Online questionnaires are being increasingly swamped by AI-generated responses, raising concerns that a vital data source for researchers is becoming polluted. Platforms like Prolific, which pay participants to answer questions, are widely used in behavioural studies.

Researchers at the Max Planck Institute noticed suspicious patterns in their work and began investigating. They found that nearly half of the respondents copied and pasted answers, strongly suggesting that many were outsourcing tasks to AI chatbots.

Analysis showed clear giveaways, including overly verbose and distinctly non-human language. The researchers concluded that a substantial proportion of behavioural studies may already be compromised by chatbot-generated content.

In follow-up tests, they set traps to detect AI use, including invisible text instructions and restrictions on copy-paste. The measures caught a further share of participants, highlighting the scale of the challenge facing online research platforms.

Experts say the responsibility lies with both researchers and platforms. Stronger verification methods and tighter controls are needed for online behavioural research to remain credible.