Does machine authorship affect news perceptions? A meta-analysis

The growing adoption of artificial intelligence in journalism has dramatically changed the way news is produced. Despite the recent proliferation of research on automated journalism, there is still no consensus on how audiences perceive and evaluate news purportedly written by machines compared to humans. Based on a review of 24 experimental studies, this meta-analysis showed that machine authorship had a negative, albeit small, effect on credibility perceptions. The effect was weaker for studies using news articles on a single topic than multiple topics. Furthermore, machine authorship had a negligible effect on news evaluations, although this effect was significantly stronger (more negative) when (a) news covered socio-political topics, (b) information about news automation systems was hidden (vs. disclosed), and (c) the actual source of news articles was a machine (vs. a human). These findings are discussed in light of theoretical accounts of human-machine communication and practical implications for news media.