More people use chatbots for writing, but fact checks are still needed

In short: AI chatbots are widely used as writing helpers, but they do not “know everything” and they can make up false details, so humans still need to check their work.

What's going on

Many people talk about AI chatbots as if they contain the sum of all human knowledge. In reality, tools like ChatGPT and Claude learn from large but limited collections of text, mostly from the internet and licensed sources, up to a cutoff date. Without added tools like web search, they do not automatically know what happened today.

That helps explain a common problem called “hallucinations,” which is when a chatbot confidently states something that is not true. Think of it like a helpful intern who writes quickly, but sometimes fills in missing information with guesses. That is why writers are told to verify quotes, numbers, names, and claims.

Even with those limits, chatbots can be strong writing assistants. ChatGPT is often used for research-heavy drafts, outlines, and structured writing, and many users pay about $20 a month for the Plus plan. Claude is known for natural sounding prose and for handling very long inputs, which can help with long documents and careful tone.

Studies suggest the tradeoff is real. One 2023 study found ChatGPT cut writing time by about 40% and improved quality by about 18%, especially for less experienced writers. But 2024 research found AI-assisted stories were rated higher individually while becoming less original overall, unless humans added their own ideas and edits.

What to watch

Expect more “AI writing” products that wrap these chatbots with extra features, like templates for marketing and search friendly pages. For most people, the key question will be whether the tool saves time without encouraging sloppy facts, bland writing, or both. Tools such as Gemini also matter here because tight links to other services can make it easier to pull in current information, but they still need careful checking.

Source: NYTimes