Writing an article with ChatGPT

ChatGPT can produce an outline or a rough draft in seconds. That speed is useful, but it also makes it easy to publish writing that sounds plausible without saying anything specific—or that confidently includes a false claim.

This post was written shortly after ChatGPT appeared in 2022. The models and interface have changed since then, but the editorial problem is the same: generated text still needs human judgment.

Give it a narrow job

Start with a topic you understand and decide what the article should help the reader do. Then use specific prompts for small parts of the process rather than asking for a finished article in one step.

For an article about meditation, for example, the questions might include:

The answers can help expose gaps in an outline. They are not sources. For factual claims, especially health, legal, financial, or current information, go back to reliable primary material and cite it.

Rewrite the output, do not just tidy it

Generated drafts tend to use broad introductions, predictable headings, repeated conclusions, and safe statements that avoid a real opinion. Fixing a few words does not solve that. Rebuild the article around the useful point, remove sections that do not earn their space, and add the context only you can provide.

Do not invent personal experience to make generated writing sound human. If you did not test something, say what the claim is based on. If you have no useful opinion or example, research the subject further instead of manufacturing one.

Keep responsibility with the author

Check every claim, link, quotation, code sample, and product detail. Also check whether the article accidentally resembles source material too closely and whether private information was included in the prompt.

I would use ChatGPT to question an outline, suggest alternatives, or help get past a blank page. I would not let it decide what is true, what matters, or what I am willing to put my name on. Those are the parts that make someone the author.