medium.com 7 min read Original Source

9 better ways to use ChatGPT than collecting more prompts

Herajeng Gustiayu

October 27, 2026 · 13 readers

Quick summary

Use simpler prompts, then keep refining in the same chat
Show ChatGPT the actual file, image or screenshot
Use Projects and memory to keep useful context
Use Temporary Chat when you want a clean slate
Search when facts may have changed, and let recurring tasks check things later

You probably already know the basic ChatGPT routine: ask a question, get an answer, try again if it’s not quite right. That’s how I used it when it first came out too. But ChatGPT has changed a lot since then.

It can now work with files and images, remember useful context, keep ongoing work together, search for current information and run tasks later. Yet it’s still easy to use it the old way: open a blank chat, type one question, copy the answer and leave.

We’re using a 2026 product with a 2023 mental model.

And that’s happening at huge scale. According to Similarweb’s current rankings, ChatGPT is still the most-used standalone AI chatbot by global web traffic, ahead of Gemini and Claude.

My own habits changed too. I used to collect prompts and spend more time trying to word them perfectly. I even wrote about the 10 ChatGPT prompts I used regularly as a product designer.

I still use prompts, obviously. I just care less about making them clever now. The bigger difference comes from knowing what context to give ChatGPT, when to show it the actual file or image, what to let it remember, and when to check the answer.

These are the habits I find more useful now.

1. Don’t overthink the first prompt

A useful prompt usually needs three things:

What you need → Useful context → What you want back

Instead of:

Plan dinner for this week.

Try:

Plan five simple dinners for two adults. I only want to cook for around 30 minutes each night, and I’d like to reuse ingredients where possible. Give me the meals first, then one combined shopping list.

That is usually enough. You don’t need to tell ChatGPT that it is a “world-class expert” or turn a simple request into a page of instructions.

When the result is wrong, I usually find it easier to correct the conversation than rewrite the whole prompt.

2. Keep the conversation going

I still see people treating every ChatGPT prompt like a Google search: ask once, judge the result, then start again.

You can simply continue:

Too detailed. Give me the three things that matter.

Keep the plan, but make it cheaper.

I don’t understand the second point. Explain it without technical terms.

Sometimes the first answer is close and only needs a few tweaks. Other times, seeing the answer makes me realise that my original request wasn’t clear enough.

I’ve found this more useful than endlessly polishing the initial prompt. The conversation itself becomes part of the instruction.

3. Show ChatGPT the actual thing

This is probably the easiest upgrade. If you already have the source, just show it to ChatGPT: upload the document or spreadsheet, send the screenshot, or take a photo of what you’re looking at.

On desktop, you can usually just drag and drop the file or image into the chat. You can also use the attachment button if that’s easier. Explaining something manually often removes useful context.

Imagine writing:

My electricity bill seems much higher this month. It has several sections and different rates. Can you explain why?

If you already have the bill, showing it is more useful. You can ask:

Explain this bill in plain English. What seems to have changed, and what should I check?

Before spending five minutes describing something to ChatGPT, I now ask myself whether I can simply show it.

4. Keep ongoing work in a Project

I used to have separate chats for things I kept returning to. One for an article, another for a trip, another for a work project. Then two weeks later, I would open a new conversation and explain half of it again.

Projects solve much of that because you can keep related chats, files and instructions together. The context builds instead of disappearing into dozens of separate conversations.

A Project could be:

Your job search
A holiday
A house renovation
Learning a new subject
Writing a book
A long-running work project

Imagine you’re planning a trip to Japan. You could keep your flights, hotels, saved places and budget in one Project, then later ask:

We now have a free afternoon in Kyoto. What fits with the rest of our trip?

You no longer need to explain the trip from the beginning. I started using Projects for the same reason, and eventually wrote about the three ChatGPT Projects I set up once and keep returning to.

For things that develop over weeks or months, this has been much more useful than having one giant prompt.

5. Let ChatGPT remember the useful context

Some things are useful once, but annoying to repeat in every chat. You can tell ChatGPT things like:

I prefer short answers unless I ask for detail.

I usually cook for two.

I’m a beginner learning Spanish.

That way, future conversations can start with more useful context already in place, instead of you retyping the same preferences.

I still wouldn’t use memory for every personal detail. I keep it for things that genuinely improve future answers.

6. Use Temporary Chat when you want a fresh start

Sometimes I want the opposite. I don’t want previous context or an ongoing project influencing the conversation.

That is what Temporary Chat is useful for.

I find it handy for researching a surprise present, exploring a completely unrelated subject, or testing something without wanting the conversation to become part of the context I regularly use.

My simple rule is: if the context will become more useful over time, I keep it together. If I deliberately want a clean conversation, I use Temporary Chat.

7. Let it organise the mess

One of my favourite uses of ChatGPT has very little to do with getting expert advice. I give it things I already know but haven’t organised.

For example:

I need to renew my insurance, return a package before Thursday, book the dentist, clean the spare room before my family visits on Saturday, buy groceries and pick someone up on Friday evening.

Then:

Sort this into what I should do today, tomorrow and later this week. Put anything with a deadline first.

I already knew what needed to get done. The annoying part was working out what to do first. The same works for messy notes, packing lists, emails or research.

I use ChatGPT this way a lot to organise messy thoughts without handing over the thinking, which I wrote more about in how I use ChatGPT without outsourcing my thinking.

Sometimes the most useful thing it does is simply turn a messy pile of thoughts into something easier to act on.

8. Ask it to search when things may have changed

I wouldn’t rely only on what an AI model already knows for things such as:

Current prices
Travel requirements
Laws and regulations
Product specifications
News
Recent statistics

When freshness matters, I usually make that part explicit:

Check current sources before answering.

Or:

Find the latest information and link the original sources.

If the subject is messy:

Tell me if reliable sources disagree.

The difference is simple. If I ask, “Help me think through whether I should buy a tablet,” I want reasoning. If I ask, “What are the current prices and specifications of these three tablets?”, I want fresh information.

If the answer includes a number, ranking, or claim I may want to verify, I want the source link too.

9. Let ChatGPT check things for you later

This is one of the newer behaviours that feels genuinely different from how I used ChatGPT a few years ago. ChatGPT can now run some tasks later or repeatedly instead of waiting for you to remember to return.

For example:

Every Monday morning, give me the week’s weather forecast.

Remind me on the 25th of every month to pay this bill.

Every Friday, summarise the biggest news in an industry I follow.

You can also use recurring checks for something you’re waiting for:

Check whether tickets for this event have gone on sale. Tell me when they do.

Previously, ChatGPT depended on me to remember to check something. For some tasks, I can now hand that part over too.

Before I trust the answer, I ask one more question

ChatGPT is much more capable than when I started using it. It can also still give a confident answer that turns out to be wrong.

When the consequence matters, I sometimes ask:

What part of your answer are you least certain about?

Separate the facts from your assumptions.

Or simply:

What should I verify before relying on this?

I do this especially with research, numbers, and subjects I don’t know well enough to spot a bad answer myself.

This has also changed how I write prompts for work. In my more recent article on 12 ChatGPT and Claude prompts for senior product designers, many prompts focus less on getting a smarter first answer and more on asking AI to show uncertainty, separate assumptions, and stop when evidence is missing.

That habit works outside design too.

I spend less time thinking about prompts now

When ChatGPT first appeared, much of the advice focused on prompts: how to phrase a request, what role to assign the AI, and which wording produced the best output.

I still care about prompts, but much less than I used to.

The bigger improvement now comes from continuing the conversation, giving ChatGPT the actual files and images, keeping related work together, letting useful context carry forward, searching when information could have changed, and asking it to check things again later.

That’s what’s changed most for me. If you still mostly open a chat, ask one question, get the answer, and leave, you’re probably using only the simplest part of what ChatGPT can do.