When ChatGPT launched in November 2022, it crossed one million users in five days. For context, Netflix took three and a half years to hit that number. Something about talking to a machine that actually talked back — coherently, in full sentences, about almost anything — landed differently than every AI product before it.
Two years later, the hype has settled into something more useful: a clearer picture of what ChatGPT is genuinely good for, where it regularly fails, and how it compares to the alternatives that have emerged since.
This isn't a press release. Here's the honest version.
What ChatGPT Actually Is
ChatGPT is a conversational AI built by OpenAI. It's powered by a large language model (LLM) — specifically the GPT series, now at GPT-4o and GPT-5 depending on which version you're accessing — which is trained on an enormous amount of text from the internet, books, and other sources.
What this means practically: ChatGPT has read more text than any human ever could, and it uses patterns from that reading to generate responses that sound fluent, knowledgeable, and contextually relevant.
What it doesn't mean: ChatGPT isn't thinking. It isn't reasoning the way humans do. It's producing statistically likely continuations of text based on your input. Most of the time, that's impressively useful. Sometimes, it's confidently wrong.
That tension — between the impressive and the unreliable — is what makes a balanced assessment worth reading.
What ChatGPT Is Actually Good At
Writing assistance and drafting
This is where ChatGPT earns its reputation. First drafts of emails, blog outlines, cover letters, product descriptions, social media captions — anything that requires generating coherent text quickly, ChatGPT handles well. It won't replace a skilled writer, but it removes the blank-page problem effectively.
Explaining complex topics
Ask ChatGPT to explain quantum entanglement like you're twelve years old, or break down how a mortgage amortisation works, and it usually does it well. The ability to adjust complexity and tone on demand is genuinely useful for learning and for communicating technical information to non-technical audiences.
Coding help
Developers have adopted ChatGPT heavily for writing boilerplate code, debugging, explaining error messages, and converting code between languages. It's not flawless — it can produce code that looks right but doesn't work — but for common patterns and standard problems, it saves real time.
Brainstorming and ideation
Stuck on a name for a product? Need fifteen angles for a marketing campaign? Trying to think of objections a client might raise? ChatGPT is a solid thinking partner for generative tasks. It won't have the best idea in the room, but it'll have twenty ideas quickly, and one of them is often useful.
Summarising long documents
Paste in a long article, report, or contract and ask for a summary. ChatGPT handles this reliably and saves significant time on reading tasks.
Where ChatGPT Regularly Falls Short
It makes things up — confidently
This is the most important limitation to understand. ChatGPT "hallucinates" — it generates plausible-sounding information that is simply false. Fake citations. Incorrect statistics. Events that didn't happen described with complete confidence.
This isn't a bug that will eventually be fixed. It's a structural characteristic of how language models work. They generate likely-sounding text; they don't verify facts. For casual use this is manageable. For anything consequential — legal, medical, financial, academic — it's a serious problem that requires verification of every claim.
Lawyers have submitted court filings with ChatGPT-generated case citations that didn't exist. That's not a hypothetical risk.
It has a knowledge cutoff
ChatGPT's training data has a cutoff date. The free version runs on older models; even the paid versions aren't real-time. Ask about something that happened recently and it either doesn't know or, worse, generates something plausible that isn't accurate. The web-browsing feature in ChatGPT Plus helps with this, but it's not seamless.
It struggles with genuine reasoning
For straightforward logic and math, ChatGPT is fine. For multi-step reasoning problems, nuanced ethical questions, or tasks that require sustained logical consistency across a long conversation, it degrades. On standardised reasoning evaluations, models of this scale make errors in more than 40% of cases, reflecting reliance on statistical patterns rather than causal reasoning.
The responses can be generic
Ask ChatGPT for a blog post or a business plan without detailed prompting and what you get is technically correct and completely forgettable. The outputs are statistically average — which means they read like the average of everything ever written on a topic. Useful as a starting point, rarely good enough as an endpoint.
ChatGPT Pricing: What You Actually Get
| Plan | Cost | What's included |
|---|---|---|
| Free | $0 | GPT-4o (limited), basic features |
| ChatGPT Go | $8/month | Everyday chat, image generation |
| ChatGPT Plus | $20/month | Full GPT-4o, file uploads, web browsing, image generation |
| ChatGPT Pro | $200/month | Highest-tier models, extended limits |
| ChatGPT Enterprise | Custom pricing | Business features, privacy controls, admin tools |
For professional workflows, ChatGPT Plus tends to pay for itself quickly among people who rely on it for work. For casual use, the free tier is genuinely capable — more so than it was a year ago.
How ChatGPT Compares to the Alternatives
Since ChatGPT's launch in late 2022, tech giants like Microsoft and Google unveiled their own chatbots — and the market has matured considerably. Here's how the main players sit today.
Claude (Anthropic)
Claude was launched in March 2023 and quickly became known for its human-like interaction. Many praise Claude for its more eloquent, less robotic answers to queries.
Where Claude stands out is in writing quality and long-document handling. Claude's 200K context window handles entire manuscripts, and projects keep your style guide and reference materials persistent across conversations. If your primary use case is writing, editing, or analysing long documents, Claude is worth trying alongside ChatGPT.
Claude also tends to be more careful about what it says — it's more likely to flag uncertainty rather than confidently state something wrong. For some users that's a feature; for others who want direct answers, it can feel like hedging.
Google Gemini
Gemini's main advantage isn't being the best standalone chatbot — it's about being embedded where your work already lives. If your team runs on Gmail, Google Docs, Google Sheets, and Google Drive, Gemini integrates directly into those tools without any copy-pasting or context-switching.
For personal use or teams already in the Google ecosystem, Gemini is a natural fit. For raw conversational capability compared to ChatGPT or Claude, the gap is smaller than it used to be.
Microsoft Copilot
Copilot is a strong fit for organizations that want AI embedded in Microsoft 365 workflows. It runs on OpenAI's models (so it's essentially ChatGPT with Microsoft integration) and is most valuable if your daily work lives in Word, Excel, PowerPoint, and Teams.
For individual users without a Microsoft 365 subscription, Copilot offers less differentiation.
Perplexity
Perplexity sits in a different category — it's primarily a research and search tool rather than a general assistant. Perplexity still has an edge for time-sensitive topics like current events. Every response includes citations to sources, which addresses the hallucination problem partially. For fact-finding and research tasks, it's worth having alongside a general-purpose chatbot.
Quick Comparison
| Tool | Best for | Free tier | Starting paid price |
|---|---|---|---|
| ChatGPT | Versatility, coding, general use | Yes (capable) | $8/month |
| Claude | Writing, long documents, nuanced analysis | Yes | $20/month |
| Gemini | Google Workspace users | Yes | $20/month |
| Copilot | Microsoft 365 users | Yes | Included in M365 |
| Perplexity | Research, current events, cited answers | Yes | $20/month |
Most power users in 2026 don't rely on a single AI tool. A practical setup for most people: use ChatGPT or Claude as your primary assistant, and Perplexity when you need verified, sourced answers on current topics.
So Is ChatGPT Good or Bad?
It's the wrong question, but it's an understandable one.
ChatGPT is a powerful tool with real limitations. It's not good or bad in itself — it's useful or harmful depending on how it's applied and whether the person using it understands what it can and can't do.
It's clearly useful when:
- You're drafting, brainstorming, or summarising and you'll review the output
- You're asking it to explain something you'll verify or already understand
- You're using it as a starting point, not an ending point
- You need code written for common tasks you'll test yourself It creates real problems when:
- People treat its outputs as facts without verification
- It's used for high-stakes decisions in medicine, law, or finance without expert review
- The confidence of the output is mistaken for accuracy
- It's used to generate content at scale without human judgment in the loop The technology is genuinely impressive. The risks are also genuine. Both things are true, and understanding both is what makes the difference between using it well and using it badly.
The Practical Takeaway
If you haven't used ChatGPT yet, try the free version. Spend an hour with it on tasks you actually do — writing, research, coding, problem-solving. Form your own opinion based on your own use cases rather than the breathless coverage in either direction.
If you're already using it, try one of the alternatives. Claude handles writing and long documents particularly well. Perplexity is excellent for research. The best tool depends on the task, and knowing your options makes you a better user of all of them.
AI isn't going away. Getting fluent with these tools — including their limitations — is increasingly a practical skill rather than a niche interest.
Have questions about a specific use case or want to know which tool fits your workflow? Feel free to reach out.