Did you know that an experiment measured the effect of a chatbot on professional writing tasks? The researchers studied four hundred and fifty-three graduates. The results published in two thousand twenty-three relate to specific exercises, not to the quality of books or authors. Participants write professional texts with or without access to the tool. Our office illustrates this work: a proposal can be proofread, corrected and retained. The fictitious words of an interface must not become a scientific citation or a document produced in that experiment.
On average, the completion time decreases by about forty percent and the quality assessed increases by about eighteen percent. These are two distinct measures in these tasks. They do not mean that a person becomes a writer, or that any generated text is better or accurate.
What matters is the protocol: tasks, a comparison group and an evaluation. The tool can help formulate, but the author still has to check the meaning and facts. The observed performance does not dispense with rereading a sentence, when it seems well written.
What we know
The experience covers 453 graduates and some professional writing tasks. Time and quality averages do not automatically apply to all texts.
To go to the source
References consulted on October 12, 2026.
Signaler une erreur ou apporter une précision
