How to Use ChatGPT in 2026: The Complete Guide With 4 Proven Prompt Templates

Updated June 2026  •  14 min read  •  All pricing verified from OpenAI’s official pages in June 2026

If you want to learn how to use ChatGPT properly in 2026, the biggest problem is not the tool — it is the advice. Most guides tell you to “just type a question,” show you a screenshot from 2023, and stop there. Meanwhile, the product itself has changed completely: GPT-4o is gone, there is a ₹399 plan built for India, and features like Projects and memory have quietly become more useful than the chat box itself.

This guide replaces two older ChatGPT articles on AIInsider. Instead of three overlapping posts, we have rebuilt everything into one guide based on how we actually use ChatGPT daily — for client emails, GST paperwork, exam prep material, and small business marketing. Every price is verified from OpenAI’s official pages, and every prompt example here was written for this article, tested, and refined. Nothing is copy-pasted from anywhere.

One honest note before we start: ChatGPT will confidently get things wrong. We cover exactly when not to trust it in the limitations section — read that part even if you skip everything else.

Getting Started: How to Use ChatGPT for the First Time

You need exactly one thing: a free account at chatgpt.com or the official ChatGPT app (Android/iOS). Sign up with Google, Apple, or any email. You can technically chat without logging in, but skip that — without an account you lose chat history, memory, and Projects, which are the features that make ChatGPT genuinely useful.

The real first decision is whether to pay. India gets a pricing structure that most countries do not, so here is the honest breakdown.

Free vs Go vs Plus in India (June 2026)

PlanPrice (India)What you actually getWho should pick it
Free₹0GPT-5.5 with a limited number of messages per 5-hour window, then a smaller fallback model. Limited image generation and file uploads. Projects included.Anyone starting out. Use free for 2–3 weeks before paying anything.
Go₹399/monthRoughly 10x the free tier’s messages, image generation and file uploads, longer memory, 4 GB file storage. No full model picker control.Students, daily users who keep hitting free limits. India was the launch market for this plan.
Plus₹1,999/monthEverything in Go, plus the model picker (choose GPT-5.5 Instant or Thinking manually), higher limits, deep research, agent features, and early access to new tools.Professionals using ChatGPT for hours daily, or anyone who needs the Thinking model on demand.
Pro~₹19,900/monthNear-unlimited access to the heaviest reasoning models.Almost nobody reading this. Skip it unless your employer pays.

Source: chatgpt.com/pricing and OpenAI’s ChatGPT Go announcement. India pricing is charged in rupees via UPI or card.

Pricing note: OpenAI changes plan limits more often than prices. The “limited messages per 5 hours” on the free tier is deliberately dynamic — it shrinks during peak load. If your free limit feels different week to week, that is normal, not a bug.

Our actual recommendation, having paid for both: start free. If you hit the message wall daily, Go at ₹399 is the best-value AI subscription available in India right now — cheaper than two months of a Netflix mobile plan. Upgrade to Plus only when you find yourself specifically needing the Thinking model for complex work, because that manual control is the genuine difference, not the marketing bullet points.

The ChatGPT Interface in June 2026: A Quick Tour

If you last used ChatGPT in 2024 or even mid-2025, the interface has changed underneath you. The most important shift: GPT-4o and the original GPT-5 were retired in February 2026. Everything now runs on GPT-5.5, and the old habit of model-shopping is mostly gone. Here is what each part of the screen does now.

The model picker (paid plans only)

Free users get no picker — ChatGPT routes your message automatically. On Go, Plus and Pro you can choose between GPT-5.5 Instant (fast, conversational) and GPT-5.5 Thinking (slower, reasons step-by-step before answering). The practical rule we use: Instant for writing, translation, and everyday questions; Thinking for anything involving calculations, multi-step planning, code, or “find the flaw in this” tasks. Even on Instant, ChatGPT will silently escalate hard questions to Thinking — so free users are not as disadvantaged as the plan table suggests.

The sidebar

Left side holds your chat history, Projects (folders with their own files and instructions — covered in the advanced section), and GPTs (custom bots, now a minor feature in practice). The search bar at the top of the sidebar searches your old chats — genuinely useful once you have a few hundred conversations.

The input box

The “+” button attaches files and photos. The microphone icon does quick dictation; the waveform icon opens full voice mode, a real-time conversation. The globe/search toggle makes ChatGPT browse the live web — turn this on for anything involving prices, news, or current rules, because the model’s built-in knowledge has a cutoff date.

Temporary chat

Top-right menu. Conversations in temporary chat are not saved to history and do not feed memory. Use it for anything sensitive — salary negotiations, medical questions, anything you would not want surfacing in a future “remembered” detail.

How to Use ChatGPT Effectively: Prompting Fundamentals

Here is the single most useful thing in this entire guide. Almost every disappointing ChatGPT answer comes from a one-line prompt. The fix is a five-part structure we will call RCTFC:

PartWhat it meansExample fragment
R — RoleWho should ChatGPT be?“You are a CA who handles GST compliance for small traders.”
C — ContextYour situation, in 2–4 lines“I run a 6-person garment export unit in Tirupur. My client is 45 days late on payment.”
T — TaskThe specific output you want“Draft a payment reminder email.”
F — FormatShape of the answer“Under 150 words, with subject line, in polite but firm English.”
C — ConstraintsWhat to avoid or include“Do not threaten legal action yet. Mention invoice number and due date placeholders.”

You do not need all five parts every time. But Context is the one Indians skip most, and it is the one that matters most — ChatGPT’s default answers are written for a generic American user. Tell it you are in India, what city, what budget, what audience, and the quality jumps immediately.

Two more fundamentals before the examples:

Iterate instead of restarting. ChatGPT remembers the conversation. If the first answer is 70% right, say “make it shorter and less formal” rather than rewriting your prompt from scratch. The second response is usually where the good output lives.

Ask it to ask you questions. Adding “Ask me 3 clarifying questions before answering” to any complex request is the cheapest quality upgrade that exists. It forces the model to fill context gaps you did not know you had.

Before/After Prompt Examples for Indian Users

These four scenarios cover what Indian users actually do with ChatGPT, based on our reader emails and our own use. Each shows the lazy prompt, the rebuilt prompt, and why the rebuild works. Copy the green boxes and replace the bracketed parts.

Example 1: GST invoice payment reminder (small business)

❌ What most people type

write an email asking client to pay pending invoice

✅ What works

You are an accounts manager at a small Indian business writing to a long-term client.

Context: We supplied goods against Invoice [INV-2026-041] dated [2 April 2026] for ₹[1,18,000] including 18% GST. Payment terms were 30 days. It is now 40+ days overdue. The client relationship is good and I want to keep it that way, but my own GST liability on this invoice is already paid, so the delay is costing me working capital.

Task: Draft a payment reminder email.

Format: Subject line + body under 140 words. Polite, professional, but with a clear ask and a specific date.

Constraints: Mention the invoice number, amount and original due date. Mention (gently) that GST on this invoice has already been deposited from our side. Do not mention legal action. End by offering to share the invoice copy again. Then give me a second, firmer version I can send 10 days later if needed.

Why this works: the GST detail (“we’ve already deposited the tax”) is the kind of leverage a generic prompt never produces — it gives the client a concrete, face-saving reason to prioritise you. And asking for the firmer follow-up version in the same prompt saves you a second session.

Example 2: IELTS Writing Task 2 practice

❌ What most people type

help me prepare for IELTS writing

✅ What works

You are an IELTS examiner. I am targeting Band 7.5 in Writing; my last mock scored 6.5, and feedback said my Task 2 essays have weak paragraph development.

I will paste a Task 2 essay below. Evaluate it like an examiner:
1. Give a band score for each of the four criteria (Task Response, Coherence & Cohesion, Lexical Resource, Grammar) with one line of justification each.
2. Quote my 3 weakest sentences and rewrite each at Band 7.5 level.
3. Tell me the ONE habit to fix before my next essay — not a list, just one.

Essay question: [paste question]
My essay: [paste essay]

Why this works: it converts ChatGPT from a vague “tips” machine into a feedback loop on your own writing. The “just one habit” constraint matters — without it, you get a 15-point list you will never act on. Do this with one essay daily for three weeks and you have a personal examiner for ₹0. (Cross-check band logic against the official IELTS descriptors occasionally; the model is a good coach but not an official scorer.)

Example 3: UPSC study plan (working professional)

❌ What most people type

make a study plan for UPSC prelims

✅ What works

You are a UPSC mentor who specialises in working professionals.

My situation: I work 9:30–6:30, Mon–Sat, in Pune. I can study 2 hours on weekdays (6–7 AM and 10–11 PM) and 6 hours on Sundays. Prelims is in [May 2027]. I have completed Polity (Laxmikanth, one reading) and Modern History. I am weakest in Economy and Environment. Optional decided: [Sociology]. I have not started answer writing or PYQs.

Task: Build a 12-week plan for the next quarter only — not the full journey.

Format: Week-wise table with columns: weekday focus, Sunday focus, weekly PYQ target, current affairs time.

Constraints: Every 4th week must be a revision-only week. Current affairs maximum 30 min/day. Tell me explicitly what to IGNORE for now. After the table, ask me 3 questions that would help you improve the next quarter’s plan.

Why this works: real constraints (job hours, completed subjects, weak areas) produce a plan you might actually follow, instead of the fantasy “8 hours daily” schedules that fill YouTube. The “what to ignore” instruction is critical for UPSC specifically — the syllabus anxiety is the real enemy, and ChatGPT is surprisingly good at scoping when asked directly.

Example 4: Small business marketing (Instagram for a local brand)

❌ What most people type

write instagram captions for my clothing boutique

✅ What works

You are a social media writer for small Indian D2C brands.

Context: I run a kurti and co-ord boutique in Jaipur. Price range ₹800–₹2,400. Audience: women 22–38 in tier-1/tier-2 cities who follow fashion reels. Brand voice: warm, slightly playful Hinglish — think “office se shaadi tak” energy, never corporate. Raksha Bandhan is 3 weeks away.

Task: Give me 5 caption options for a Rakhi-gifting reel showing our new festive collection.

Format: Each caption = hook line (under 8 words) + 2-line body + CTA + 5 hashtags mixing big and niche tags.

Constraints: No “Elevate your wardrobe” type phrases. At least 2 captions should use Hinglish naturally. CTA should push DMs, not link-in-bio, because DMs convert better for us.

Why this works: voice description plus a banned-phrase list is the difference between captions that sound like every AI account and captions that sound like your brand. The “office se shaadi tak” reference does more work than three paragraphs of brand guidelines — give ChatGPT a vibe anchor and it extrapolates well.

Pattern across all four: notice that every good prompt contains a number (₹1,18,000, Band 7.5, 12 weeks, 5 captions) and at least one thing to avoid. Specific inputs in, specific outputs out. That is 80% of prompting skill; the rest is iteration.

Common Mistakes People Make With ChatGPT (Stop Doing These)

We have watched friends, readers, and frankly ourselves make every one of these. In rough order of damage:

1. Treating it like Google

Typing a 4-word query, reading the first response, closing the tab. ChatGPT is a conversation tool — its value is in the follow-up. If you are not sending at least one “now change X” message, you are using maybe 30% of the product.

2. Trusting numbers, dates, and legal facts without checking

GST rates, IPC/BNS sections, mutual fund returns, visa rules, medicine dosages — ChatGPT will state wrong versions of all of these with total confidence. The rule: use it to understand, never to cite. If a number ends up in an invoice, a court document, or a prescription decision, verify it from the official source. Turning on web search helps but does not make it citation-safe.

3. Pasting confidential data

Client lists, employee salaries, unpublished financials, Aadhaar/PAN details — do not paste these into a regular chat. By default, consumer ChatGPT conversations can be used to improve OpenAI’s models unless you opt out in Settings → Data Controls. At minimum, use temporary chat for sensitive material and strip identifying details.

4. Publishing ChatGPT output as-is

This one we will say with some humility: Google’s 2026 spam updates hammered sites that published lightly-edited AI text at scale — including, candidly, parts of this site, which is why this guide is a rebuilt, consolidated article. The same applies to your LinkedIn posts, college assignments, and client deliverables. ChatGPT output is a draft. The moment readers (or professors, or Google) detect the unmistakable default AI voice, your credibility is spent. Edit ruthlessly, add what only you know, delete every sentence you would not say aloud.

5. Asking for everything in one giant prompt

“Write me a complete business plan with marketing strategy, financials and a website copy” produces 2,000 words of confident mush. Break big jobs into a sequence: structure first, then one section per message, critiquing as you go. This is exactly what Projects are for — next section.

6. Never opening Settings

Ninety percent of users have never set custom instructions or checked what memory has stored about them. Five minutes in Settings → Personalization improves every future conversation. Details below.

Advanced Ways to Use ChatGPT: Projects, Memory, Voice and Files

Once prompting basics are comfortable, these four features are where ChatGPT stops being a chatbot and starts being a workspace.

Projects: folders with a brain

Projects (in the sidebar, available on the free plan too as of 2026) let you group chats, upload reference files, and set project-specific instructions. Every chat inside the project automatically knows the files and follows the instructions.

Concrete setup we use — a “Client: [Name]” project for each ongoing client: upload the proposal PDF and rate card, set instructions like “All emails formal but warm, sign off as Shekhar, amounts always with GST shown separately.” Every future email, summary, or follow-up for that client starts pre-loaded with context. Same pattern works for “UPSC Prep” (upload your notes, instruct it to always quiz before explaining) or “Boutique Marketing” (upload your best-performing captions as voice reference).

Memory: useful, slightly creepy, worth managing

ChatGPT remembers details across chats — your profession, preferences, ongoing projects. This is why it might suddenly know you are a teacher in Indore. Go to Settings → Personalization → Memory to see and delete individual memories. Two tips: explicitly say “remember that I prefer answers in simple English with Indian examples” to seed it deliberately, and audit the list monthly — it accumulates stale facts (it kept believing one of us was still preparing for CAT a year after the exam).

Custom instructions: set once, benefit forever

Settings → Personalization → Custom Instructions has two boxes: who you are, and how ChatGPT should answer. A genuinely effective pair for an Indian professional: About you: “Small business owner in [city], [industry]. Comfortable in English and Hindi. Time-poor.” How to respond: “Be direct. Use Indian context — ₹, GST, Indian examples — by default. Give the answer first, explanation after. Flag uncertainty instead of guessing.”

Voice mode: the most underused feature in India

The waveform icon opens a real-time spoken conversation that handles Hindi, Hinglish and accented English well. Free users get a limited daily allowance; paid plans get much more. Best real uses we have found: practising English interviews aloud on a commute, explaining a concept back to it and having it correct you (brilliant for exam prep), and helping parents who type slowly but speak fluently use AI at all.

File analysis: your underpaid analyst

Attach PDFs, Excel/CSV files, Word docs, or photos with the “+” button. Strong uses: “Summarise this 40-page tender document and list every deadline and penalty clause,” “Here is my sales CSV — which product line is dragging margin down?”, or photographing a contract clause and asking for a plain-English explanation. Free tier allows limited uploads; Go raises the cap roughly 10x with 4 GB of storage. One caution: for precise totals across large spreadsheets, ask it to show its calculation — extraction is reliable, arithmetic across hundreds of rows sometimes is not.

Honest Limitations: When NOT to Trust ChatGPT

A guide on how to use ChatGPT is incomplete — and frankly dishonest — without this section. After years of daily use, here is where the tool genuinely fails:

It hallucinates, fluently. ChatGPT generates plausible text; it does not “look things up” unless web search is on. It has produced, for us personally: a non-existent income tax section, a confidently wrong PPF interest rate, and a fabricated quote attributed to a real RBI governor. GPT-5.5 hallucinates less than older models, but “less” is not “never,” and the errors are harder to spot because the surrounding text is so polished.

Do not let it decide anything medical, legal, or financial. Understanding a diagnosis your doctor gave you — fine, helpful even. Deciding whether to take medication, sign a contract, or move savings based on its answer — no. It cannot examine you, read the full contract context, or know your risk profile, and it will not be liable when it is wrong. You will be.

India-depth is thinner than it looks. It speaks fluent Hinglish but its training skews heavily toward US/UK sources. State-specific rules (stamp duty, professional tax, local body procedures), regional language nuance beyond Hindi, and anything hyperlocal — quality drops sharply. The more specifically Indian your question, the more you should verify.

It mirrors you. Ask “why is X a good idea?” and it argues for X. Ask “why is X a bad idea?” and it argues against, equally fluently. For real decisions, always ask for the case against your own preference — and consider getting a second opinion from a different AI entirely (see our ChatGPT vs Claude vs Gemini comparison).

Recency has edges. Without the search toggle, its knowledge stops at its training cutoff. With search on, it is only as good as the pages it finds — which for Indian topics are sometimes outdated themselves. Anything involving “current price/rule/rate”: search on, then verify the source link it cites.

Frequently Asked Questions

Is ChatGPT free to use in India in 2026?

Yes. The free tier includes GPT-5.5 with a limited number of messages per 5-hour window, limited image generation and file uploads, and Projects. If you keep hitting limits, ChatGPT Go costs ₹399/month in India — see the plan comparison above.

What is the difference between ChatGPT Go and ChatGPT Plus?

Go (₹399/month) gives roughly 10x the free tier’s messages, uploads and image generation, plus longer memory. Plus (₹1,999/month) adds the manual model picker — you can force GPT-5.5 Thinking for hard problems — along with higher limits and tools like deep research. Most students and casual daily users are fully served by Go; Plus is for heavy professional use.

Which model does ChatGPT use in 2026?

GPT-5.5. Older models including GPT-4o, GPT-4.1 and the original GPT-5 were retired from ChatGPT in February 2026. Paid users can choose between GPT-5.5 Instant (fast) and GPT-5.5 Thinking (deeper reasoning); free users are routed automatically.

How do I write better ChatGPT prompts?

Use the RCTFC structure from this guide: Role, Context, Task, Format, Constraints. Always include your situation (India, city, budget, audience), at least one number, and one thing to avoid. Then iterate — the second or third response after your feedback is usually the keeper.

Can I trust ChatGPT for GST, legal, or medical questions?

Use it to understand concepts and draft documents, never as the final authority. It states wrong tax rates, fabricated legal sections, and outdated rules with full confidence. Verify anything consequential against official sources (CBIC, India Code, your CA or doctor) before acting on it.


Related Articles

All plans, prices and feature availability verified from OpenAI’s official pricing page and Help Center in June 2026. OpenAI changes limits frequently — check chatgpt.com/pricing before subscribing. This guide consolidates and replaces our two earlier ChatGPT tutorials.

Leave a Reply

Your email address will not be published. Required fields are marked *