ORCID
https://orcid.org/0009-0009-3717-039X
Year
2025
Season
Fall
Paper Type
Doctoral Dissertation
College
Brooks College of Health
Degree Name
Doctor of Clinical Nutrition (DCN)
Department
Nutrition & Dietetics
NACO controlled Corporate Body
University of North Florida. Department of Nutrition & Dietetics
Committee Chairperson
Dr. Andrea Y. Arikawa
Second Advisor
Dr. Indika Kahanda
Third Advisor
Dr. Alan Flanagan
Department Chair
Dr. Andrea Y. Arikawa
College Dean
Dr. Mei Zhao
Abstract
Background: Nutrition misinformation on social media has intensified in recent years, particularly around dietary fats. Despite its reach, few studies have evaluated misinformation prevalence using an evidence-based reference standard or employed human–AI hybrid workflows to manage large-scale post classification.
Methods: This cross-sectional content analysis examined 1,024 Instagram posts discussing seed oils, saturated fats, omega-6 fatty acids, polyunsaturated fats, and unsaturated fats. A human–AI hybrid workflow was used, in which AI-assisted tools automated data extraction and supported coding. Accuracy was coded using a four-level scale aligned with the Dietary Guidelines for Americans, 2020–2025 and the 2020 Dietary Guidelines Advisory Committee Scientific Report. Tone, misinformation themes, professional credentials, functional/holistic identifiers, bio keywords, and engagement metrics (likes, comments, reel views, engagement rate) were also captured. Statistical tests examined associations among accuracy, tone, engagement, and author characteristics.
Results: 66.5% of posts were mostly inaccurate or inaccurate (p < .001), with misinformation heavily concentrated in posts about seed oils, omega-6 fats, and saturated fats. Unsaturated fat posts showed the most favorable accuracy profile. Accurate posts were predominantly neutral or informational in tone, whereas inaccurate posts were more often fear-based, prescriptive, or sensational (p < .001). Inaccurate posts received significantly more likes, comments, and reel views than accurate posts (all p ≤ .002), while engagement rate did not differ by accuracy classification (p = .42). Profession was strongly associated with accuracy (p < .001): credentialed dietetics professionals produced mostly accurate content (~75% accurate), as did doctoral-trained nutrition and science professionals (~88% accurate). In contrast, chiropractors, self-proclaimed nutritionists, influencers, and many brands posted predominantly inaccurate content (>75% inaccurate). Physicians were not more accurate than non-physicians; approximately 70% of physician-authored posts were mostly inaccurate or inaccurate. Functional/holistic credentials and biography language were strongly associated with misinformation (p < .001), including within the dietetics subgroup.
Conclusions: Dietary fat misinformation is pervasive on Instagram and disproportionately produced by non-expert and functionally branded creators. Inaccurate content also receives higher engagement, potentially amplifying its reach. Human–AI hybrid evaluation frameworks offer a scalable and evidence-based approach for monitoring nutrition misinformation and supporting targeted intervention strategies on social media platforms.
Suggested Citation
Martin, Charlotte Amanda, "Saturated facts: Leveraging artificial intelligence to identify and analyze dietary fat-related nutrition misinformation on social media" (2025). UNF Graduate Theses and Dissertations. 1377.
https://digitalcommons.unf.edu/etd/1377