Easy-to-read experts from nine countries met online to talk about artificial intelligence (AI) and easy-to-read.
“It is not the saviour and it is not the end of the world. It is a tool like any other. But you have to use it smartly,” said Judith Timoney from FUB, Sweden.
Why now
The way easy-to-read is designed is changing, and AI is the main reason. More texts presented as easy-to-read exist than ever before.
But we also regularly see documents labelled as easy-to-read that are nothing like it. No person with an intellectual disability was involved in making these documents, let alone asked whether they understood the information.
With AI, the question is whether easy-to-read documents really are made in a way that people with intellectual disabilities can understand the information, and whether people with intellectual disabilities are involved in making this information accessible.
This discussion on 20 October 2026 was a first step in our work to revises the European easy-to-read guidelines. We will work with our members and others to make these guidelines clearer, up-to-date, and relevant to current technology and communication.
How members use AI today.
During this discussion, we heard about three main ways organisations use AI in easy-to-read.
- Some use it to write the first draft of a text, as they produce more reports than they can afford to adapt by hand.
- Some use it to support validators. In Luxembourg, Klaro lets validators ask an AI what a difficult word means; validators then explain it in their own words.
- Some organisations build their own tools. In France, Unapei created Cap’FALC, a tool designed so that it cannot adapt a whole document in one click. This helps keep the person, not the machine, as the author. Klaro is also testing an app that turns a difficult letter, or a conversation with a doctor, into easy-to-read.
People do the evaluation
There are different ways of using AI to make information accessible. But everyone agreed AI should not replace people with intellectual disabilities role in making easy-to-read.
“We help ourselves with some kind of AI, but we never trust it. We always check if it is correct. And then we validate it with people with intellectual disabilities,” said Karolina Makowiecka from PSONI, Poland.
Ensuring people with intellectual disabilities are included, credited and properly paid for their work on easy-to-read and understand was also considered a priority. Easy-to-read is also about creating job opportunities for people with intellectual disabilities.
“Our costs are higher, because we are putting in the job of the people who are testing,” said Silvi Laidro from Estonian Association of Persons with Intellectual Disabilities. The higher costs are a problem when trying to win public contracts – services that don’t involve people with intellectual disabilities come cheaper.
Personal data
“We have to be very careful about sensitive data, because we work on self-determination. We do not have to expose personal ideas or medical cases about the people we are working with,” said Alice Urvoy de Portzamparc from Nous Aussi, France.
What’s next
Participants in the discussion said they would welcome guidance on the use of AI in easy-to-read. They want to know how to use AI well, not to be told to avoid it.
Another common thread was also to explain why easy-to-read is needed. Guidance on easy-to-read should describe the thinking behind it, not only the rules.
Inclusion Europe and Unapei will organise a workshop in France in 2027, to go deeper into this conversation.
By the end of the year, we will published updated guidelines for easy-to-read.
AI for Easy-to-Read is a project of Inclusion Europe with Unapei, supported by the European Disability Forum and Google.org.
