What Is Generative AI and How Does ChatGPT Work? (Simple Explanation with Examples)

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

Generative AI course · Chapter 1 · Estimated reading time: 22 minutes

1. Suppose we are a student writing a resume with ChatGPT

Suppose we are Priya, a final-year BCom student in Pune. Campus placements start next week. We have a resume, but the top part, the "Profile summary", is empty. We do not know what to write.

So we open ChatGPT on our phone and type this:

I am a final-year BCom student from Pune. I know Excel, Tally and basic accounting.
I did a 2-month internship in the accounts team of a CA firm.
Write a 3-line profile summary for my resume. Keep it simple and professional.

In about 3 seconds, ChatGPT writes something like this:

Final-year BCom student from Pune with hands-on knowledge of Excel, Tally and basic
accounting. Completed a 2-month internship with a CA firm, supporting entries,
reconciliations and GST paperwork. Eager to start a career in accounts and finance.

Look at what just happened:

1. We wrote a request in normal English. No coding.
2. The tool wrote new text that did not exist before.
3. The text is about our details, not a copy from some website.
4. It took seconds, not 30 minutes.

That is Generative AI. It generates (creates) new content when we ask for it in plain language.

2. What is Generative AI? (now the simple definition)

Now that you have seen it, the definition is easy.

Generative AI is a type of AI that creates new content, like text, images, code, audio or video, when you give it an instruction in normal language.

The instruction you type is called a prompt. The answer it gives is called the output (or the response).

Normal software vs Generative AI

What can Generative AI create?

In this chapter we use only text, because it is free, fast and the easiest place to start. Images and video come in later chapters.

3. How does ChatGPT make an answer? (the "next word" idea)

This is the one idea you must understand. Everything else in this course is built on it.

Start with your phone keyboard

Open WhatsApp and type "Happy". Your keyboard (Gboard or the iPhone keyboard) suggests the next word: "birthday", "Diwali", "new year".

How does it know? It has seen millions of messages. After "Happy", the word "birthday" comes very often. So it guesses the most likely next word.

ChatGPT does the same thing, at a very big scale

ChatGPT is a Large Language Model (LLM). "Large" because it learned from a huge amount of text: books, websites, articles and code. From all that text, it learned which words usually come after which words.

When you send a prompt, it writes the answer one small piece at a time:

1. It reads your full prompt.
2. It guesses the most likely next word (actually a small piece of a word, called a token).
3. It adds that word to the answer.
4. It reads everything again, prompt plus the new word, and guesses the next word.
5. It repeats this hundreds of times, very fast, until the answer is complete.

That is why you see the answer appear word by word on the screen, like someone typing.

Worked example: guess the next word yourself

Read each line and guess the last word before you look at the right column.

You just did what an LLM does. You did not "search" anything. You guessed from what you have read before. The AI does the same thing, but it has read far more than any human.

Two important results of the "next word" idea

1. Answers change every time. The AI picks from several likely words, not always the top one. So if you ask the same question twice, you get two different (but similar) answers. This is normal.
2. It can sound right and still be wrong. The AI is guessing words that sound likely. It is not checking a fact database. So sometimes it writes a confident sentence that is false. This is called a hallucination.

4. What Gen AI is good at, and where it goes wrong

Good at

1. Writing first drafts: emails, leave letters, resume lines, captions, product descriptions.
2. Explaining a hard topic in simple words: "Explain GST input credit like I am 15."
3. Changing the style of text: shorter, more polite, more formal, bullet points.
4. Ideas: "Give me 10 names for a tea stall near a college."
5. Summaries: paste a long article and ask for 5 key points.
6. Small code and formulas: Excel formulas, simple Python, SQL queries.

Not good at (be careful)

1. Fresh facts: today's train timing, today's gold price, today's news. Its knowledge stops at a certain date, unless the tool searches the web.
2. Exact numbers and calculations in long problems. Always re-check totals.
3. Facts about small or local things: a small shop in your town, a local college rule.
4. Knowing you: it only knows what you tell it in the chat.

Worked example: see an old or wrong answer with your own eyes

1. Open ChatGPT, Gemini, Claude or Copilot.

2. Type this prompt:

   Give me the exact departure time of every train from Pune to Mumbai tomorrow morning.
   

3. Read the answer.

What you will usually see: either the tool says "please check IRCTC or the NTES app for live timings", or it gives a list that looks correct but may be old or wrong. Both answers teach the same lesson: for live data, use the official app. Use AI to write and explain, not as the final source of facts.

Worked example: it sounds right, but is it right?

Now try a question where the AI must "remember" a small detail.

1. Open a new chat.

2. Type:

   Give me the full names of 3 books written by Ravindra Bagale, with the year each was published.
   

3. Read the answer.

What you may see: some tools say "I could not find books by this person". Others invent three nice-sounding book titles with years. Those titles look real, but the AI made them up, because guessing likely words is all it does. This is a hallucination.

How to protect yourself, every time:

1. Ask: Which part of your answer are you not sure about?
2. Ask the tool to show sources: Give me links I can check. (Gemini, Copilot and ChatGPT with search can show links.)
3. Open the link yourself and read it. If there is no real source, do not use the fact.

5. Open your first AI tool (free)

You need only a phone or laptop, internet, and a Google or Microsoft account. All four tools below have a free plan.

What is free in October 2026? All four tools have a free plan. ChatGPT's free plan gives everyday text chats, with limits on file uploads, images and voice. The cheapest paid plans in India, ChatGPT Go and Google AI Plus, cost about ₹399 a month. You do not need any paid plan for this chapter or the next one. Plans and prices change often, so check the tool's own pricing page before you pay.

Which one should I start with?

Any one is enough. Every prompt in this chapter works in all four.

Steps to open ChatGPT (the others are almost the same):

1. Open the browser on your laptop or phone.
2. Go to chatgpt.com.
3. Click Sign up (or Log in if you already have an account).
4. Choose Continue with Google and pick your Gmail account.
5. You will see a big text box that says something like "Ask anything".
6. That box is where you type your prompt.
7. Press Enter (or the arrow button) to send it.

6. Worked example 1: write a leave email (your first useful prompt)

Situation: you work at an office in Pune. You have a fever and need leave tomorrow.

1. Open ChatGPT.

2. Copy and paste this prompt exactly:

   Write a short email to my manager Anil asking for sick leave tomorrow, 20 October.
   I have a fever. I will be available on phone for anything urgent.
   My colleague Sneha will handle my daily report.
   Keep it polite and under 80 words.
   

3. Press Enter.

4. Read the answer.

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

Subject: Sick Leave Request for 20 October

Dear Anil,

I am not feeling well and have a fever, so I would like to request sick leave
for tomorrow, 20 October. Sneha has kindly agreed to handle my daily report.
I will be available on phone for anything urgent.

Thank you for understanding.

Regards,
[Your Name]

Why this prompt worked: look at what we told it.

1. Who gets the email: manager Anil.
2. What we want: sick leave, tomorrow, 20 October.
3. Details: fever, available on phone, Sneha will do the report.
4. Limits: polite, under 80 words.

The AI did not have to guess. It used our details. This is the main secret of good prompts, and it is the full topic of the next chapter.

7. Worked example 2: explain a hard topic in simple words, in 4 tools

Situation: your younger brother asks, "How does UPI money go from my phone to the shopkeeper?" You want a simple answer.

1. Open ChatGPT in one tab, Gemini in a second tab, Claude in a third and Copilot in a fourth.

2. Paste the same prompt in all four:

   Explain how a UPI payment works, from my phone to the shopkeeper's bank account.
   Explain it to a 12-year-old. Use 5 short steps. Use one example with Rs 50 for chai.
   

3. Press Enter in each tab.

4. Compare the answers using the table below.

What you will usually see:

The exact answers will differ. The point is: all four understand normal English, and the same good prompt works in all of them. Pick the tool you like. In this course, every prompt works in any of the four, unless we say otherwise.

A typical answer looks like this:

1. You scan the shopkeeper's QR code with Google Pay, PhonePe or Paytm.
2. You type Rs 50 and enter your UPI PIN. The PIN proves it is really you.
3. Your app sends the request to NPCI, the system that connects all banks for UPI.
4. NPCI asks your bank to take Rs 50 from your account and tells the shopkeeper's bank to add Rs 50.
5. Both of you get a message: "Rs 50 paid". It takes about 2 seconds. Chai time!

8. Worked example 3: keep talking in the same chat (follow-up prompts)

A chat remembers what you said earlier in the same conversation. So you do not need to type everything again. You can improve the answer step by step.

Continue in the same chat as Worked example 1 (the leave email):

1. Type: Make it even shorter, 3 lines only.
2. Read the new answer. It is the same email, shorter.
3. Type: Now write a WhatsApp message to Anil with the same information.
4. Read the answer. It is now short and informal, with no subject line.
5. Type: Add that I will share a doctor's note when I am back.
6. Read the answer. The new line is added.

Expected output after step 5:

Hi Anil, I have a fever and need sick leave tomorrow (20 Oct). Sneha will handle
the daily report. I am available on phone if anything is urgent, and I will share
the doctor's note when I am back. Thanks!

What we learned:

1. You do not need a perfect prompt the first time.
2. Ask, read, then fix with a short follow-up.
3. If you click New chat, the AI forgets the old chat. Start a new chat when you change topic, so old details do not mix in.

9. Worked example 4: a study plan from a messy request

Situation: suppose we are Rahul, a fresher who just joined Infosys in Pune. Your team says you must clear the AWS Cloud Practitioner exam in 4 weeks. You have 1 hour every evening.

1. Open any of the four tools.

2. Paste this prompt:

   I am a fresher at Infosys. My team wants me to pass the AWS Certified Cloud Practitioner
   (CLF-C02) exam in 4 weeks. I can study 1 hour every evening, Monday to Friday, and 3 hours on Sunday.
   Make a simple week-by-week study plan in a table with columns: Week, Topics, Practice.
   

3. Press Enter.

4. Check the answer: it should be a table with 4 rows (Week 1 to Week 4).

Expected output (shortened):

5. Now type a follow-up: Turn week 1 into a day-by-day plan.
6. You get a plan for Monday, Tuesday and so on.

Check the plan against the official facts. On the official AWS page (October 2026), CLF-C02 has 65 questions in 90 minutes and costs 100 USD. In India the foundational voucher shows as ₹8,553 before 18% GST on the Pearson VUE AWS store. The AI plan may not know the current numbers, so always confirm on aws.amazon.com/certification before you book.

10. Worked example 5: ask the AI to ask you questions

Sometimes you do not know what details to give. Then let the AI ask you.

1. Open a new chat.

2. Paste:

   I want to write a LinkedIn post about finishing my first Excel course.
   Before you write anything, ask me 4 short questions to understand what to write.
   

3. The AI replies with questions, for example: "Which course? What did you learn? Any project? Who do you want to thank?"

4. Answer them in one message, for example:

   1. Excel for Business on ravindrabagale.com
   2. VLOOKUP, pivot tables and charts
   3. I made a monthly expense tracker
   4. Thank my trainer Ravindra Bagale
   

5. Now the AI writes the post using your answers.

This trick is very useful for beginners. When you do not know what to tell the AI, tell it to ask you first.

11. What you must never type into an AI tool

What you type may be stored by the company and, on some free plans, may be used to improve their models (you can turn this off in settings in most tools). So follow these rules from day one.

Never paste:

1. Passwords, OTPs or UPI PINs.
2. Aadhaar number, PAN number, bank account numbers or card numbers.
3. Your company's private code, client names, salary sheets or contracts, unless your company has approved that AI tool for work.
4. Other people's private details (phone numbers, addresses, medical reports).

Turn off "use my chats for training" (optional, takes 1 minute). The menu names change often, so look for these words:

Even with this switched off, never paste the items in the list above.

Safe way: replace private details with dummy words before you paste. For example, write [CLIENT NAME] instead of the real client name, and Rs X instead of the real salary.

12. Words you learned in this chapter

13. Beginner mistakes and quick fixes

14. Try at home

Recap

Did you try all five examples? Good. In the next chapter we turn "who, what, details and limits" into a simple 4-part formula for every prompt.

Next chapter: Prompt Engineering for Beginners: Role, Task, Context and Format (with 5 Examples)

Ravindra Bagale, trainer: linkedin.com/in/ravindra-bagale. The people in the stories (Priya, Rahul, Anil, Sneha) are made up for learning. Company names are used only to make the situations real; nothing here describes how those companies actually work. AI answers change every time, so your output will be similar, not the same. Plans and prices were checked on 19 October 2026 and change often.