Overview of key insights from main research on AI-generated content, presented chronologically from the oldest to the most recent
| Author(s) | Context | Method | Key insights/findings |
|---|---|---|---|
| Waddell (2018) | AI-generated news | Quantitative approach | The study revealed that news attributed to a machine is perceived as less credible than news attributed to a human journalist. The authors also observed a negative effects of machine authorship through the indirect pathway of source anthropomorphism and negative expectancy violations, with evidence of moderation by prior recall of robotics |
| Graefe and Bohlken (2020) | AI-generated news | Meta-analysis | This meta-analysis’ results from 12 studies involving a total of 4,473 participants revealed no significant difference in perceived credibility between human and AI-generated news. However, a slight edge for human-written news regarding quality and a substantial advantage for human-written news in terms of readability were found. Experimental comparisons indicated that participants rated credibility, quality, and readability higher when informed that they were reading a human-written article |
| Kim et al. (2020) | AI-generated content (text, audio, and video) | Quantitative approach | The study examined how different content generators (human vs. AI) and information delivery methods (text, audio and video) influence users’ perceptions of content. The findings suggest that both the type of generator and the delivery method significantly impact the perceived quality, satisfaction and readability of the content |
| Moravec et al. (2024) | AI-crafted journalism | Quantitative approach | The study identified gender, age, and socioeconomic status as significant factors influencing respondents’ ability to recognize the source of text. Females were more successful at identifying human-generated texts, while males excelled at recognizing AI-generated texts. Younger respondents were generally better at identifying AI-generated content, and higher education and income levels correlated with improved accuracy. Attitudes toward AI in journalism varied by age: those aged 18–29 showed ambivalence, 30–49 were uncertain, 50–69 had diverse views and those 70 and older were skeptical. Males, particularly in older age groups, were more optimistic about AI’s potential in journalism compared to females |
| Wu et al. (2020) | AI-generated artistic content | Quantitative approach | An experiment was conducted to explore subjects’ explicit and implicit perceptions of AI-generated content in the U.S. and China. The two countries showed differing attitudes towards AI’s performance in artistic work. U.S. subjects were more critical of AI-generated content compared to human-generated content, both explicitly and implicitly. In contrast, while Chinese subjects expressed overt positivity towards AI-generated content, they valued it less than human-authored content |
| Wu and Wen (2021) | AI-generated ads | Quantitative approach | This study explored the factors that affect consumers’ overall appreciation of AI-created advertisements. The findings revealed consumers’ perception of the objectivity of the advertisement creation process positively influences the machine heuristic - a belief that machines are more secure and trustworthy than humans. This perception enhanced consumer appreciation for AI-generated advertisements. However, the perceived objectivity of the ad creation process negatively affected the perceived eeriness of AI advertising, which in turn diminished appreciation for these ads. Additionally, consumers’ discomfort with robots positively influenced both the machine heuristic and the perceived eeriness of AI advertising |
| Arango et al. (2023) | AI-generated charitable giving ads | Quantitative approach | The research found that potential donors reacted differently to children’s faces when they were aware that these images were AI-generated. Knowing an image is artificial negatively affected donation intentions, with this effect being mediated by empathy and anticipatory guilt, as well as by empathy and emotion perception. AI-generated images can enhance their effectiveness by emphasizing their ethical intentions and in extraordinary situations, the use of AI-generated images by charities is deemed acceptable by consumers |
| Chaisatitkul et al. (2024) | AI-generated storyboards | Interview | The research findings revealed that consumers had a positive outlook and greater liking for works created by generative AI, as they perceived it as “unbiased” |
| Hitsuwari et al. (2023) | AI-generated poetry* | Quantitative approach | The study’s results indicated that the beauty rating of the AI-generated haiku created with human intervention was the highest, while the ratings for both human-made and AI-generated haiku without human involvement were identical. Participants were unable to differentiate between human-made and AI-generated haiku. These findings imply that human-AI collaboration enhances creativity in haiku production. Furthermore, a negative correlation was observed between discrimination performance and beauty ratings in AI-generated haiku, indicating that high-quality AI-generated works are often perceived as being human-created |
| Chen et al. (2024a) | AI-generated ads | Quantitative approach | Consumers have more positive attitudes toward AI-generated ads with agentic appeals, and the effect is mediated by task self-efficacy, while they have more positive attitudes toward human-created ads with communal appeals, and the effect is mediated by social self-efficacy. Assigning a social role to the AI advertising generator, a partner, or a servant role, helps mitigate or even reverse the negative effects of AI-generated ads with communal appeals |
| Grassini and Koivisto (2024) | AI-generated artwork | Quantitative approach | The study’s findings indicated that individual characteristics, such as creative personal identity and openness to experience, influence how people perceive artworks based on their believed source. Participants struggled to consistently differentiate between human and AI-generated images. While they generally preferred AI-generated artworks over human-made ones, a negative bias emerged when considering subjective source attribution. As a result, artworks perceived as AI-generated were rated as less preferable, regardless of their actual source |
| Park et al. (2024) | AI-generated content | Mixed-method approach | Participants struggled to differentiate between AI accounts and human accounts. Additionally, there were notable differences in how users perceived the three types of accounts. Participants found both AI and influencer accounts to be more appealing than public accounts, and they rated the quality of AI-generated content similarly to that of content produced by influencers |
| Song et al. (2024) | AI-generated advertisement | Quantitative approach | The study’s results demonstrated that the advertisements with rational appeals improved visit intention for AI-generated ads more effectively. In contrast, those with emotional appeals were more attractive when the declared creator was human |
| Zhang et al. (2024) | AI-generated content | Mixed-method approach | This study employed a mixed-method approach including grounded theory and content analysis. Grounded research, conducted on an innovative digital tourism platform in China, identified three key dimensions of content quality in digital tourism interpretation: informativeness, emotional appeal and empathy. Additionally, content analysis and ANOVA results revealed that AI-generated content displayed lower quality across all three dimensions compared to professionally generated content. The findings support the conclusion that AI cannot replace human professional interpreters in terms of content quality for interpretation |
| Kirk and Givi (2025) | AI-generated follow-up email | Quantitative approach | Seven preregistered studies show that consumer belief that marketing communications are AI-generated (vs. human) reduces positive word-of-mouth and loyalty. The “AI-authorship effect” is mediated by perceived authenticity and moral disgust, and is weaker for factual (vs. emotional) messages, AI-edited content, AI-signed communications, and when consumers perceive most marketing as AI-generated |
| Author(s) | Context | Method | Key insights/findings |
|---|---|---|---|
| AI-generated news | Quantitative approach | The study revealed that news attributed to a machine is perceived as less credible than news attributed to a human journalist. The authors also observed a negative effects of machine authorship through the indirect pathway of source anthropomorphism and negative expectancy violations, with evidence of moderation by prior recall of robotics | |
| AI-generated news | Meta-analysis | This meta-analysis’ results from 12 studies involving a total of 4,473 participants revealed no significant difference in perceived credibility between human and AI-generated news. However, a slight edge for human-written news regarding quality and a substantial advantage for human-written news in terms of readability were found. Experimental comparisons indicated that participants rated credibility, quality, and readability higher when informed that they were reading a human-written article | |
| AI-generated content (text, audio, and video) | Quantitative approach | The study examined how different content generators (human vs. AI) and information delivery methods (text, audio and video) influence users’ perceptions of content. The findings suggest that both the type of generator and the delivery method significantly impact the perceived quality, satisfaction and readability of the content | |
| AI-crafted journalism | Quantitative approach | The study identified gender, age, and socioeconomic status as significant factors influencing respondents’ ability to recognize the source of text. Females were more successful at identifying human-generated texts, while males excelled at recognizing AI-generated texts. Younger respondents were generally better at identifying AI-generated content, and higher education and income levels correlated with improved accuracy. Attitudes toward AI in journalism varied by age: those aged 18–29 showed ambivalence, 30–49 were uncertain, 50–69 had diverse views and those 70 and older were skeptical. Males, particularly in older age groups, were more optimistic about AI’s potential in journalism compared to females | |
| AI-generated artistic content | Quantitative approach | An experiment was conducted to explore subjects’ explicit and implicit perceptions of AI-generated content in the U.S. and China. The two countries showed differing attitudes towards AI’s performance in artistic work. U.S. subjects were more critical of AI-generated content compared to human-generated content, both explicitly and implicitly. In contrast, while Chinese subjects expressed overt positivity towards AI-generated content, they valued it less than human-authored content | |
| AI-generated ads | Quantitative approach | This study explored the factors that affect consumers’ overall appreciation of AI-created advertisements. The findings revealed consumers’ perception of the objectivity of the advertisement creation process positively influences the machine heuristic - a belief that machines are more secure and trustworthy than humans. This perception enhanced consumer appreciation for AI-generated advertisements. However, the perceived objectivity of the ad creation process negatively affected the perceived eeriness of AI advertising, which in turn diminished appreciation for these ads. Additionally, consumers’ discomfort with robots positively influenced both the machine heuristic and the perceived eeriness of AI advertising | |
| AI-generated charitable giving ads | Quantitative approach | The research found that potential donors reacted differently to children’s faces when they were aware that these images were AI-generated. Knowing an image is artificial negatively affected donation intentions, with this effect being mediated by empathy and anticipatory guilt, as well as by empathy and emotion perception. AI-generated images can enhance their effectiveness by emphasizing their ethical intentions and in extraordinary situations, the use of AI-generated images by charities is deemed acceptable by consumers | |
| AI-generated storyboards | Interview | The research findings revealed that consumers had a positive outlook and greater liking for works created by generative AI, as they perceived it as “unbiased” | |
| AI-generated poetry* | Quantitative approach | The study’s results indicated that the beauty rating of the AI-generated haiku created with human intervention was the highest, while the ratings for both human-made and AI-generated haiku without human involvement were identical. Participants were unable to differentiate between human-made and AI-generated haiku. These findings imply that human-AI collaboration enhances creativity in haiku production. Furthermore, a negative correlation was observed between discrimination performance and beauty ratings in AI-generated haiku, indicating that high-quality AI-generated works are often perceived as being human-created | |
| AI-generated ads | Quantitative approach | Consumers have more positive attitudes toward AI-generated ads with agentic appeals, and the effect is mediated by task self-efficacy, while they have more positive attitudes toward human-created ads with communal appeals, and the effect is mediated by social self-efficacy. Assigning a social role to the AI advertising generator, a partner, or a servant role, helps mitigate or even reverse the negative effects of AI-generated ads with communal appeals | |
| AI-generated artwork | Quantitative approach | The study’s findings indicated that individual characteristics, such as creative personal identity and openness to experience, influence how people perceive artworks based on their believed source. Participants struggled to consistently differentiate between human and AI-generated images. While they generally preferred AI-generated artworks over human-made ones, a negative bias emerged when considering subjective source attribution. As a result, artworks perceived as AI-generated were rated as less preferable, regardless of their actual source | |
| AI-generated content | Mixed-method approach | Participants struggled to differentiate between AI accounts and human accounts. Additionally, there were notable differences in how users perceived the three types of accounts. Participants found both AI and influencer accounts to be more appealing than public accounts, and they rated the quality of AI-generated content similarly to that of content produced by influencers | |
| AI-generated advertisement | Quantitative approach | The study’s results demonstrated that the advertisements with rational appeals improved visit intention for AI-generated ads more effectively. In contrast, those with emotional appeals were more attractive when the declared creator was human | |
| AI-generated content | Mixed-method approach | This study employed a mixed-method approach including grounded theory and content analysis. Grounded research, conducted on an innovative digital tourism platform in China, identified three key dimensions of content quality in digital tourism interpretation: informativeness, emotional appeal and empathy. Additionally, content analysis and ANOVA results revealed that AI-generated content displayed lower quality across all three dimensions compared to professionally generated content. The findings support the conclusion that AI cannot replace human professional interpreters in terms of content quality for interpretation | |
| AI-generated follow-up email | Quantitative approach | Seven preregistered studies show that consumer belief that marketing communications are AI-generated (vs. human) reduces positive word-of-mouth and loyalty. The “AI-authorship effect” is mediated by perceived authenticity and moral disgust, and is weaker for factual (vs. emotional) messages, AI-edited content, AI-signed communications, and when consumers perceive most marketing as AI-generated |
Note(s): *For a more comprehensive literature review comparing human-made and AI-generated poetry, refer to the study by Hitsuwari et al. (2023)
Sharing content requires targeting cookies to be enabled. Please update your cookie preferences to use this feature.