Risks and benefits of artificial intelligence deepfakes : systematic review and comparison of public attitudes in seven European countries

This study provides an evidence-based integrated appraisal of artificial intelligence (AI)-generated deepfakes by integrating a cross-disciplinary literature synthesis with original opinion-poll evidence from seven European countries. A SWOT matrix distils convergent concerns-weaponised disinformation, privacy erosion, and the detection arms race-alongside under-explored opportunities in education, therapy, and creative industries. To test whether these scholarly themes resonate with citizens, a computer-assisted web survey (N = 7,083) measured perceived risks and benefits across 10 specific scenarios for each theme. Correspondence analysis and Bonferroni-adjusted means reveal a pronounced age gradient for benefits, whereas risk perceptions vary by country-younger cohorts are noticeably less alarmed only in Sweden, France, and Czechia. Geographically, Dutch, German, British, and Italian publics prove the most enthusiastic: the United Kingdom (UK) couples similar enthusiasm with markedly higher risk vigilance, whereas Czech and Swedish respondents remain consistently sceptical, underscoring a broad, though imperfect, west/south versus central/north divide. The Netherlands, Germany, the UK, and Italy value pro-social applications (i.e., realistic crisis drills, public-interest campaigns, and therapeutic 'mental-health' avatars), with the Netherlands topping four benefit items and Italy favouring commercial/entertainment uses such as virtual brand ambassadors. By contrast, Czech and Swedish respondents assign uniformly low benefit scores. Juxtaposed with risk perceptions, the UK and Czechia register the greatest vigilance, Sweden the most relaxed, and others intermediate. Divergence seems associated with digital literacy levels and regulatory maturity. The survey reveals a statistically and practically significant gap between perceived risks and benefits: across all seven countries, respondents, on average, rate risks higher than advantages. Regression estimates indicate that advancing age, lower household income, and gender (woman) enlarge this gap-primarily by undermining perceived benefits-whereas tertiary education and residence in certain western or southern European countries-notably, Germany and Italy-are associated with more balanced appraisals. This study concludes that layered governance, interoperable detection standards, and targeted literacy programmes are urgently required.