The rise of accessible generative AI streamlines deepfakes generation and diffusion, further challenging information quality in the digital age. This paper investigates how deepfake salience, combining deepfake plausibility and platform measures aimed at signaling deepfake presence, affects news-sharing of news from reliable sources on social media platforms. Departing from the MAIN model, we show how deepfake salience alters heuristics availability, changing users' perceived credibility of news content and consequently affecting news-sharing. We do so through two studies. The first exploratory study examines, through observatory data, how deepfake salience affects engagement with deepfakes, while the second study, through an online experiment, tests the effects of deepfake plausibility on news-sharing. Our findings reveal that while news from reliable news sources is generally perceived as more credible and consequently shared more, the presence of high-plausibility deepfakes negatively moderates this perception, especially when countermeasures like labeling are applied to the deepfake. Furthermore, these countermeasures do not affect engagement with the deepfake content. This study contributes to the literature on platforms' second-order effects on credibility and the dissemination of reliable news. It also contributes to the discourse on online information quality and offers practical implications for social media platforms seeking effective strategies against misinformation.
