Ullrich K. H. Ecker

Realistic, misleading AI-generated audio content is a new threat to digital-media consumers. This study examined the effects of misleading AI-generated podcasts that presented false claims about nanoparticles in sunscreen (Experiment 1; N = 463), or microwaves damaging foods (Experiment 2; N = 500). U.S. participants either received a misleading podcast, misleading article, or a neutral podcast (control). To test interventions, misleading podcasts were preceded by either a passive inoculation explaining how AI-generated information can deceive, an active inoculation that added an educational activity, or no intervention. The podcast and article produced comparable increases in misinformed beliefs. Both inoculations significantly reduced general trust in AI-generated information and attenuated the podcast’s impact on beliefs. Results demonstrate that AI-generated audio can influence health-related beliefs as strongly as text-based misinformation, and that brief AI-source-focused inoculations can mitigate, but not eliminate, its persuasive impact. Findings inform interventions targeting emerging forms of AI-driven misinformation.

Fact-checks reliably reduce belief in misinformation, but their effects diminish over time, partially due to forgetting—this is known as belief regression. Across two experiments, we tested whether memory-enhancing techniques could improve the long-term efficacy of fact-checks. In Experiment 1 (N = 555; one-week delay) we examined a mental-imagery-based mnemonic linking false and true claims to a ‘liar’ and ‘truthteller’, respectively, alongside conditions varying the extent to which active encoding of the fact-check was encouraged. The mnemonic did not improve fact-check memory or reduce belief regression. In contrast, briefly summarizing fact-checks improved memory and sustained belief change relative to a control condition. Experiment 2 (N = 269) replicated the enhanced effectiveness of fact-checks in the summarization condition over a longer delay period (two weeks); however, belief regression did not significantly differ from control. These findings suggest that low-barrier interventions implemented at the time of encoding can improve the effectiveness of fact-checks but may not slow the rate of belief regression. To provide theoretical and practical insight into the drivers of belief regression, future research may benefit from examining the impact of retrieval-based interventions.

Misinformation has become a much‐discussed issue in public and academic domains. One harmful aspect of misinformation is that it can persistently influence people's judgements after clear and credible correction—a phenomenon known as the continued influence effect. There has been speculation that age affects people's susceptibility to the continued influence effect, but previous research has produced inconclusive evidence, specifically across testing modalities. To try and reconcile conflicting findings, the current study used a continued‐influence paradigm wherein causal event misinformation was retracted. Two samples were analysed using correlation and linear mixed‐effects modelling: an online sample aged 21–79 ( N = 259) and an in‐person sample aged 56–92 ( N = 143). Results showed that, when controlling for encoding of experimental materials in the online sample, age had no significant effect on the effectiveness of misinformation retractions, though declining effectiveness was observed with advanced older age in the in‐person sample.

Misinformation influences cognition—shaping memory, beliefs, attitudes, reasoning, and decision making. While intentionally disseminated misinformation (i.e., disinformation) has long served persuasive functions, its strategic use to generate epistemic uncertainty has grown in prominence. Such strategic manipulation by incentivised actors serves to construct “alternative realities” that undermine shared truth and institutional trust. This paper reviews recent evidence on the direct cognitive and behavioural impacts of misinformation, examines the transformative roles of social media and generative artificial intelligence, and considers broader epistemological consequences, including the erosion of truth norms and implications for democratic discourse. Addressing these challenges requires coordinated individual and systemic interventions that integrate psychological, technological, educational, and regulatory approaches to strengthen epistemic resilience and safeguard evidence-informed public reasoning and decision-making.

Fact-checks can reduce—but scarcely eliminate—misinformation impacts, and their effects are often short-lived. Across two experiments (total N = 471), we examined whether graphs could bolster the effectiveness and longevity of fact-checks, through increasing their memorability. Participants rated their belief in accurate and misleading headlines, then received corresponding fact-checks, some of which included a graph. Participants then rated their belief in each headline again as well as their memory of each fact-check, both immediately and after a 1-week (Experiment 1) or 4-week (Experiment 2) delay. In Experiment 2, graphs were also sometimes accompanied by a multiple-choice question to encourage deeper engagement. Graphs alone improved memory for fact-checks but did not consistently improve discernment between accurate and misleading headlines; graphs presented with a multiple-choice question, however, improved discernment immediately and after a 4-week delay. Findings demonstrate the potential of interventions combining visual content with active engagement for combatting misinformation impacts.

We examined peer-to-peer transmission of misinformation and corrections, an understudied mechanism of misinformation spread. Participants (n = 120 per wave; total N = 600) viewed and wrote summaries of four news articles (conditions: control, misinformation-only, weak correction, strong correction). These summaries were passed down four subsequent waves of participants, each wave summarizing for the next. Summaries were coded for the presence of misinformation and correction details, and misinformation reliance. Initially (position one), both weak and strong corrections reduced misinformation reliance, although strong corrections were more effective. However, both misinformation and correction details were lost during transmission, leading to increased misinformation reliance in the correction conditions. By position five, there was no significant difference in misinformation reliance between the weak correction and misinformation-only conditions, showing that corrections can “lose out” to misinformation during peer-to-peer transmission and highlighting the importance of using detailed corrections to prevent misinformation reliance re-emerging.

Generative Artificial Intelligence (AI) systems are increasingly used to generate police-incident reports. Yet such systems can produce errors, which may detrimentally impact legal processes. The current study examined the impact of an AI-generated report on memory and perceptions of a police incident, and whether safeguards can reduce those impacts. Participants (N = 432) read an accurate summary of a domestic-violence incident; 24 hours later, they read an AI-generated report that was either accurate or contained misinformation, before completing memory tests and rating their perceptions of the incident. Two safeguards were tested: (1) active error-monitoring, and (2) watching body-worn camera footage. Misinformation significantly impaired memory and influenced perceptions of the incident—especially in participants who strongly identified with police—and the safeguards were largely effective at reducing these impacts. Findings demonstrate the risks of generative-AI usage in forensic settings, requiring the implementation of appropriate safeguards for combatting its detrimental effects.