ChatGPT

ChatGPT launched November 30, 2022 as a free research preview from OpenAI, the company Sam Altman, Greg Brockman, Ilya Sutskever, and others founded in 2015. It has since grown into the default entry point most people have to generative AI, reportedly reaching 900 million weekly active users by early 2026 according to figures OpenAI has disclosed. What started as a single chat window is now a bundle: text chat, image generation, voice, a Deep Research mode, a coding agent (Codex), a browser-automation agent, and — as of October 2025 — its own Atlas web browser. ## Who it's actually for ChatGPT is the right default for someone who wants one subscription that covers writing, research, light coding, and image generation without picking a specialist tool for each. It's a weaker choice for anyone whose primary use case is a narrow, high-stakes one: developers doing serious agentic coding are better served by Claude, which independent benchmark comparisons and coding-focused reviewers consistently rank ahead of GPT-5.x on tasks like SWE-bench; teams already living in Google Workspace get more from Gemini's native integration and larger context window. ChatGPT's advantage is breadth and the size of its plugin/integration ecosystem, not being the best model at any single task. ## What's actually good about it The model lineup moves fast — by mid-2026 OpenAI had already cycled through GPT-5.1 through GPT-5.5 (each deprecated within its ChatGPT app on a rolling schedule) into the GPT-5.6 family (Sol, Terra, Luna), giving Plus and above a current flagship plus cheaper, faster variants for routine tasks. That cadence is a real advantage for staying near the frontier without switching products. The agent and tool ecosystem is the most mature of the major chatbots: Codex for coding tasks, a browser-use agent, Deep Research for multi-step web research, and now the Atlas browser, all inside one account. For business and enterprise customers, OpenAI states plainly that inputs and outputs are not used for model training by default — a real and useful default, distinct from personal Free/Plus accounts where training use is opt-out rather than opt-in. ## What's actually wrong with it Three things stand out from independent coverage and user reports, not from OpenAI's own marketing. First, OpenAI began showing ads to logged-in adult users on the Free and Go tiers in the United States starting February 2026, expanding through the year into multiple ad formats (sponsored answer cards, product placements, sidebar units). Free users can reportedly opt out of ads only by accepting fewer daily messages instead — a tradeoff, not a fix. This is a meaningful shift for a product whose free tier previously had no commercial layer at all. Second, hallucination and reliability complaints have not gone away, and by some measures have gotten more visible rather than less. G2's aggregate review data lists "Inaccuracy" as ChatGPT's single most-cited negative theme, and multiple independent write-ups this year point to newer model updates producing faster but less carefully hedged answers — a pattern users on Reddit and Hacker News describe as the assistant answering confidently rather than correctly. None of this means ChatGPT is unusually bad at accuracy compared to other frontier chatbots — every major model still hallucinates on checkable facts — but it undercuts any claim that the problem is solved, and OpenAI's own research has at points shown newer models hallucinating at rates comparable to or higher than older ones on certain benchmarks. Third, support for consumer tiers is bot-only with no clear path to a human for account or billing problems, which matters more than it sounds like it should once money is involved — Plus and Pro are real monthly charges, and users who hit a billing snag have limited recourse. Fourth, the plan lineup itself has become genuinely hard to compare. Between Free, Go ($8/month), Plus ($20/month), Pro ($200/month), Business, and Enterprise, feature and limit differences shift with enough regularity that even attentive users report losing track of what their tier includes — OpenAI's own release notes show model availability changing multiple times within a single year. ## Pricing reality As reported across multiple current pricing breakdowns (OpenAI's own pricing page could not be directly retrieved during this research pass and these figures should be re-verified against it before being treated as final): Free is $0 with ads and tight limits; Go is roughly $8/month with higher volume but still ad-supported and missing advanced features; Plus is $20/month, ad-free, with the fuller model set, Deep Research, Sora, Codex, and agent mode; Pro is $200/month for the top-tier model access and the highest Deep Research allowance. Business pricing was reportedly cut in April 2026 to around $20/seat/month billed annually or $25/month billed monthly, down from $25/$30, and now includes unlimited flagship-model use per seat. Enterprise remains custom-quoted. Given how often these numbers move, a reader making a purchasing decision should confirm current pricing directly on OpenAI's site rather than treating any of the above as fixed. ## Verdict ChatGPT earns its default status through breadth, ecosystem maturity, and an aggressive model-release cadence, not through being definitively the best chatbot at any one job — independent comparisons this year consistently place Claude ahead on coding and long-document writing, and Gemini ahead on multimodal work and context window size, while describing the gap at the top as the smallest it has been. For a generalist user who wants one tool for most things and values the surrounding ecosystem (Codex, Atlas, Deep Research, third-party integrations), it remains a reasonable default. For a reader whose need is narrow and specific — serious coding work, deep Workspace integration, or a strict no-ads free tier — a competitor is now a better first choice, and OpenAI's introduction of ads to its free product in 2026 is the clearest sign yet that "good enough for everyone, forever, for free" is no longer the plan. ## Quick take **Verdict:** ChatGPT is the broadest, most ecosystem-rich general AI assistant available, but it no longer has a clear model-quality edge over Claude or Gemini, and its free tier now carries ads. **Choose it if:** - You want one subscription that covers chat, research, light coding, and image generation instead of juggling specialist tools - You value the largest third-party integration and agent ecosystem (Codex, Atlas browser, Deep Research, plugins) - You're a business buyer who needs a documented default of no training on your data without negotiating an enterprise contract **Skip it if:** - Your primary use case is serious agentic coding, where independent benchmarks currently favor Claude - You're deep in Google Workspace and would benefit more from Gemini's native integration and larger context window - You want a genuinely ad-free free tier — OpenAI began showing ads to Free and Go users in the US in February 2026 - You need reliable human support for billing or account issues; consumer tiers are bot-only **Watch out:** Free and Go tiers now show ads (rolled out in the US starting February 2026), and the only stated way to avoid them on Free is to accept a lower daily message limit instead.