How AI and privacy work in CapturedIt

CapturedIt uses AI to make your library searchable and useful — summaries, categories, semantic search, transcription, and answers grounded in what you saved. That only works if you trust us with your content. This guide explains what we do with your data, what you control, and what we never do — without the engineering jargon.

Your content is yours

Everything you capture — links, notes, files, photos, voice memos — belongs to you. CapturedIt stores it in your personal library so you can search, share, and connect it to your tools. You can read our privacy policy, export your data anytime from Dashboard → Security, and delete your account when you want a clean exit.

We do not sell your personal data. We do not use your captures to train AI models for other customers or for general-purpose model improvement. AI runs on your content only to serve your library — better search, summaries, and assistant integrations you turn on.

You choose when AI runs

AI is included in every plan, but automatic processing is never forced on you. In Profile → Preferences you pick:

• Automatically when I capture — After you save, enrichment runs in the background when your preferences allow it.
• Only when I tap AI Assist — Captures save as-is. You decide item by item when to run full AI.

You can also turn off Enable AI processing globally to stop automatic enrichment while keeping AI Assist available per item. Per-type toggles (links, text, images, voice, and so on) control automatic processing only — not AI Assist when you explicitly request it.

What AI does — and what it does not

When AI runs on a non-private item, it works on that item to improve your library. Typical outputs include:

• Links — Extracts the main article text and may add a short summary and categories. Link previews (title, image) are added for everyone regardless of AI settings.
• Text and notes — Long notes may get a summary and categories so you can find them later.
• Images and photos — May get a plain-language description for search and context.
• Voice memos — Speech can be converted to text when transcription is enabled.
• Search — Semantic search uses meaning-based representations so you can find related content even when the words differ. See semantic search explained.
• Just Ask — Library-wide questions retrieve relevant items first, then generate an answer with citations. Private items are excluded. See Just Ask.

AI does not rewrite your originals, publish anything on your behalf, or run on items you mark private. If nothing relevant exists in your library, Just Ask says so — it does not invent content from your saves.

Full vs minimal enrichment

For most capture types you can set an enrichment level in Preferences:

• Full — Summaries, descriptions, categories, and search embeddings. Uses more of your monthly AI allowance.
• Minimal — Categories and search only; no summaries or image descriptions. Uses less of your allowance.

Each item that completes full AI processing counts as one use toward your plan limit. Current usage is visible on Billing. When you reach the limit, automatic processing and AI Assist pause until the next period or you upgrade. You can still capture and browse everything you already saved.

Private items: your off switch

Some captures should never leave your vault — credentials, medical notes, legal drafts, or anything sensitive. Mark an item Private on its detail page and it is excluded from:

• Automatic AI processing and AI Assist
• Sharing and public links
• API, MCP, and connected assistant access
• Just Ask and semantic search over your library

Private is per item, not all-or-nothing. The rest of your library can still use AI and integrations normally.

When content is sent for processing

Your captures are stored securely in your account. Content is sent to AI services only when you allow it — through automatic processing, AI Assist, transcription, or features like Just Ask that you use on purpose. Private items are never included.

Processing happens over encrypted connections. Results — summaries, categories, search embeddings, transcripts — are saved back to your library so you benefit on your next search or query. We use trusted infrastructure providers to operate the service; details are in our privacy policy.

Integrations and API access

If you connect ChatGPT, Claude, Cursor, Slack, or your own scripts via the Context Hub API, access is gated by API keys you create in Dashboard → Security. Best practices:

• Create separate keys per integration so you can revoke one without affecting others
• Use read-only scopes when write access is not needed
• Mark sensitive items private before connecting external tools
• Review API activity periodically to see which apps accessed your library

Setup guides for specific assistants live on our Learn hub. Private items remain excluded from all API and MCP responses.

Transparency and your rights

Security is not hidden behind a support ticket. From Dashboard → Security you can:

• See a summary of what CapturedIt stores (items, categories, connections, shares)
• Manage API keys and view recent API activity by app
• Review active sessions and revoke any you do not recognize
• Download your data or delete your account

Cookie consent, marketing email preferences, and AI toggles live under Profile → Privacy & preferences.

Quick reference