Having the AI assistant read a photo or document
A note from school in the backpack, a PDF notice, a prescription, a garage receipt with the next inspection date: copying this information by hand is the main source of missed items. The assistant can read a photo like a document and directly extract events and tasks from it.
What the assistant can read
| Content type | What is understood |
|---|
| Photo taken on the spot | Handwritten or printed text, dates, times, places, amounts |
| Screenshot | Message, text, displayed email, web page, schedule |
| PDF | Notice, invoice, contract, school schedule, prescription |
| Text file (.txt, .csv, .md) | To-do lists, exports, raw notes |
Reading covers printed text and legible handwriting. A photographed paper planner, a schedule table, or an activity calendar are interpreted as sequences of dates, not just as a plain image.
Sending a photo or file in three steps
- Open the Assistant tab.
- Tap the camera icon to take a photo, or the paperclip to attach an image, a PDF, or a text file.
- Add a short instruction if needed: “create the events” or “what should I remember?”
- Send: the assistant replies with what it understood and proposes the corresponding actions.
Without an instruction, the assistant summarizes the document and proposes the most obvious actions. With an instruction, it limits itself to what you ask: extract only the dates, only the amount due, only the list of items to prepare.
Examples that work well
- Photographed school note: “add the field trip to the calendar and create a task to sign the slip.”
- Prescription: “create a daily reminder for the treatment, morning and evening, for one week.”
- PDF invoice: “create a payment task with the stated due date.”
- Term activity schedule: “add every Wednesday practice.”
- Garage receipt: “remind me of the inspection one month before the date.”
Photographing instead of copying: the information moves from paper to the calendar with no manual entry step, so no risk of a copying mistake.
Best practices for reliable reading
- Frame the whole document, flat, with even lighting and no cast shadow.
- One photo per document: two overlapping sheets mix up the dates.
- For a long document, prefer PDF over photo: the text is already usable there.
- Always check the year: a school note stating “Thursday the 12th” is interpreted based on context, and a correction remains possible in one sentence.
Limits and privacy
Attached files are sent to be analyzed, then the retained elements (title, date, due date) are saved in your account; the document itself is not archived in a library. Word and Excel office formats aren't read directly: export them to PDF first. Very degraded handwriting or a blurry photo may produce a partial reading — the assistant then states what it couldn't decipher rather than inventing a date.
Frequently asked questions
Does the assistant read handwriting?
Yes, as long as it's legible: a photographed school note or a handwritten note is interpreted, dates and times included.
Which file formats are accepted?
Images (photo, screenshot), PDFs, and simple text files (.txt, .csv, .md). Word and Excel files must be converted to PDF.
Can multiple events be created from a single photo?
Yes. A photographed schedule can generate several events in one request, for example all of a term's practice sessions.
Are my documents kept?
No. They're analyzed to extract actions; only the created elements (events, tasks) remain in your account.