How to Use ChatGPT Step by Step: Chats, Follow-ups, Memory and Projects

By Ravindra Bagale, trainer · LinkedIn · Last updated: 19 October 2026

Generative AI course · Chapter 4 · Estimated reading time: 15 minutes

1. Suppose we need a letter to the society secretary

Suppose we are Rohan. We live in a housing society in Thane. The parking spot below our building has had a broken light for 3 weeks, and two-wheelers are getting scratched at night. We want to write a letter to the society secretary.

We open ChatGPT and type:

Write a letter to my society secretary about a broken parking light.

Expected output (yours will be similar, not the same):

Dear Secretary,

I am writing to bring to your attention that the light in the parking area is broken.
This is causing inconvenience to residents. I request you to kindly look into this matter
at the earliest.

Yours sincerely,
[Your Name]

It is polite, but weak. No building name, no dates, no reason why it matters. Most beginners now delete the chat and start again with a new long prompt. Do not do that. Stay in the same chat and fix it with short follow-ups.

Follow-up 1 (add the facts):

Add these facts: I am Rohan Desai, flat B-304, Shanti Heights, Thane.
The light in the B-wing parking has been broken for 3 weeks.
Two scooters were scratched at night last week.

Follow-up 2 (change the tone):

Make it firm but still polite. Ask for the repair within 7 days.

Follow-up 3 (change the format):

Keep it under 120 words. Add a subject line. At the end add "Copy to: Society Chairman".

Expected output after 3 follow-ups:

Subject: Repair of broken light in B-wing parking

Dear Secretary,

I am Rohan Desai from flat B-304, Shanti Heights. The light in the B-wing parking
has not been working for 3 weeks. Last week, two scooters were scratched at night
because the area is dark.

I request you to get the light repaired within 7 days, so that residents and their
vehicles are safe.

Thank you for your help.

Yours sincerely,
Rohan Desai, B-304

Copy to: Society Chairman

Each follow-up changed only one thing: facts, then tone, then format. ChatGPT remembered the earlier letter because we stayed in the same chat. This is the most useful habit in this chapter.

2. Sign in and learn the main buttons

1. Open chatgpt.com in a browser, or install the ChatGPT app by OpenAI from the Play Store or App Store. Check that the publisher is OpenAI.
2. Tap Sign up or Log in. You can use an email address, or a Google, Microsoft or Apple account.
3. After you sign in, you will see the main screen.

About the free plan (checked October 2026): OpenAI's pricing page says the free plan has unlimited everyday text chats, with limits on uploads, image creation, voice and deep research. When you reach a limit, ChatGPT tells you and shows when it resets.

3. Chats and follow-ups: how ChatGPT "remembers" inside one chat

Inside one chat, ChatGPT reads the whole conversation so far every time you send a message. That is why "make it shorter" works: it knows what "it" is.

Short follow-ups that you can copy:

When a chat becomes very long (many pages), it can start to lose track of early details. Then ask: Summarise everything we decided so far in 5 bullet points. Copy that summary into a new chat and continue there.

4. Edit your message, try again, and branch

Three buttons save you a lot of typing.

Edit a message you sent:

1. Move your mouse over your own message (on the phone, long-press it).
2. Tap the pencil (edit) icon.
3. Change the words and press Send.
4. ChatGPT writes a new answer from that point. The later messages are replaced in this view.

Try again (a new answer to the same prompt):

1. Under ChatGPT's answer, find the try again / regenerate icon (a circular arrow).
2. Tap it. You get a fresh answer to the same message.

Branch in new chat (try two ideas):

1. Under an answer, open the more (•••) menu.
2. Choose Branch in new chat.
3. You now have a copy of the chat up to that point. Try one idea here, and a different idea in the original chat.

Worked example: two versions of the same letter. In the letter chat from section 1, branch after follow-up 3. In the new branch, type Make it friendly, as if I know the secretary well. In the original chat, type Make it very formal. Now you have both versions, and neither one was lost.

5. Temporary chat: a chat that is not saved to history

Use a temporary chat when you want a quick conversation that should not stay in your history.

1. Start a New chat.
2. Select Temporary at the top.
3. Choose Unpersonalized (does not use your memory or custom instructions) or Personalized (can use them). Choose before your first message, because you cannot change it later.
4. Type your message.

What OpenAI's help page says about temporary chats (checked October 2026):

• It stays out of your chat history unless you save it.
• It does not create or update memories.
• It is not used to improve models while it stays temporary.
• A copy may still be kept for up to 30 days for safety reasons.

So temporary chat is not a place for passwords or OTPs. Those never go into any AI tool.

6. Memory: what ChatGPT keeps across chats

Inside one chat, ChatGPT reads the conversation. Memory is different: it carries some details from one chat to another, like your name, your job, or "I like answers in short bullet points".

Worked example: turn memory on, test it, then turn it off

Step A: see your settings

1. Open Settings.
2. Select Personalization.
3. Select Memory.
4. Look at the switches you have. Depending on your account you may see Memory, Reference saved memories, Reference chat history, a memory summary or Manage.

Step B: save something on purpose

1. Start a new chat.

2. Type:

   Remember this: I am a BCom student in Nagpur, and I like answers in 5 short bullet points.
   

3. ChatGPT usually replies that it has updated its memory.

Step C: test it in a new chat

1. Start another New chat.
2. Type: Explain what GST is.
3. Look at the answer. If memory is on, it will often be in short bullet points and may use a commerce example.
4. Ask: What do you remember about me? ChatGPT shows what it can use.

Step D: remove it

1. Ask: Forget that I am a BCom student in Nagpur. Or delete it from Settings → Personalization → Memory → Manage (or the memory summary).
2. To stop memory completely, turn the Memory switch off.

Facts from OpenAI's memory help page (checked October 2026):

• Turning memory off does not delete your old chats.
• Deleting a chat does not always delete a saved memory that came from it. Delete the memory too.
• If you turn off Reference chat history, details remembered from past chats are scheduled for deletion within 30 days. Your chats stay unless you delete them.

Data controls: is my chat used for training?

1. Open Settings → Data controls.
2. Look for Improve the model for everyone.
3. If it is on, OpenAI may use your chats to train its models. Turn it off if you do not want that.

7. Projects: a folder with its own files and instructions

A project keeps related chats, files and instructions together. ChatGPT uses what is in the project for every chat inside it. OpenAI's help page lists Projects on the free plan, with up to 5 files per project (checked October 2026).

Worked example: a "Hostel meal plan" project

Suppose we are Sneha, a student staying in a hostel in Pune. We have a small budget, a shared kitchen with one induction stove, and we are vegetarian. Every week we want a simple meal plan and a shopping list.

Create the project:

1. In the sidebar, select New project.
2. Name it Hostel meal plan. Pick any icon and colour.
3. Open the project and find the place for instructions (in the project menu or settings).

4. Paste these instructions:

   I am a vegetarian student in a Pune hostel.
   Kitchen: one induction stove, one pressure cooker, one pan. No oven, no microwave.
   Budget: about ₹700 a week for groceries.
   Answers: simple English, a table first, then a shopping list grouped by shop (vegetable vendor, kirana, dairy).
   This is general meal planning, not medical or diet advice.
   

5. Add a file if you like, for example a short text file with the vegetables you like and dislike.

Use the project:

1. Inside the project, start a chat and type: Make my meal plan for Monday to Friday, lunch and dinner.
2. Next week, start another chat in the same project and type: Same as last week, but no potatoes and add one paneer dish.

Expected output (shortened, yours will be similar, not the same):

Shopping list
Vegetable vendor: onions 1 kg, tomatoes 1 kg, cucumbers 4, ...
Kirana: rice 2 kg, toor dal 500 g, poha 500 g, ...
Dairy: curd 1 kg, paneer 200 g, ...

You did not repeat the hostel, the stove or the budget in the second week. The project remembered them. Check the shopping list prices yourself at your local shop; the AI does not know today's prices in your area.

Project memory setting: in the project menu (•••) select Project settings. Under Memory you can choose Default memory or Project-only memory. Project-only memory keeps this project's chats separate from your other chats.

8. Chat, project or memory? Which one to use

9. The same ideas in Gemini, Claude and Copilot

The names change, but the ideas are the same. Menus move often, so look for these words.

Facts from each company's help pages, checked October 2026. Gemini's memory of past chats needs you to be 18 or over, signed in with a personal Google account, with Keep Activity on.

10. Beginner mistakes and quick fixes

11. Try at home

Recap

Open ChatGPT now, make one project for a job you do every week, and write its instructions. Next week you will thank yourself.

Previous chapter: ChatGPT vs Gemini vs Claude vs Copilot: Which Free AI Tool Should You Use?

Next chapter: How to Use AI on Your Phone: Voice, Camera and Gemini Live

Ravindra Bagale, trainer: linkedin.com/in/ravindra-bagale. The people, flats, society, letters and meal plans on this page are made up for learning. The meal plan is a general example, not medical or diet advice. Button names and settings were checked on the official help pages on 19 October 2026 and can move. AI answers change every time, so your output will be similar, not the same.