ChatGPT Review (2026): Still the Best All-Round AI Assistant?

ChatGPT is the product that turned large language models from a research curiosity into a daily tool for hundreds of millions of people. Four years on, it faces serious competition from Claude, Gemini and a crowd of specialist tools. We spent several weeks using it as our primary assistant for writing, research, coding and everyday tasks to see whether it still deserves the default spot.

What ChatGPT is, in one paragraph

ChatGPT is OpenAI’s conversational assistant. You type (or speak) a request and it responds with text, code, images, charts or spoken audio. It can search the web, read uploaded files, run Python for data analysis, remember facts about you across conversations, run scheduled tasks, and be extended with custom GPTs and connectors to apps such as Gmail, Google Calendar, Notion and SharePoint. Since October 2026 it runs on GPT-6, and answers can include “Intelligent UI”: interactive charts, buttons and forms rather than plain text (OpenAI’s GPT-6 announcement). It is available on the web, on iOS and Android, and through a desktop app for macOS and Windows that combines Chat, the agentic Work mode and the Codex coding agent.

Pricing at a glance

Plan	Price (US, at time of writing)	Model in chat	Who it is for
Free	$0 (may show ads)	GPT-6 Luna	Casual use, trying it out
Go	$8/month (may show ads)	GPT-6 Luna	Regular users who want more uploads, images and voice on a budget
Plus	$20/month	GPT-6 Sol	Daily users who want the stronger model, deep research and no ads
Pro	$100, $200 or $500/month	GPT-6 Sol + GPT-6 Pro	Power users who live in ChatGPT and Codex all day
Business (formerly Team)	$25/seat/month, or $20 billed annually; Premium seats $125 / $100	GPT-6 Sol + GPT-6 Pro	Teams of 2-200 that need shared workspaces and data controls
Enterprise / Edu	Contact sales	GPT-6 Sol + GPT-6 Pro	Larger organisations with compliance needs

Sources: ChatGPT pricing, Pro tiers, Business pricing.

The free tier is genuinely useful: everyday text chat is not capped by a published message count, though uploads, image generation, voice, data analysis and deep research have their own limits that OpenAI does not publish and adjusts over time. The catch since early 2026 is that Free and Go may show ads. Plus is still the sensible paid tier for most people: it switches you to the stronger GPT-6 Sol model, removes ads, and adds projects, scheduled tasks and more deep research. Pro is now three tiers rather than one, which only makes sense if you regularly hit Plus limits.

Writing: excellent, with a recognisable voice

For emails, outlines, blog drafts and rewrites, ChatGPT is fast and reliable. It follows structural instructions well (“three sections, each with a one-sentence takeaway”) and handles tone shifts between formal, casual and playful without much coaxing.

The main weakness is a house style. Left to its own devices, ChatGPT drifts toward tidy lists, balanced “on the one hand, on the other” paragraphs and a slightly over-polished tone. You can fix this with a short style instruction in custom instructions or a project, and we recommend doing so before judging its writing. See our guide on how to write better prompts for the exact instructions we use.

Research: strong when browsing is on, risky when it is off

With web browsing enabled, ChatGPT answers current-events and product questions with citations, and its deeper research mode produces long, sourced reports that are a good starting point for any brief. Without browsing, it answers from training data and will sometimes state outdated or invented facts with full confidence.

Our rule: for anything with a date, a price or a statistic, turn browsing on and click the sources. For pure research work, we found Perplexity a more focused tool, though ChatGPT is close behind and does far more besides.

Coding: great explainer, decent builder

ChatGPT explains unfamiliar code clearly, finds bugs in pasted snippets quickly and writes solid self-contained functions, scripts and SQL. The built-in Python sandbox is excellent for data cleaning and quick charts from a CSV.

Where it falls short is sustained work on a real codebase. Copying files in and out of a chat gets old fast, and it loses track of project structure in long sessions. If you write code for a living, an editor-integrated assistant is a better fit. Our getting-started guide to AI coding assistants walks through the options.

Voice, vision and images

Voice mode is the feature that most surprises people who have not tried it. Conversation is natural, interruptions work, and it is useful for brainstorming while walking or practising a language. Vision is reliable for reading screenshots, whiteboards and documents. Image generation is good for diagrams, mockups and illustrations that include text, which older image models struggled with. Video is the gap: OpenAI shut down the standalone Sora app in April 2026 and ChatGPT does not generate video, so if that matters, look at Google’s Gemini and its Veo models.

Memory, projects and custom GPTs

Memory lets ChatGPT remember your preferences and context across chats. It is convenient, and you can review or delete what it has stored. Projects group chats, files and instructions for a single piece of work, which is the feature we use most. Custom GPTs let you package instructions and files into a reusable assistant; the store is enormous, though quality varies.

Reliability and limits

Response speed is consistently good. The irritation on Plus is hitting usage caps on the strongest model partway through a working session; OpenAI does not publish exact numbers, and heavy users will still notice them. The other irritation is churn. In the past few months OpenAI has added a Go tier, split Pro into three price points, renamed Team to Business, replaced the desktop app, retired its Atlas browser and moved every chat tier to GPT-6. GPT-5.5 leaves ChatGPT on 14 October 2026. None of that is bad in itself, but it makes ChatGPT harder to keep track of than its rivals.

Privacy

You can opt out of training on consumer plans by switching off “Improve the model for everyone” in Data controls, and Temporary Chats are not used to improve models. Business, Enterprise and Edu exclude data from training by default and add admin controls (OpenAI data controls). For regulated work, read the data-processing terms rather than relying on a review.

Who should use ChatGPT

• Most people who want one assistant that handles writing, questions, images and voice. It is the safest default.
• Students and learners, thanks to the strong free tier and voice mode.
• Small teams who want shared projects and connectors without a complex rollout. Business starts at two seats.

Who should look elsewhere

• Writers who want a more natural prose style by default may prefer Claude. Our ChatGPT vs Claude comparison goes into detail.
• Professional developers should pair it with, or replace it by, an editor-integrated coding assistant.
• Researchers who need citations on every claim may prefer a dedicated answer engine.
• Google Workspace users may get more from Gemini, which lives inside Gmail, Docs and Drive.

Verdict

ChatGPT earns 4.5 out of 5. Breadth is its superpower: no other assistant combines text, voice, vision, images, data analysis, browsing, agents and an app ecosystem this smoothly, and GPT-6 with Intelligent UI makes answers more useful than a wall of text. Competitors beat it in individual categories, and its default writing voice needs taming, but if you can only pay for one AI subscription, this is still the one to pick.