ChatGPT Plus Review

Reviewing ChatGPT Plus is like reviewing a city: by the time you publish, something has been torn down and something else has opened. So instead of cataloguing features that may be renamed next quarter, we judged it the way a subscriber actually experiences it — as the one tab that ends up open all day, doing a dozen unrelated jobs with no setup.

That breadth is the honest reason it remains the default first subscription. In one paid plan you get a strong general-purpose model for writing and reasoning, voice conversations that work well enough to use while driving or cooking, image generation for the occasional visual, and the ability to upload files and ask questions about them. None of these is always best-in-class. All of them are good, and they live in one place, which matters more than benchmark wins for most people.

Custom GPTs are the sleeper feature. Most subscribers ignore them; the ones who don't end up with small saved assistants for repeated jobs — a meal planner that knows your constraints, a proofreader with your style rules baked in. It is the closest thing the product has to a memory of how you like to work, and it converts the assistant from a clever toy into a tool you'd miss.

Now the flaws, because they are structural rather than cosmetic. Usage caps on the most capable models exist and have shifted repeatedly — if you lean hard on the best model during a heavy week, you can hit a ceiling mid-task. The interface and model lineup change often enough that a workflow you rely on may behave differently after an update you didn't ask for. And the privacy defaults — whether your conversations feed model training — are worth checking in settings rather than assuming; most subscribers never look.

There are also tasks where competitors are simply better. Very long documents and sustained writing projects favour rivals built around larger working memory. Research questions with citations favour Perplexity's approach. If your work is dominated by one of those two patterns, the default choice may not be your best choice — see our head-to-head guide.

One habit separates subscribers who feel the money is well spent from those who quietly cancel, and it has nothing to do with which model is selected. The people who get value give the assistant the context a competent colleague would need — what the thing is for, who reads it, what has already been tried, what the constraints are — and then push back on the first answer rather than accepting it. The people who don't ask a one-line question, receive a generic one-line-question answer, and conclude the product is overrated. Both are getting exactly what they asked for.

It is worth knowing where the failure modes live, because none of them are announced. It will produce a confident wrong answer as readily as a right one, particularly on anything numerical, recent, or specific to a jurisdiction. It will agree with a premise you smuggled into your question. And it has no reliable knowledge of what it does not know. None of this makes it unusable; it makes it a tool that needs a supervisor, and the supervision is the part that is on you.

But for the person asking where to start: start here. It covers the widest range of everyday work, the free tier lets you verify that before paying, and if you later discover a specialised need, you'll know exactly which rival to test. Most people never do, because the everything-store keeps being enough.