What Is ChatGPT? A Plain-Language Guide

ChatGPT is an artificial intelligence chatbot built by OpenAI that answers questions and produces text, code, and images from plain-language instructions. You type what you want the way you would ask a person, and it replies in conversation. It launched on November 30, 2022, reached 100 million users within two months, and has since become the most widely used AI product in the world.

What Does ChatGPT Actually Do?

In short, ChatGPT turns instructions into finished work. Ask it a question and it explains. Give it a messy paragraph and it rewrites it cleanly. Paste a contract and it summarizes the key terms. Describe a spreadsheet formula in words and it writes the formula. The same chat window handles writing, editing, translation, brainstorming, math, coding, image generation, and analysis of files you upload.

The most common everyday uses fall into a few buckets:

• Writing and editing: emails, reports, job postings, social posts, and rewrites of anything in a different tone or length
• Summarizing: long documents, meeting transcripts, research papers, and PDFs condensed to the points that matter
• Answering and explaining: questions on nearly any topic, with newer versions searching the web and citing sources
• Coding: writing, debugging, and explaining code in most programming languages
• Analysis: uploaded spreadsheets and data files turned into summaries, charts, and answers
• Images: generating pictures and diagrams from text descriptions

How Does ChatGPT Work?

ChatGPT is powered by a large language model, or LLM, from OpenAI's GPT family (the letters stand for Generative Pre-trained Transformer). The model was trained on an enormous amount of text, and what it learned from that training is a deep statistical sense of how language fits together. When you send a message, the model generates its reply one small chunk at a time, repeatedly predicting what should come next given everything written so far.

Two extra steps turn that raw prediction engine into a useful assistant. First, human trainers rated and corrected model outputs during development, teaching it to be helpful and to refuse harmful requests (a process called reinforcement learning from human feedback). Second, modern versions of ChatGPT can use tools: they search the web for current information, run code to do real math, and read files you upload, which grounds answers in actual data rather than memory alone.

Understanding this design explains both the magic and the failure modes. The model is genuinely excellent with language, structure, and patterns. But because it predicts rather than looks up, it can also produce confident statements that are simply wrong, which is why important answers should be verified.

How Much Does ChatGPT Cost?

ChatGPT is free for casual use, and paid tiers add capability. The free tier includes the smaller GPT models with daily limits and no credit card required. Paid plans, billed in USD, are Go at US$8 per month, Plus at US$20 (about $27 CAD), Pro at US$200, and Business at US$25 to US$30 per user per month, with custom Enterprise pricing above that. Canadian banks convert the USD charge and typically add a 2.5 to 3 percent foreign transaction fee.

For a full breakdown of every plan in Canadian dollars, including the hidden fees and the point where the API becomes cheaper than subscriptions, see our ChatGPT pricing guide. Students should also read the ChatGPT for students guide, since the free tier covers most student use cases.

What Are ChatGPT's Limitations?

No, ChatGPT is not always right, and that is its most important limitation. It can hallucinate: state false information with full confidence, invent citations, or misremember details. Web search and source citations in newer versions reduce this substantially, but the risk never reaches zero.

• Accuracy: verify anything consequential against a primary source, especially legal, medical, financial, and tax answers
• Privacy: on consumer tiers, conversations may be used to improve the models unless you opt out; business tiers exclude training by default
• Recency: without web search enabled, the model's built-in knowledge has a cutoff date and misses recent events
• Judgment: it optimizes for a plausible answer, not for telling you that your question is the wrong question

How Is ChatGPT Different From Claude, Gemini, and Copilot?

ChatGPT's main competitors are Claude (from Anthropic), Gemini (from Google), and Microsoft Copilot, which embeds OpenAI's technology inside Office apps. All four do the same core job with different strengths: Claude is favoured for long documents and careful writing, Gemini integrates tightly with Google Workspace, and Copilot lives inside Word, Excel, and Teams. Individual paid plans for all of them land near the same US$20 per month price point.

If you are choosing between them, our AI tools pricing comparison puts every plan side by side in CAD, and the ChatGPT vs Claude vs Gemini comparison covers capability differences in depth.

How Do Businesses Use ChatGPT?

Businesses generally start with individual seats for drafting and research, then graduate to automated workflows where the real savings live: AI that triages a shared inbox, drafts quotes from job notes, processes invoices, or answers routine customer questions around the clock. That second step usually involves the API rather than the chat interface, connected to existing systems. Our AI consulting practice builds exactly these workflows, typically starting with a single high-payback process.

Frequently Asked Questions