ChatGPT review: Two years in, it's still the one most people should use

Three years after the world lost its mind over ChatGPT, the question has shifted. It used to be "is AI any good?" Now it's "which AI?" - and that's the question I've actually been trying to answer for the last six months.

I run ChatGPT every day. I also run Claude, Gemini, Perplexity, and a rotating cast of others depending on the task. After enough head-to-head testing, here's where I've landed: ChatGPT is no longer the best AI at any single thing. It's still the AI most people should default to.

That's a weirder position than it sounds.

What ChatGPT actually does well

The thing that surprised me on a recent re-evaluation: ChatGPT's biggest strength is something OpenAI rarely talks about. It's not the model intelligence (Claude often beats it on long reasoning, and competitors match it on raw IQ). It's not the multimodality (Gemini ships images and code with comparable competence). It's the fact that everything works.

You can paste in a 200-page PDF and ask questions about page 84. You can upload a CSV and have ChatGPT write the Python to analyze it, run that Python, and show you the chart, all without leaving the chat. You can generate an image, then ask it to change one specific element. You can talk to it with your voice, watch it talk back, interrupt it mid-sentence, and have it pick up your thread without missing a beat. Custom GPTs, scheduled tasks, Projects, memory across conversations. None of these are groundbreaking individually. Combined, they create a flywheel where ChatGPT becomes the place you keep returning to.

Claude is sharper at coding and writing nuance. Perplexity is far better at finding current information with citations. Gemini integrates more tightly with Google's stack. But none of them ties depth of features together this seamlessly. When I want to do something fast and don't want to think about which tool to open, I default to ChatGPT and trust it'll work.

Where it falls short

The honest case against: GPT-5 didn't quite ship the leap people were hoping for. The reasoning models in the o-series are smart but slow, often taking 30+ seconds to answer questions Claude handles in three. The free tier feels increasingly stingy compared to what competitors offer. Google's Gemini gives away its top-tier model for free; OpenAI gates its best behind Plus ($20/mo) and Pro ($200/mo) plans.

Hallucinations haven't gone away either. ChatGPT will still invent citations, fabricate quotes, and confidently make up biographies of obscure people. It's better than two years ago, but for anything reputation-critical, you have to verify everything. That's true of all AI, but ChatGPT's confident prose makes the hallucinations especially convincing. I've watched it produce a paragraph about a court case that didn't exist, in such polished language that a non-lawyer would have shipped it.

The Plus plan is $20/month, matching Claude Pro and Perplexity Pro. Pro is $200/month and adds unlimited access to the slow reasoning models. At that price you're really asking whether it's still cheaper than hiring a human, and for 99% of users the answer is the $20 plan.

How it stacks up

Here's where I'd pick ChatGPT over the alternatives:

• Casual everyday use, fast answers, voice mode, generating images in the same chat where you're talking
• Building Custom GPTs for your team or personal workflows
• Working across files (PDFs, CSVs, code) and getting answers that reference all of them
• Multimodal projects that need image plus text plus code in one place

Here's where I'd switch tools:

• For coding, open Claude or run it inside Cursor instead
• For research that actually needs current sources, use Perplexity
• For long-form analysis where reasoning quality matters more than features, Claude
• For anything tied tightly to Google Workspace, Gemini

The bigger picture

ChatGPT has 800 million weekly users. That's a number that should sober anyone trying to "disrupt" it. The product has become competent at so many things that most users will never discover the better tools for each specific task, because ChatGPT does it well enough.

That sounds like faint praise. It's actually the entire game. The tool you reach for matters more than the tool that wins a specific benchmark. ChatGPT is the AI most people open without thinking about it, and that habit advantage compounds harder than most reviewers acknowledge.

Verdict

I'll keep paying $20 a month for ChatGPT. I'll also keep running Claude for actual writing, Perplexity for research with sources, and Cursor for code. The model that wins on raw quality changes every few months. The product you trust to handle 80% of your AI needs without you thinking about it doesn't change as often, and right now that product is still ChatGPT.

If you're new to AI tools and don't want to comparison-shop, this is the right place to start. If you've already settled into a stack, there's still a reasonable argument for keeping ChatGPT as your default and switching to specialists only when a task demands it.