorrect. Click a property header, select Edit property, and change the type from the dropdown. For instance, change a Text property to Phone or Email if the AI misidentified the data. - Add missing properties manually
Click the + button at the end of the property row to add any columns the AI missed. Choose the correct property type and name it. - Delete sample rows
Hover over the sample rows generated by AI, click the three-dot menu, and select Delete to remove them. The rows are only for preview and should not remain in a production database. - Rename the database
Click the database title at the top and type a meaningful name such as “CRM Database” or “Project Task Tracker”.
Common Issues After AI Generates a Database Schema
AI Created Too Many or Too Few Properties
If the output has more columns than needed, delete the extra properties. Click the property header, select Delete property, and confirm. If the output has too few columns, add missing properties manually using the + button. The AI cannot guess context beyond your prompt, so a short description yields a minimal schema.
Property Types Are Wrong for the Data
Notion AI sometimes assigns a generic Text property to fields that should be Number, Date, or Select. For example, a “deal value” field may become Text instead of Number. Click the property header, choose Edit property, and change the type. The AI does not validate data types against real-world usage.
AI Generated a Relation to a Non-Existent Database
If your prompt mentions a related database, Notion AI may create a Relation property that points to a database that does not exist. The relation appears broken with a red indicator. Either create the related database first or delete the Relation property and create a manual relation later.
Sample Data Contains Placeholder Text
The AI inserts fictional sample data such as “Company A” or “jane@example.com”. This data is not real and should be deleted before using the database for actual work. Delete each sample row individually or delete all rows at once by selecting them and pressing Delete.
Notion AI Database Generation vs Manual Schema Setup
| Item | AI-Generated Schema | Manual Schema Setup |
|---|---|---|
| Time to create | 10 to 20 seconds | 5 to 15 minutes |
| Property types | Basic types only (no Formula, Rollup, Relation to existing DB) | All types including Formula, Rollup, Relation, and Created By |
| Property accuracy | Depends on prompt detail; often needs manual correction | Full control over each property name and type |
| Sample data | Generated automatically with placeholder values | None; you enter real data |
| Database views | Default Table view only | Any view type (Board, Gallery, Timeline, Calendar, List) |
The AI method is faster for starting from scratch but requires cleanup. Manual setup is slower but gives you complete control over schema design, especially for complex databases with formulas, rollups, and multiple views.
You can now generate a Notion database schema from a text description using the AI Database from description feature. For best results, write a prompt that lists every column name and the type of data it holds. After generation, rename properties, correct property types, and delete sample rows. To speed up future database creation, save your best prompts in a Notion note for reuse.