Fact · Voice and AI
Even when an LLM is asked only for grammar edits, the same 2026 preprint finds, it still changes the text in a way that significantly alters its semantic meaning.
Checked against its source on October 17, 2026
In the source's own words
“even when LLMs are prompted with expert feedback and asked to only make grammar edits, they still change the text in a way that significantly alters its semantic meaning”
How LLMs Distort Our Written Language, Marwa Abdulhai et al., arXiv, 2026, Abstract.
Open the original- What it measures
- The test prompted the models with expert feedback and asked for grammar edits only, then measured the meaning of the output.
- What it does not show
- A controlled prompt setup. It shows that the change can happen, not how often a cleanup prompt will cause it.