Plain language: what it is and why AI makes it more necessary

Written by
Josep Maria SanchónDirector de contenidos
Published
8 min read
Tags
Narrative and content

Understanding a diagnosis, a court ruling or an official procedure shouldn't depend on how it is written. Why plain language has become a right and what changes now that so much text is written by AI.

Ilustración en acuarela de dos manos que recogen agua transparente de un arroyo; a través del agua se ven la palma y unos guijarros de colores.
Clear water lets you see every last stone on the bottom; so does a good text (DB/AI).

For years, writing well was hard work. You needed to know your grammar, have the craft and put in the time. Today any AI tool will give you, in seconds, a text with no mistakes, properly punctuated and in a pleasant tone. What it won’t give you is the decision about what needs saying, to whom and why. And that, which was always the hard part, is exactly what defines plain language.

What plain language is

There is a definition shared by the International Plain Language Federation and the ISO standard that regulates it: a communication is in plain language if its wording, structure and design are so clear that the intended readers can easily find what they need, understand what they find and use that information.

Look at the order of the verbs: find, understand, use. The definition says nothing about writing beautifully or simplifying to the point of treating readers like children. It talks about what happens to that person after they read. A clear text is one that lets them do what they came to do: book an appointment, understand a ruling, find out whether they’re entitled to a grant. And sometimes that takes more words than it seemed.

Plain language started as a concern of the courts, which saw that people didn’t understand what was being communicated to them. In Spain it is now law. Since December 2024, the Organic Law on the Right of Defence requires, in its Article 9, procedural acts and communications to be written in plain language, in a simple and universally accessible way. And it extends the requirement to court rulings, which must be understandable bearing in mind who they are addressed to.

Outside the courts, the institution that has crystallised the movement is the Royal Spanish Academy (RAE). Together with the Association of Spanish Language Academies and the Ibero-American General Secretariat, it published the Pan-Hispanic Guide to Plain and Accessible Language, which takes plain language into public administration, business, consumer affairs and healthcare. The Pan-Hispanic Network it coordinates brings together more than 400 institutions, and private companies have already joined, such as Mapfre.

There is also a technical standard. In 2024 the Spanish standards body UNE published the Spanish version of ISO 24495-1, with the principles of plain language, and in April this year the second part, devoted to legal communication. AENOR certifies texts and methodologies with its own seal, and companies such as Iberdrola have already obtained it.

In the English-speaking world, none of this is new. In the United Kingdom, the Plain English Campaign has been fighting official gobbledygook since 1979, and its Crystal Mark, created in 1990, now appears on more than 20,000 documents worldwide. In the United States, the Plain Writing Act, signed on 13 October 2010, requires federal agencies to write documents for the public in language that is «clear, concise, well-organized». And in Brussels, the European Commission has committed to plain language in its own texts, with training for staff and annual Clear Writing Awards.

What a few years ago was good practice among a few has become a right, a standard and a quality criterion that can be audited.

AI at the heart of the debate

What has changed in recent months is the focus. The 3rd Convention of the Pan-Hispanic Network, held in La Plata (Argentina) on 1 and 2 October, was titled «Ibero-American Dialogue on Plain and Accessible Language and Artificial Intelligence». It brought together more than 1,200 attendees and 275 papers, and ended with the signing of the La Plata Declaration, which creates an Ibero-American plain language network in Spanish and Portuguese.

At the preparatory sessions, held in Madrid in May, the director of the RAE, Santiago Muñoz Machado, argued that AI should use plain language because «soon, most communications will take place through AI». And at the opening in La Plata he explained that, although the subject is not new, it has returned over the last decade because of the revolution brought by new technologies and artificial intelligence.

It makes sense that the concern is growing right now. Language models write with flawless correctness. The problem lies elsewhere.

Correct language is free, clear language still costs

There is data that helps us understand what kind of text AI produces. A team of researchers analysed more than 15 million abstracts of biomedical papers published on PubMed and estimated that at least 13.5% of those from 2024 had been processed with language models. They detected this from the trail left behind: style words, which give a text tone but add no content. The study has limits worth pointing out, as it covers academic, biomedical prose in English. But it describes a pattern anyone will recognise: flawless sentences that sound like something and say very little.

Google, in principle, simply judges what a text offers the person reading it, whatever tool it was written with. It does warn, though, that generating pages in bulk with no value for readers may fall under its scaled content abuse policy.

And here is what matters. AI has made correct text so cheap it is almost free. Clear text costs the same as before, because grammar was never the expensive part. The expensive part is deciding what you want the reader to understand, what they need to be able to do afterwards and what can go. That decision is editorial and the model doesn’t make it for you. If you ask AI to «improve» a text that doesn’t know who it is for or what its purpose is, it will give you back a better-written text that doesn’t know either.

On a website it shows sooner

On the web, a lack of clarity costs you faster. Nielsen Norman Group’s classic studies measured it in the digital stone age: in 1997, 79% of participants scanned pages and only 16% read word by word. By 2008 they had calculated that, on an average visit, users read at most 28% of the words, and 20% is more realistic. These are old studies, but nothing we see today suggests we read any more slowly.

What’s more, the reader of a website is no longer just a person. We explained this in another Apuntes article, «El clic no ha muerto»: every text is read at once by three readers, the person, the search engine and the model that summarises the answer. I haven’t been able to find any solid study showing that generative engines cite clear texts more often, so I won’t claim it. But the reasoning is simple: a text that says what each thing is, who it is for and under what conditions is easier to understand for any of the three readers and audiences. A vague, unclear text, on the other hand, forces all three to guess. The human will stop reading. AI won’t take you into account. Nor will the search engine.

At Digital Bakers we work with some sectors where an unclear text has consequences. We polish texts, give them lustre and try to make the tone and personality of each corporate voice come through in plain texts. Which, we insist, doesn’t mean they should be impoverished or full of street slang. That’s not it. Plain language should achieve something that is very easy to explain but takes work and common sense: treating people with intelligence and respect. Making texts useful to people and serving their needs and searches.

Plain language should achieve something that is very easy to explain but takes work and common sense: treating people with intelligence and respect.

Sources

More from Field Notes
AI is making your town and your website look like everyone else's. What are you going to do about it?Yes, we use AI. But no, it won't make your project cheaper

Shall we talk about your content?

Let's talk

At Digital Bakers this is how we write: human judgment up front and technology behind, pushing.