You do not have to type everything
So far this has been a box you type into. It is also a box you can hand things to.
You can give it a document and ask what is in it. You can show it a photograph and ask what you are looking at. You can stop typing altogether and just speak. None of these are advanced features hiding behind a menu you have to hunt for, and none of them require you to convert or prepare anything first.
In plain English
- Upload:
- Handing a file from your computer or phone to the tool so it can read it. Nothing is published or shared with anyone by uploading.
- Attachment:
- The file you have handed over, sitting alongside your message, exactly like attaching something to an email.
- PDF:
- The document format that looks the same everywhere. Most letters, bills, statements and forms you are sent will be one.
- Transcribe:
- Turning spoken words into written text. It is what happens under the bonnet when you talk to it instead of typing.
Giving it a document
Look for a plus or paperclip near the message box. Buttons move around as the app is redesigned, so do not memorise a position, just look for the one that means attach. Choose your file, then type your question alongside it.
The transformative use is not summarising, it is asking. A twenty page tenancy agreement, a pension statement, a school policy, a set of terms and conditions you were never realistically going to read: you can now interrogate them.
I have attached my tenancy agreement. In plain English, tell me: how much notice I have to give, what happens to my deposit, and anything in here that is unusual or that I should be worried about.
The last part of that prompt is the valuable one. Asking what you should be worried about invites it to point at things you did not know to ask about, which is precisely the problem with dense documents.
Sanity check anything that carries consequences. If it tells you a contract clause means X and acting on that costs you money, find the clause yourself and read it. Use it to find the important parts, not to make the final call for you.
It handles spreadsheets and slides too, and it can usually produce a file back: a tidied table, a summary document, a draft to hand on. Ask for it and see.
Showing it a photograph
Take a picture, hand it over, ask about it. That is the whole workflow, and it is far more useful than it sounds.
A handwritten recipe from a relative, typed up properly. A plant with sad-looking leaves. An error message on a washing machine. A restaurant menu in a language you do not read. A cluttered garage, where you want three sensible suggestions. A form someone posted you, where you want the deadline found.
This is a photo of a letter I was sent. Tell me who it is from, what they want me to do, whether there is a deadline, and what happens if I ignore it.
Photographs of paper are the sleeper hit here. Anyone who deals with post they find intimidating, and that is a lot of people, gets something genuinely useful from this on day one.
Same caution as always: it can misread a blurry number or a scrawled word and then reason confidently from the mistake. Good light, flat paper, whole page in frame.
Talking instead of typing
There is a microphone option near the message box, and on the phone apps there is usually a fuller voice mode where you have an actual back and forth conversation out loud, no typing anywhere in sight.
This is worth trying even if you feel faintly silly doing it, because two groups find it changes everything. If typing is slow or uncomfortable for you, whether through arthritis, eyesight, or simply never having got on with keyboards, the barrier disappears. And if you think better out loud, you can ramble for two minutes about a problem and get back something structured, which is a trick that is hard to reproduce by typing.
I am going to talk at you for a bit about a problem at work and it will not be organised. Listen, then give me back the three main things I actually said and one question I have not thought about.
Yes, you will feel odd talking to your phone the first time. Everyone does. It lasts about four minutes, and then you are the person who dictates a shopping list while cooking.
The other assistants can do all of this too. Claude and Gemini both take documents and images, and both have voice options. If ChatGPT ever refuses a file or fumbles a photograph, trying the same thing elsewhere is entirely reasonable.
Checkpoint
You can hand it documents and photographs and talk to it out loud, which turns it from a writing tool into something you can point at anything.
๐ Quiz
Question 1 of 4What is the most valuable thing to do with a long document you have uploaded?