Goal: Guide professionals to use AI tools to support text simplification and problem detection while maintaining human oversight for accuracy, relevance, and accessibility.
Artificial intelligence (AI) is changing the way we write, edit, and make texts accessible. Tools like ChatGPT and other large language models (LLMs) can support plain language work in several ways—but they also raise important questions about reliability, bias, and quality control.
First, AI can help detect complex or unclear content. Readability tools powered by AI can scan texts and highlight long sentences, passive voice, jargon, or dense vocabulary. Some platforms suggest alternative phrasing (like U31 in French) or even rewrite entire sections. This can save time, especially in the early stages of simplification.
Second, generative AI—like ChatGPT—can rewrite a text in simpler language on request. For example, you can paste a complex paragraph and ask for a version that’s easier to understand. You can also request a summary, list of main points, or even different reading levels (e.g. B1, easy-to-read). This makes it a potentially useful tool for first drafts, idea generation, or comparative rewriting.
However, we must use these tools with caution. AI does not understand the real-world needs of readers. It does not know your audience, your context, or the purpose of your text. It can introduce mistakes, oversimplify, or “hallucinate” information. For now, AI lacks the judgment and empathy needed to create truly accessible content.
It’s also important to consider bias. Most large models are trained on data that reflects dominant language patterns—this can reinforce exclusionary norms or lead to culturally inappropriate suggestions. Furthermore, LLMs don’t always follow existing plain language standards like ISO 24495-1:2023 or align with national guidelines.
That’s why human expertise remains essential. AI can be part of your plain language toolkit, but not your final editor. Always test content with real users and check AI-generated suggestions against recognised frameworks.
In short: AI can assist but not replace. It’s a tool for simplification, not a guarantee of clarity.
Reflection prompt & tip
Try giving AI a role:
“You are a plain language editor. Rewrite this message in a clearer, simpler way, following ISO 24495 principles.”
Then always review and adjust before using.