ChatGPT is the artificial intelligence chatbot that started the whole generative AI revolution. It’s a multimodal tool that can perform a range of functions, from answering complex questions — and simplifying the answers for you, as if you’re 5 years old — to coding software, creating images and videos and even engaging in casual conversations. 

Launched in November 2022 by OpenAI, the company co-founded by Sam Altman, ChatGPT quickly became the most popular AI application ever released, and the one everyone tried to emulate and one-up. Since then, all the big tech players have joined the generative AI race. Perplexity, Meta AI, Google’s Gemini, Microsoft’s Copilot and Anthropic’s Claude can all do similar things. But ChatGPT’s popularity, open-ended capabilities and early start gave it a strong lead.

ChatGPT became one of the fastest-growing apps in history, amassing hundreds of millions of users. ChatGPT often stands out because of its versatility and ease of use but mostly from the sheer scale of its user base.  

Let’s dive deeper into how ChatGPT works, how to use it and why it has become the talk of the town in the past few years — and some of its more controversial points.

How does ChatGPT work?

ChatGPT is built on a transformer architecture, specifically the GPT (generative pretrained transformer) family of models, ergo the name ChatGPT. It was trained on massive amounts of data from books and the internet — websites, Wikipedia, Reddit threads, news sites and much, much more.

(Disclosure: Ziff Davis, CNET’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)

Like other AI chatbots, it uses deep learning algorithms to understand context and predicts the most likely next word in a sentence based on patterns it’s seen before. That’s one reason why it sometimes “hallucinates” or generates confident-sounding but incorrect information. 

ChatGPT relies on large language models (LLMs), enabling it to understand your questions and provide human-like responses. Sometimes it does it so convincingly and naturally, you might forget it’s not actually thinking. And this boundary is being pushed further all the time. Its latest model, GPT-4.5, currently available to Plus and above tiers in a research preview, shows greater emotional intelligence and supposedly hallucinates less.

Equal parts impressive and unsettling, several LLMs have recently passed the Turing Test, where a human evaluator talks to a machine and a human, and if the evaluator cannot distinguish between them, the machine passes. GPT-4.5 was judged to be the human 73% of the time. 

Afraz Jaffri, a senior director analyst at Gartner, argues these results don’t matter as much, though.

“You can create systems that can fool people into thinking that this thing that you’re talking to is human,” Jaffri tells CNET. “It understands what they’re saying but actually doesn’t understand what it’s saying. So it’s really a test of human gullibility rather than a test of intelligence.”

When I asked Jaffri what he thinks is ChatGPT’s biggest flaw, he said, “They’re just unreliable. So even if I ask something and I get a response, I still want to check that that is actually true, which kind of defeats the point of us getting it in the first place.” 

However, OpenAI has been quick to act on user feedback. Just days ago, it announced on X a new feature that improves citations by highlighting the relevant text and showing which part each citation refers to. 

It also rolled GPT-4o back to its prior version after users on X complained that its “sycophantic” praise felt unsettling and distressing. Sycophancy is the fake flattery of someone in power, often used to gain something from them. OpenAI says it’s working on a more balanced approach going forward.

Accessibility and integrations

You can easily access ChatGPT through a web browser or via its apps on Android and iOS devices. 

It has also been integrated into educational apps, customer service bots and even Apple products via Apple Intelligence and Siri. 

Additionally, OpenAI introduced the ability to interact via phone calls through the 1-800-CHATGPT service, simplifying use for those who prefer voice interaction over typing.

What can you do with ChatGPT?

There are many use cases for ChatGPT — to name a few, it can plan your vacations, practice job interview questions with you, help you with your budgeting, write your resume, do your meal planning, draft emails, summarize long documents, create business strategies — or just have a conversation with you. It’ll learn from your inquiries and requests, and tailor itself to what it thinks you like.

If you don’t want it to remember you, you can also use a temporary chat by selecting it in the top right corner. It won’t appear in history, update its memory or be used to train models, although it may be kept for up to 30 days for safety.

ChatGPT Canvas can help you write and code, and with the Tasks feature, you can set up reminders and schedule tasks. It can draft marketing copy and even assist you with self-diagnosis, although you should always be cautious of the results and consult your doctor.  

It’s also one of the most customizable chatbots. You can create your own GPT tailored to specific tasks or personalities, or select among the thousands of user-created ones. They are organized into categories like education, productivity, trending and programming, etc.

Beyond text, ChatGPT can generate original images and videos through its integrated tools. GPT-4o’s image generator is now the default in ChatGPT for all users, offering you the ability to generate highly realistic images from simple text prompts. If you prefer to keep using Dall-E 3 (its first image generator), you can still access it through the Dall-E GPT. 

All ChatGPT users on the web and iOS/Android now have a Library where images are automatically saved, letting you revisit, reuse or edit them without digging through old chats. You can access it from the left-hand sidebar. 

Additionally, OpenAI has introduced Sora, its text-to-video-generation model, to Plus and Pro plans, expanding ChatGPT’s multimedia capabilities.

Different models for different functions

Originally based on OpenAI’s GPT-3 architecture, ChatGPT has evolved. There are different model variations and limitations based on your subscription and your purpose for using ChatGPT.

GPT-4o is the most advanced version accessible to everyone, handling text, images and audio as input and output. GPT-4o Mini, the fastest model, is optimized for everyday tasks and available to all users (free-tier users are automatically switched to it when hitting use limits on GPT-4o). 

Models o3 and o4-mini are reasoning models, where o3 is especially strong in coding, math and visual tasks, while o4-mini is a lighter and cost-efficient version. o4-mini-high is best for coding and visual perception. 

Models o1 and o1-mini are designed to “think” longer before responding and are ideal for solving complex problems. Last, as mentioned earlier, GPT-4.5 is the largest and best model for chat and it’s available in research preview for all paid and ChatGPT Edu plans for students.

If you are on any paid plan, you can change the model before or during a chat by clicking the model selector in the top-left corner of the screen. 

ChatGPT pricing

OpenAI provides ChatGPT on a freemium model. The free version offers basic access to GPT-4o Mini and limited access to GPT-4o, o4-mini and deep research. Upgrading to a paid plan unlocks additional features. 

ChatGPT Plus costs $20 a month and provides access to Sora, deep research and multiple reasoning models (o3, o4-mini and o4-mini-high), faster response times, limited access to GPT-4.5 and additional features such as voice interaction. 

For teams, there’s ChatGPT Team starting at $25 per user per month. For heavy users, ChatGPT Pro is available for $200 a month with the highest limits and capabilities. Businesses can also choose ChatGPT Enterprise with custom pricing for enterprise-level features.

Students and nonprofits occasionally get special offers. For example, ChatGPT Plus access is often free for students in the US and Canada during exam periods.

Privacy concerns, legal battles and other controversies

ChatGPT faces ongoing legal challenges related to industry concerns regarding the ethics and the legality of data sourcing for AI training. High-profile lawsuits include those from major publishers like The New York Times, Dow Jones and CNET’s parent company Ziff Davis (see disclosure above), which allege unauthorized use of their content for training ChatGPT, as do many individual authors.

Let’s not forget when OpenAI’s “Sky” voice model mimicked Scarlett Johansson without permission, prompting her team to threaten legal action over the unauthorized use of her voice. OpenAI removed the voice, said it came from a different actress and promised clearer AI-likeness disclosures.

Privacy is another concern. Experts and privacy advocates have raised ongoing questions about data protection, how personal information is stored and used, and what users should or shouldn’t share. Even OpenAI advises against entering sensitive information like credit card numbers, medical history or Social Security numbers. 

Jaffri emphasized that malicious or careless deployment, not the models themselves, is the root of most risks. 

“Now anyone can create these things and you can spread them a lot easier and a lot quicker, so that’s the kind of worry,” he told CNET. “It’s more about how it’s used rather than it itself being something to worry about.”

OpenAI has attempted to address these issues with enhanced data protection measures. Your prompts (on Free, Plus and Pro plans) may be used to train the model unless you opt out in settings. Team and Enterprise plans are designed not to collect user data by default. OpenAI did not respond to a request for comment on potential copyright and plagiarism concerns about training data or on user privacy and data protection.

Recent developments and new features

Since its beginning, ChatGPT has grown in features and capabilities. OpenAI expanded ChatGPT’s memory feature, allowing the chatbot to recall previous interactions (which you can manage or delete), creating a more personalized user experience. 

Recently, OpenAI made ChatGPT’s image generation capabilities free for all users, albeit with a limited number of monthly uses. Furthermore, in response to competition from rivals like Chinese AI company DeepSeek R1, OpenAI rolled out trials for its newest lightweight model and deep research, accessible even to free-tier users (with limited access).

ChatGPT is becoming a default interface for how people interact with AI and even how they get information from the internet. Through its ChatGPT Search feature, it is competing more directly with Google search (even removing mandatory login requirements to make it that much easier for casual users). 

ChatGPT remains an influential player in AI. As generative AI tools become more embedded in daily life, ongoing enhancements in usability, privacy protection and content generation capabilities will determine ChatGPT’s long-term success. 

“The whole kind of technology itself is going to be transformative for a lot of different areas, which will have a direct impact on our lives,” Jaffri says. “Hopefully for the better.”