Our work is informed by literature on early adolescent moral development and Ajzen’s (1991) theory of planned behavior (TPB). Early adolescence (ages 11–14) involves significant cognitive and social changes, including growing autonomy, perspective-taking, and concern for justice, fairness, and others’ rights (Eccles, 1999; Mitali et al., 2021; Smetana & Turiel, 2006). Peers and adults influence adolescents’ moral judgments, shaping perceptions of “right” and “wrong” use of technologies such as generative AI (GenAI).
The TPB provides a framework for understanding intentions to adopt behaviors such as fact-checking online content. Intentions are shaped by attitudes, subjective norms, and perceived behavioral control (Ajzen, 1991). TPB has been applied to middle school students’ views on GenAI, online interactions, and peer engagement (Authors, in press; Obrusnikova et al., 2012; Pan et al., 2023).
Middle school students can develop skills for using and evaluating GenAI through direct instruction and hands-on practice. Tools like ChatGPT enable exploration of academic content, but research indicates that students often over-rely on AI, struggle to ask effective questions, and produce incomplete or inaccurate explanations (Abdelghani et al., 2025; Solyst et al., 2024). Instruction that highlights GenAI features, limitations, ethical considerations, and effective fact-checking strategies can improve understanding and performance (Ali et al., 2021; Chen et al., 2023; Kong et al., 2024).
Instruction in lateral reading further enhances students’ ability to validate digital content, including AI-generated outputs. Middle and high school students exposed to curricula such as CTRL-F (https://www.ctrl-f.ca/) increased both use of and preference for lateral reading, improving their ability to investigate sources, assess claims, and trace information to original sources (Brodsky et al., 2023; Walsh-Moorman et al., 2020; Wineburg et al., 2022).
These practices support “civic online reasoning,” equipping students to locate, evaluate, and verify online information in meaningful contexts (Digital Inquiry Group, n.d.; McGrew, 2020). Because GenAI can obscure sources and mimic authoritative language, guided experiences with it enable students to critically assess AI content and reflect on the ethical implications of using it while exploring real-world community issues responsibly (Wineburg & Ziv, 2024; CivXNow & Listen First, 2024).
We used a mixed methods approach to answer the research questions, combining quantitative data (from surveys) and qualitative data (from a focus group that included a performance task) to provide a comprehensive understanding of middle schoolers’ perspectives on fact-checking online content, including output from GenAI tools. The research questions were:
1. How does a fact-checking lesson influence middle school students’ abilities and intentions to evaluate online sources, including AI-generated information?
2. How do attitudes, subjective norms, and perceived behavioral control influence middle school students’ intentions to fact-check online information?
Participants and Context
A free week-long summer civics camp was held at a private university located in a large urban area in the Southwest. Two main strategies were used to recruit racially and economically diverse students for the camp: (a) inviting past participants to attend and (b) asking middle-school principals, counselors, and teachers to nominate potential participants. The camp roster included 57 students who would be entering grades 6–9 in the upcoming school year; 15 students did not provide consent (parent) or assent (youth) for participation in the study.
The demographic and educational backgrounds of the study participants were diverse. The participants were 50% male and 50% female. Their racial/ethnic identification included 14 (33%) Hispanic/Latino/Chicana, 11 (26%) Black/African American, 10 (24%) Caucasian/White, 5 (12%) multiracial, 1 (2%) Asian/Pacific Islander, and 1 (2%) who did not specify. Additionally, 33 (79%) attended public school, 4 (10%) attended private school, 3 (7%) were homeschooled, and 2 (4%) did not specify.
Data Collection and Analysis
On the second day of camp, students responded to a 16-item survey. The survey asked students about their knowledge and use of GenAI along with various other information sources, followed by 10 Likert-style questions which assessed TPB dimensions (intentions, attitudes, subjective norms, and perceived behavioral control).
After responding to the survey, the students played an iCivics game called "Newsfeed Defenders" (https://ed.icivics.org/games/newsfeed-defenders) in which they practiced judging the credibility of online news sources. Later that day, the first author presented a one-hour media literacy lesson using Peardeck, which allowed her to collect some text-based responses during the lesson. The lesson focused on finding and evaluating sources that covered both traditional searches and GenAI queries, and provided practice on using lateral reading to verify online content, including GenAI output. The lesson was based on curriculum provided by CIVIX (n.d.), Common Sense Education (n.d.), and iCivics Education (n.d.).
On the fourth day of camp, students answered another 16-item survey. This survey repeated the 10 Likert-style questions assessing TPB dimensions, plus new questions designed to evaluate students' fact-checking knowledge.
Finally, nine students participated in a focus group that included a fact-checking performance task, so that we could gather more in-depth information about their perceptions regarding fact-checking GenAI content. The first author used demographic information as well as data from the second survey to select the focus group participants. She selected students so that they varied in demographic characteristics, participated in different project groups, and demonstrated various levels of proficiency with GenAI (based on answers to the post-survey questions).
Quantitative survey data analysis included calculation of descriptive and inferential statistics, plus creation of charts and graphs to analyze survey responses. Qualitative data were analyzed using thematic analysis to identify patterns and themes in the focus group transcript.
Preliminary Quantitative Results
Survey results suggested that the middle school students approached searching for and fact-checking online sources, including GenAI output, with a mix of confidence and skepticism. They reported using a variety of tools to locate information and valued fact-checking, often in line with adult expectations, though they were less certain about whether peers shared this view. On the pre-survey, 21% reported knowing “a great deal” about GenAI, while 42% knew “a fair amount.” Over half (57%) had used GenAI at least occasionally, and most relied on search engines (83%) to find information, followed by social media (60%), Wikipedia (40%), news sites (37%), and library databases (25%). About 42% somewhat agreed that GenAI outputs were accurate, though 34% were unsure; uncertainty dropped to 21% on the post-survey. Roughly two-thirds believed fact-checking online sources, including ChatGPT, was important—slightly more so at the end of the camp—and about the same proportion linked fact-checking to becoming a more responsible citizen. While 70% thought trusted adults valued fact-checking, only 55% believed their peers did; both perceptions decreased on the post-survey (to 66% and 42% respectively).
Preliminary findings also indicated that while students intended to use fact-checking strategies, they did not always rely on effective approaches. About three-fourths expressed an intention to fact-check online information, including GenAI outputs, on both surveys. Roughly two-thirds initially felt confident identifying false or misleading information and found it easy to do so, rising to about three-fourths on the post-survey. Fewer students (74%, down from 83%) agreed they had enough time to fact-check, perhaps recognizing that lateral reading requires more time than other strategies. Pre-survey results showed that 77% would consult other sources to verify information, but many also selected less effective strategies: 64% looked for strong evidence within the source, 50% relied on whether content “seemed reasonable or made sense,” 49% considered professional appearance or domain extensions, and others turned to family, friends (53%) or online comments (42%) for help judging sources.
Preliminary Qualitative Findings
The fact-checking performance task asked students to read a partially true AI-generated paragraph about community opposition to a plan to drill gas wells near a preschool. During the focus group, participants reported that using GenAI tools is generally easy, though finding “unbiased” information can be “tricky.” One student likened using ChatGPT or other GenAI tools to a “two-sided coin,” noting complexities around biased information and the sources GenAI tools draw from. Preliminary analysis of the focus group data suggests that the middle school participants relied on three main fact-checking strategies across the performance task, their camp inquiry projects, and other projects outside of camp. Each strategy is briefly described below, with supporting evidence from the transcript.
Trusting Instincts. A few participants explained that during fact-checking, they sometimes relied on their own sense of accuracy. One participant disclosed, “The website that I saw, I just trusted. I knew it was fine. I can't really explain why.” Another said, “I couldn't find it [a number referenced in the performance task article]. I just thought it wouldn't be 1200 feet.” In these cases, fact-checking ended in conclusions by participants that were not grounded in evidence.
Using Familiar Sources. Other times, participants talked about trusting sources that were familiar to them in the fact-checking process. For example, one participant mentioned using a news website that, “I've used before, and I kind of trust that…” Another participant whose group was studying animal welfare reported using PetSmart’s website for fact-checking because,“I've been there before, and … that's where my dog goes, so I kind of know they're pretty credible.” In these instances, prior use or familiarity with a source gave participants a sense of credibility when fact-checking.
Being Strategic. Finally, participants shared strategic fact-checking strategies they either learned at camp or had used in the past. For one camper, something new learned at camp was, “...that when you [search for] something, the first one [website at the top of search results] is not always the best.” Several other participants mentioned checking the credibility of authors and organizations, as well as cross-referencing a variety of sources when evaluating information. These sources included informational videos and websites relevant to the topic (e.g., using a popular psychology source to define mental health terms associated with teen mental health).
Given the prevalence of misinformation and disinformation in today’s society, students must learn to verify information using proven strategies. This study addresses the urgent need to strengthen students’ ability to critically evaluate digital content, including GenAI outputs. Teachers, principals, and technology coaches will gain insights into how middle school students perceive and apply fact-checking strategies, as well as ideas regarding instruction that builds civic online reasoning skills. Overall, this study contributes to a broader understanding of how adolescents can engage with emerging technologies critically, responsibly, and thoughtfully.
Abdelghani, R., Murayama, K., Kidd, C., Sauzéon, H., & Oudeyer, P.-Y. (2025). Investigating middle school Students' question-asking and answer-evaluation skills when using ChatGPT for science investigation. https://doi.org/10.48550/arxiv.2505.01106
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211. https://doi.org/10.1016/0749-5978(91)90020-T
Ali, S., DiPaola, D., Lee, I., Hong, J., & Breazeal, C. (2021). Exploring generative models with middle school students. Proceedings of the CHI Conference on Human Factors in Computing Systems, 1–13. https://doi.org/10.1145/3411764.3445226
Authors. (in press).
Brodsky, J. E., Brooks, P. J., Pavlounis, D., & Johnston, J. L. (2023). Instruction increases Canadian students’ preference for and use of lateral reading strategies to fact-check online information. AERA Open, 9. https://doi.org/10.1177/23328584231192106
Chen, B., Zhu, X., & Díaz del Castillo, H. F. (2023). Integrating generative AI in knowledge building. Computers and Education: Artificial Intelligence, 5, 100184. https://doi.org/10.1016/j.caeai.2023.100184
CIVIX. (n.d.). CNTL-F dIgital media literacy classroom resources: Verification skills. https://ctrl-f.ca/en/resources/
CivXNow & Listen First. (2024). Uncharted waters: Education, democracy, and social cohesion in the age of artificial intelligence. https://civxnow.org/wp-content/uploads/2024/03/ CivxNow-Report-2.pdf
Common Sense Education. (n.d.). AI literacy lessons for grades 6-12. https://www.commonsense.org/education/collections/ai-literacy-lessons-for-grades-6-12
Digital Inquiry Group. (n.d). The cornerstone of civic online reasoning.
https://cor.inquirygroup.org/about/
Eccles, J. S. (1999). The development of children ages 6 to 14. The Future of Children, 9(2), 30-44. https://doi.org/10.2307/1602703
iCivics Education. (n.d.). Media and Influence. https://ed.icivics.org/curriculum/media-and-influence
Kong, S.-C., Lee, J. C.-K., & Tsang, O. (2024). A pedagogical design for self-regulated learning in academic writing using text-based generative artificial intelligence tools: 6-P pedagogy of plan, prompt, preview, produce, peer-review, portfolio-tracking. Research and Practice in Technology Enhanced Learning, 19, 30-48. https://doi.org/10.58459/rptel.2024.19030
McGrew, S. (2020). Learning to evaluate: An intervention in civic online reasoning. Computers and Education, 145, 103711. https://doi.org/10.1016/j.compedu.2019.103711
Mitali, T., Galarneau, E., & Peplak, J. (2021). Moral development in adolescence. Journal of Research on Adolescence, 31(4), 1097-1113. https://doi.org/10.1111/jora.12639
Obrusnikova, I., & Dillon, S. R. (2012). Students’ beliefs and intentions to play with peers with disabilities in physical education: Relationships with achievement and social goals. Journal of Teaching in Physical Education, 31(4), 311–328. https://doi.org/10.1123/jtpe.31.4.311
Pan, Y., Huang, Y., Kim, H., & Zheng, X. (2023). Factors influencing students’ intention to adopt online interactive behaviors: Merging the theory of planned behavior with cognitive and motivational factors. The Asia-Pacific Education Researcher, 32(1), 27–36. https://doi.org/10.1007/s40299-021-00629-y
Smetana, J. G., & Turiel, E. (2006). Moral development during adolescence. In G.R. Adams & M. Berzonsky (Eds.), Blackwell Handbook of Adolescence (pp. 247-268). Blackwell Publishing. https://doi.org/10.1002/9780470756607.ch12
Solyst, J., Yang, E., Xie, S., Hammer, J., Ogan, A., & Eslami, M. (2024). Children’s overtrust and shifting perspectives of generative AI. https://doi.org/10.48550/arxiv.2404.14511
Walsh-Moorman, E. A., Pytash, K. E., & Ausperk, M. (2020). Naming the moves: Using lateral reading to support students’ evaluation of digital sources. Middle School Journal, 51(5), 29–34. https://doi.org/10.1080/00940771.2020.1814622
Wineburg, S., Breakstone, J., McGrew, S., Smith, M. D., & Ortega, T. (2022). Lateral reading on the open internet: A district-wide field study in high school government classes. Journal of Educational Psychology, 114(5), 893–909. https://doi.org/10.1037/edu0000740
Wineburg, S. & Ziv, N. (2024). What makes students (and the rest of us) fall for AI misinformation? Edweek. https://www.edweek.org/technology/ opinion-what-makes-students-and-the-rest-of-us-fall-for-ai-misinformation/2024/10
Other presentations in this group:
Back
Trips and Tours
