Only 7 of 40 people spotted the fake story a news robot told them, a study finds

The same people rated the robot as credible, with mean trust scores above 5 on a 7-point scale for ability, integrity and benevolence. Higher trust went with greater willingness to accept the robot as a newscaster (Spearman's rho .41, p = .009). The authors call the mismatch a 'credibility paradox' and say designing for credibility alone may not be enough.

A 10-15 minute session also did not change what people feared or valued about robot news presenters. The authors say longer-term studies are needed to see how views shift over sustained interaction.

Each participant completed a pre-interaction questionnaire, heard three stories from Roboldo in randomised order with 10 seconds between them, then completed a post-interaction questionnaire that added the question of whether any story was false. Sessions lasted 10 to 15 minutes, all in Brazilian Portuguese, and trust was rated on a 16-item, 7-point scale.

Trust scores were high from the start (pre-interaction ability mean 5.15, integrity 5.07, benevolence 5.02). Among acceptance measures, trust correlated most strongly with willingness to trust Roboldo's news (rho .51) and weakest with seeing it as more reliable than human newscasters (rho .33, p = .035).

On the fabrication test, the binomial test found detection did not exceed chance (p = .989). The authors read the spread of answers as uncertainty rather than confident detection. They conclude that news organisations using robot presenters need transparency measures and ways to encourage critical engagement.

Women's baseline concern about lack of empathy (65.0%) was more than double men's (31.6%), but gender differences were not statistically significant in any category.