Notion added AI autofill to databases in late 2023, and it’s one of those features that looks brilliant in a demo but quietly wrecks your workflow if you don’t understand what it’s actually doing.
The pitch is simple: Notion watches how you fill in database properties, then offers to complete the rest for you using pattern recognition and its AI model. It’s meant to save time on repetitive data entry—content calendars, client trackers, product roadmaps, that sort of thing.
But autofill doesn’t just speed things up. It also makes assumptions. And when those assumptions are wrong, you end up with corrupted records that take longer to fix than if you’d entered everything manually.
How autofill actually works
Notion’s autofill pulls from three sources: the structure of your database, the content already in other rows, and its underlying language model’s general knowledge.
If you’re filling in a content calendar and you’ve already logged five blog posts with categories like “SEO,” “monetisation,” and “hosting,” autofill will suggest categories for new rows based on the title or notes you’ve entered. If you type “How to optimise WordPress caching” into a new row, it’ll probably suggest “hosting” as the category.
That’s useful when your database is consistent and your naming conventions are tight. It falls apart when your data is messy, when you use the same words to mean different things, or when you’re tracking anything that requires context Notion doesn’t have.
The AI doesn’t understand your business. It understands patterns in text. If your database has irregular naming, overlapping categories, or nuanced distinctions that matter to you but aren’t obvious from the text, autofill will guess wrong more often than it guesses right.
When to use it
Autofill works best in databases with:
- Repetitive, predictable patterns. Client intake forms, meeting notes, content calendars where the categories are stable and the naming is consistent.
- Low stakes. If a wrong guess costs you five seconds to fix, autofill saves time. If it silently corrupts a financial tracker or product spec, it’s not worth the risk.
- Text-heavy fields with obvious relationships. If you’re logging blog titles and Notion can reliably infer the category from the title, autofill will probably get it right 80% of the time. That’s a good trade.
I use it for content idea backlogs where speed matters more than precision. I don’t use it for anything tied to revenue, client work, or product specs.
When it ruins your data
Autofill becomes a liability when:
- Your database has overlapping or evolving categories. If “monetisation” sometimes means sponsorships and sometimes means courses, autofill will pick one arbitrarily and you won’t notice until weeks later.
- You’re tracking anything with subtle distinctions. Client status (“warm lead” vs. “active conversation”), product stages (“spec’d” vs. “in development”), or anything where the difference matters but isn’t obvious from the row’s title or description.
- Multiple people use the same database. Autofill learns from everyone’s input. If one person uses loose naming and another is strict, the AI will average out their habits and suggest garbage to both.
I’ve seen operators lose entire afternoons cleaning up autofill mistakes in client trackers because Notion confidently filled in the wrong pipeline stage for 30 rows and no one noticed until invoices didn’t match expectations.
How to control it
You can toggle autofill per database. Open any database, click the ••• menu in the top right, then look for Autofill properties. Turn it off globally or disable it for specific properties.
If you keep it on, spot-check the first 10–20 rows after Notion starts making suggestions. If it’s getting things wrong more than 20% of the time, turn it off for that property or that database.
And if you’re working with a team, set a naming convention first. Autofill isn’t a substitute for clean data—it’s a shortcut that only works when your data is already consistent.
One non-obvious tip: if you’re using autofill in a content calendar, create a separate “AI suggestion” property and let Notion fill that instead of overwriting your canonical category field. You can review suggestions in bulk, accept the good ones, and ignore the rest without risking your source of truth.
Got a Notion workflow that’s slowing you down? Reply and tell me what you’re tracking—I’ll cover database setups and automation tricks in a future issue.