ChatGPT vs Claude

They are not the same product any more

ChatGPT is an assistant platform. Images, voice, browsing, custom GPTs, memory, apps, a free tier that now carries advertising. Claude is a work surface: long documents, careful reasoning, code, and very little else competing for the screen.

If your day is drafting, editing, analysing documents or writing software, Claude's output quality is the better buy and it is not close on long-form prose. If your day is a bit of everything and you want the assistant to also handle images, search and voice, ChatGPT is the better buy and Claude will feel bare.

The limits are the real difference

Anthropic added weekly rate limits on top of its rolling five-hour windows in August 2025, with extra usage sold at API rates. Heavy users cannot treat the subscription as unlimited any more, and Anthropic has tightened limits repeatedly, including testing whether Claude Code belongs on the entry tier at all. That is why Claude's scores are split so violently: 4.6 on G2 from people describing the output, 1.5 on Trustpilot from people describing the wall they hit at 4pm.

ChatGPT's limits move too, with demand, and OpenAI retires models paying customers depend on. The GPT-4o removal was the clearest example: people had built habits, prompts and workflows on a specific model and lost it. You are renting capability from both companies, not buying it.

Things that should affect the decision and usually do not

• •OpenAI began testing advertising inside ChatGPT for free and Go tier users in February 2026 and has since rolled it out more widely. The paid tier is not ad-supported, but the direction of travel is worth noticing.
• •In the consolidated copyright litigation in the Southern District of New York, the court affirmed an order in January 2026 compelling OpenAI to hand a twenty-million-conversation sample of ChatGPT logs to plaintiffs. Ordinary chats are inside that kind of sample.
• •Anthropic settled the authors' class action Bartz v. Anthropic for $1.5 billion over training on pirated books, which is the largest number of its kind so far and did not change the product at all.

Neither company comes out of the last two years looking careful. Pick on fit, and assume your conversations are not private in the way you would like.

How we would buy it

• •Writing, editing, long documents, code: Claude.
• •General assistant, images, voice, search, sharing with a household: ChatGPT.
• •Both at once, which plenty of people do: pay for one and use the other's free tier for a month before deciding the second $20 is real.
• •Hitting limits constantly on either: you are the customer both companies are trying to move onto usage billing. Model what the API would cost you before buying a higher tier.