Researchers from Oxford and Potsdam universities say AI writing tools are twisting the meaning of users' drafts on politically sensitive subjects, potentially reshaping public opinion over time.
Key facts
- A study from Oxford and Potsdam universities examined AI redrafting and summarising tools.
- Researchers found the tools inject political biases into messages on sensitive topics such as abortion and climate change.
- Some tools leaned distinctly rightwing while others were more liberal, according to the study.
- Experts warn small changes in drafting could spread rapidly and cause long-term shifts in public opinion.
AI tools are altering the meaning of online messages on sensitive political topics, ranging from abortion to climate change, in ways that could reshape long-term public opinion, according to researchers cited by the Guardian.
The findings come from a study conducted by Oxford and Potsdam universities, which examined how AI tools redraft and summarise the large volume of daily online messages. Tech companies have promoted these tools as convenient ways to manage the growing influx of communication.
According to the study, many of these tools inject their own political biases into users’ text. Some lean distinctly rightwing, while others are more liberal, the researchers found.
Experts involved in the research warn that even small changes introduced during the drafting process could spread rapidly across online communication. Over time, they say, these incremental shifts could snowball and create long-term changes in public opinion.
The concern centres on the scale at which AI writing assistants are now used. As these tools become embedded in everyday messaging and summarising, subtle alterations to meaning could reach large numbers of people without users being aware their words have been changed.
The Guardian reported the study’s findings on 6 July 2026, highlighting the potential consequences for public debate on some of the most contested issues, including abortion and climate change.
Why it matters
AI writing tools are increasingly used to draft and summarise everyday messages, so even subtle political bias in their output could influence how millions of people communicate. Researchers warn these small shifts could accumulate and reshape public opinion on contested issues over time.
Frequently asked questions
Which universities conducted the study?
The study was carried out by researchers at Oxford and Potsdam universities, according to the Guardian.
What kinds of bias did the study find?
The study found that some AI tools leaned distinctly rightwing while others were more liberal when redrafting or summarising messages on sensitive topics.
Why do researchers say this is a concern?
Experts warn that small changes introduced during drafting could spread rapidly and snowball into long-term shifts in public opinion.