SUN-818 AI- Chatbots: Do they help our obese teenaged Type 2 DM patients?

Abstract Disclosure: G. Wu: None. I. Chim: None. M. Nguyen: None. B. Hoang: None. N. Mansubi: None. V.V. Toram: None. R. Toram: None. A. Madala: None. Background: Many patients ask healthcare questions to AI-mediated chatbots/LLMs, such as ChatGPT, Gemini, Meta AI and Character.ai, seeking answers instead of visiting physicians. Teenagers log onto Character.ai to find personalized and “friend-like” responses. In 2024, in Florida, one teenager committed suicide after his interactions with his chatbot on Character.ai. Obesity in adolescence can lead to Type 2 Diabetes. In addition, obesity alone, causes anxiety among teenagers. In WHO, boys aged 5-19 have a higher prevalence of obesity than girls in high and upper middle income countries worldwide. For female teenagers, obesity causes more psycho-social morbidity than in boys in the US. In the US, chatGPT has 123.5 million active users in 2024. Purpose: How effectively do various AI platforms communicate informative self-care plans to overweight, teenage patients with Type 2 diabetes?Methods: We queried ChatGPT, Gemini, Meta AI, Character.ai with the question: “I am a chubby, 18 year old girl with a new diagnosis of Type 2 Diabetes. I feel insecure about my body image. What should I do?” The response was evaluated with reading metrics (Flesch-Kincaid, Gunning Fog, Smog Index, Dale-Chall) and text analysis with cosine similarity score. For cosine similarity, we compared the responses from the other LLMs to ChatGPT because ChatGPT is the most queried of all the LLMs. Results: Mention of GLP-1: 0/4; “Go see a professional:” 4/4. “Go see a doctor/healthcare provider:” 4/4. Medical Disclaimers: 1/4( ChatGPT and Character.ai); Weblinks: 1/4 (Gemini 1.5); Mention of DMES/ADCES7: 0/4. ChatGPT, Gemini, Meta.AI provided more detailed actionable steps compared to Character. The lowest grade level was Character.ai at 4th grade. ChatGPT had the highest Grade Level scores (9th grade) and the longest response. Total word count: range 60-668, avg=310.8, sd=266.8, median 281. Character.ai had the shortest word count (60 words). All LLM’s used words such as “I understand,” “It is important to have a support system,” yet none mentioned ADCES7 which has healthy coping as part of diabetes self-care behaviors. Flesch-Kincaid Grade Level: range= 4.8-9.4, avg= 7.24, sd=2.22; median 8.3 for all LLMs and Character.ai. Conclusion: All the LLMs and Character.ai are sympathetic and have useful information for the users. There needs more training of these LLMs on the DMSES and ADCES7 would be helpful for our teenaged, overweight Type 2 Diabetic patients. Presentation: Sunday, July 13, 2025