The one-line answer
Without AI, the core features are all still there — importing screenshots, recognizing text, searching content, exporting notes, syncing across devices. None of these have ever been tied to AI, from day one of the product's design. AI is an optional layer stacked on top, not a gate you have to unlock the core features through — which is the exact opposite of a lot of "AI-first" productivity tools.
What you already get without AI
- On-device OCR: Apple's own Vision framework recognizes English and Chinese text inside a screenshot right on your device. The whole process is offline, with nothing uploaded — it's a capability of Apple's own framework and never touches a third-party server.
- Full-text search: Search matches against the OCR text recognized inside the image. Remember one word that appeared in the screenshot, and you can find it — no AI-generated title or summary required.
- Three ways to get screenshots in: bulk-import your existing library, send them over from any app's share sheet, or auto-capture via a Shortcut. Importing itself has nothing to do with AI.
- The note detail page can expand the OCR text: the original text OCR recognized can be expanded and viewed on the detail page, then manually selected and copied. Even without AI pulling out a field like "verification code: 123456" for you, you can still press and hold the raw text itself to copy it.
- Export and iCloud sync: your data goes through your own private iCloud library — Apple's end-to-end sync mechanism — so the developer can't see the content, and there's no account to register and no server relay involved.
- The Mac version goes even further: staying true to local-first, the Mac version connects to zero third-party AI. Day-to-day use runs entirely on three native shortcuts — ⌘N to import, ⌘F to search, ⌘E to export — while all three platforms still share the same data through iCloud.
What configuring AI adds on top
Honestly, the AI layer does save real effort. Specifically, here's what it adds:
- auto-generated titles and summaries, so you know what a shot is at a glance in the list without opening the original image to figure it out;
- automatically splitting screenshots into "just glancing at it" versus "needs to be kept" — without AI configured, new screenshots stay conservatively in the main, visible list rather than being misjudged as unimportant and folded away; you simply lose this layer of auto-sorting. See this piece on organizing approaches for the full mechanism;
- pulling key fields like verification codes, order numbers, and amounts out of receipt-style screenshots into a one-tap-to-copy tag, so you don't have to hunt through the whole block of text yourself — see this piece for details;
- pulling out lines from the OCR text worth keeping and collecting them into the Highlights tab, where they can be exported as shareable cards.
These all share one thing in common: AI works on text that local OCR has already recognized — it never uploads the original screenshot image itself for analysis. For which provider to configure and how to pick one, see this piece on BYOK and the 17 supported providers.
Who should hold off on AI
If your screenshots are mainly reference material for your own eyes, you don't care about auto-sorting or summaries, and you'd rather nothing pass through a third party, you can skip AI entirely and use it as a local screenshot repository with full-text search — that usage has been a complete product from day one, not a stripped-down version. If you take a lot of screenshots and care more about not having to sort them yourself, configuring an AI provider is worth it, and the cost usually turns out lower than you'd expect.
