Blog ·

How to Give ChatGPT More Context Without Typing It All

ChatGPT only knows what you tell it. The facts that change an answer, where custom instructions, memory and Projects keep them, and how to say them aloud.

Typally Insights tiles: 476,777 words dictated, 91:31:07 of talk time, 5,655 transcripts

To give ChatGPT more context, tell it the facts that would change the answer, and keep lasting facts in custom instructions, memory or a Project. Saying it out loud is faster than typing it: people speak about 150 words a minute and type about 40 (common averages, not our measurement).

In short:

I talk to ChatGPT and Claude instead of typing to them. On September 30, 2026, my Typally Insights screen showed 476,777 words dictated and 91 hours 31 minutes of talking. I had used it on 97 of the last 98 days. In my own words: "I use it purely for prompting Claude, ChatGPT, other AI." I build Typally, and it comes up near the end.

Typally Insights tiles: 476,777 words dictated, 91:31:07 of talk time, 5,655 transcripts
My own Insights screen: 476,777 words, 91:31:07 of talking and 5,655 dictations, which I use for prompting AI. My numbers, read September 30, 2026; not a typical result.

What context does ChatGPT actually need?

It needs the facts that would change its answer. Test each fact with one question: if this were different, would the reply be different? If yes, say it. If no, leave it out.

Take a reminder email about an unpaid invoice. "They've paid on time for three years" sets the tone: friendly, assuming an oversight. "I already sent one reminder" makes this the second one. "I want to keep them as a client" decides how firm it can be. The accounting app you sent the invoice from changes nothing, so skip it.

Two lists for an unpaid-invoice email: facts that change ChatGPT's reply and facts that don't
The change test on one task. Four facts change the reply; three don't, so they stay out. Illustrative example written for this article.

ChatGPT can't ask about a fact it doesn't know exists. Leave out the first reminder and it has to guess whether this is the first or the second.

Why does the context get left out when you type?

Because every typed sentence costs time, and context feels optional. Typed, it's tempting to cut the invoice request to one line. Said out loud, the facts come along, because they're already in your head.

Typed (12 words)

Write a polite reminder email to a client about an unpaid invoice.

Said, after cleanup (90 words)

I need a reminder email to a client about an unpaid invoice. They're a dental clinic. We've run their social media for three years and they've always paid on time, so I think this is an oversight. The invoice is $1,200 and it's 21 days late. I sent a short reminder last week and got no reply. I want to keep them as a client, so friendly, but clear that it's due. Under 100 words. End with a polite way out, like asking if it went to the wrong person.

The same request typed as 12 words, then said out loud with the ums struck through and the added context marked
Typed, ChatGPT has to guess four things. Said, the same request takes about 40 seconds at a common average of 150 words a minute. Struck through: what the cleanup removes. Illustrative, not a real session.

In the picture, the struck-through words are the ums, a "like" and one restart, which the cleanup takes out. At common averages, typing those 90 words takes about 2 minutes 15 seconds, and saying them about 40.

Where can ChatGPT keep your context so you don't repeat it?

ChatGPT holds context at four levels: every chat, one project, one chat, and the message you're writing. The feature names below are as of October 2026, from OpenAI's help pages. What your plan includes may differ, so check Settings in your own account.

Table of where ChatGPT holds context: every chat, one project, one chat, this message, and what to put in each
Put each fact at the widest level where it stays true, and you say it once. Feature names as of October 2026; what your plan includes may differ.

The setup is the longest context you'll ever give ChatGPT, and it's only text boxes. Custom instructions and a Project's instructions take dictation like any message. Click the field, talk for a minute about your work and your readers, then read it back and trim.

How do you say the context instead of typing it?

  1. Click where the text should go: ChatGPT's message box, or the custom instructions field.
  2. Run the change test in your head, then say the facts that passed, in any order.
  3. Keep talking when you stumble. With a cleanup tool, the ums and restarts never reach the box. Plain dictation leaves them in.
  4. Read it once before you send. In ChatGPT, Enter sends the message, so check before you press it.
  5. If the answer misses, say why: "Too stiff. We're on first-name terms with the owner." That's a fact ChatGPT didn't have.

For a longer walk-through of speaking a whole prompt, see Voice Prompting: How I Speak My ChatGPT and Claude Prompts.

Which free tools let you say it instead of typing?

You can start without buying anything. Three free options, as of October 2026:

For a short prompt inside ChatGPT, its own button may be all you need.

Where does Typally fit?

Typally, voice typing for Mac and Windows, is what I use to talk to ChatGPT and Claude. Hold your shortcut key, say the context, let go, and clean text lands where your cursor is: ChatGPT, Claude, Gemini or a custom instructions field. It records for as long as you hold the key, drops the ums and false starts, and fixes the punctuation.

It listens only while you hold the key. The audio is discarded after transcription, and your history stays on your computer.

Run the change test on your next prompt, then say it instead of typing it. Download Typally free at typally.com for Windows or Mac. The free plan gives you 100 minutes a month and doesn't ask for a card.

What else do people ask about ChatGPT and context?

Does ChatGPT remember context from other chats?

Partly, if memory is on. It can use saved memories and, where your plan has it, details from past chats. Custom instructions apply to every new chat; check Settings, then Personalization, to see what's on.

What's the difference between custom instructions and memory?

You write custom instructions yourself, and they apply to every new chat. Memory is a list of facts saved along the way, by you or by ChatGPT. Use custom instructions for what rarely changes and memory for facts that come up as you work.

Should I use a Project or custom instructions?

Custom instructions for facts true of all your work. A Project for one piece of work with its own files and instructions, which apply to the chats inside it.

Is Typally free?

There's a free plan with 100 minutes of dictation a month, forever, and it doesn't ask for a card. Pro has a 14-day free trial, and the trial needs a card.

Stop typing, start speaking Typally types what you say in every app on Windows and Mac. 100 free minutes every month.
Try Typally free