Generative AI and Eating Disorders
Using generative AI safely: what people experiencing eating disorders, disordered eating and body image concerns and their supporters should know
The key message
People are increasingly using generative artificial intelligence (AI) tools, such as ChatGPT, Gemini and Claude, to ask questions about their health and well-being. There are many reasons someone might use AI. It is easy to access, available at any time and may feel easier or more private than talking to another person. For some people, professional healthcare may also be difficult to access or afford.
We do not want people to feel ashamed or judged for using AI tools. Our aim is to help people use them more safely and critically, find trusted information, stay connected with other people and seek professional support when needed.
However, AI can provide information that is incorrect or unsupported, including summaries that misrepresent the original source. Its responses may also be incomplete, biased or unsafe.
AI is not a replacement for professional healthcare or human support.
Remember
AI creates responses based on patterns in large amounts of information. It can sound confident, caring and understanding, but it does not know you personally or fully understand your health, experiences or needs.
The information used to develop AI tools may also contain biases. This can include weight bias, assumptions about who does or does not experience eating disorders, and Western or culturally narrow ideas about food, health and bodies.
AI may be useful for
- Finding general information about eating disorders, disordered eating and body image
- Explaining health information in simpler language
- Helping you think of questions to ask a healthcare professional
- Researching organisations and support services
- Connecting to eating disorder helplines
- Helping you prepare to talk with a healthcare professional or trusted person
- Taking a first step towards seeking help when talking to someone feels difficult
Take extra care if asking AI about
- Food, eating and meal planning
- Exercise
- Weight or changing your body or appearance
- Symptoms of eating disorders
- Whether you might have an eating disorder
- Treatment
- Recovery
- Advice that is specific to you
It is understandable that people may want to ask AI about these things. However, answers to these questions can sometimes reinforce eating disorder thoughts, behaviours and unrealistic beauty/appearance ideals, including for people who do not have an eating disorder but are experiencing disordered eating or body image concerns.
If you use AI, use it to find information or prepare to talk with people, not to make important health decisions for you.
Do not rely on AI to
- Diagnose an eating disorder
- Decide whether you are "sick enough" to seek help
- Tell you what or how much you "should" eat
- Tell you how much you "should" exercise
- Tell you what your "ideal" weight might be or how to change your body
- Replace healthcare, treatment or ongoing human support
- Provide crisis support or manage urgent safety concerns
If something an AI tool tells you makes you worried about your health, consider talking about it with a healthcare professional, support service or trusted person. You can also report inappropriate or unsafe responses using the tool's feedback or reporting function, where available.
Ask yourself before trusting an AI response
- Can I verify this information?
- AI can give convincing answers that are wrong and may provide sources that do not exist or do not support its claims. Check that any sources are genuine and verify important information using reputable health organisations or other trusted sources.
- Could this information be harmful?
- Advice that seems harmless can sometimes reinforce eating disorder thoughts, disordered eating or body image concerns.
- Could there be bias in this answer?
- Think about whether the response makes assumptions about weight, bodies, gender, culture, food or who experiences eating disorders.
- Does this really apply to me?
- General purpose AI does not have a complete understanding of your health, circumstances or experiences.
- Would it help to talk about this with someone?
- AI can help you organise your thoughts, but another person can listen, understand your individual situation, notice when something may be wrong and provide ongoing human connection and support.
Stay connected with people
An AI conversation can sometimes feel very personal. Some tools are designed to communicate in ways that sound warm, supportive and empathetic.
AI may feel helpful, but it is not a friend, healthcare professional or personal relationship. It cannot provide understanding and connection that come from relationships with other people.
Be aware if using AI starts to replace talking with people, makes you more isolated or becomes your main source of emotional or health support.
Where possible, let AI open the door to human connection rather than replace it. For example, you could ask it to help you:
Protect your privacy
AI tools may collect, store or use information you enter. Think carefully before sharing personal or sensitive information.
Be particularly careful about entering
- Your full name, contact details or other identifying information
- Detailed medical or mental health information
- Identifying information about another person
- Sensitive photographs, documents or other personal information
Before using an AI tool, look at its privacy settings and privacy information. Consider what information is collected, whether your conversations are stored, who may have access to them and how the information you provide may be used.
For parents, carers and supporters
Keep being someone they feel comfortable talking to
If someone you support is using AI for eating disorder, eating, body image or health questions, try to approach the conversation with curiosity rather than judgement.
You can help by
- Learning about the AI tools and platforms they are using
- Asking open questions about how and why they use them
- Avoiding shaming them or immediately trying to stop their AI use
- Helping them check important information against trusted sources
- Talking together about information that may be harmful or biased
- Encouraging them to stay connected with friends, family and other trusted people
- Supporting them to speak with healthcare professionals when needed
Conversation starters
Signs an AI response may be unsafe
Be cautious if AI:
- Promotes weight loss, dieting, restriction or compensatory behaviours
- Focuses heavily on appearance, weight or comparing bodies
- Provides specific food, exercise, calorie or weight recommendations
- Uses judgemental, stigmatising or shame-based language
- Claims to know for certain whether you have an eating disorder or what treatment you need
- Encourages secrecy
- Suggests avoiding healthcare or other support
- Encourages you to rely on AI instead of other people
- Gives health information as fact without reliable sources or acknowledging uncertainty
Not all AI and digital health tools are the same
There is an important difference between general purpose generative AI tools and purpose-built digital health tools, including chatbots, that have been specifically designed and evaluated for particular health purposes. If you are considering an AI-based or digital health tool, look for clear information about:
Purpose
What is the tool designed to do — and what is it not designed to do?
People
Were people with relevant lived experience and appropriate health expertise involved in its development?
Evidence
Has the tool been evaluated and is there evidence supporting its safety and effectiveness?
Safety
What safeguards are in place, particularly for eating disorder and mental health risks?
Privacy
What happens to the information you provide?
Support
What happens if you need professional or urgent help?
Transparency
Is the organisation clear about when and how AI is being used?
A tool being described as “AI for health” does not by itself mean that it has been tested or shown to be safe.
Overall
Using AI to ask about health is increasingly common. For some people, it may be one of the easiest or most accessible places to start.
The goal is not simply to tell people not to use AI. Instead, we want people to use it critically and safely: recognising its limitations, checking important information, protecting their privacy, staying connected with other people and seeking professional support when needed.
AI can sometimes be useful as a first step, but it should help open the door to reliable information and human connection — not close it.
Given the dynamic evolution of AI, this resource represents our understanding of the known dangers at this time (19 September 2026). Future revisions will be necessary.
Key references
- World Health Organization. Ethics and governance of artificial intelligence for health: guidance on large multi-modal models. Geneva: WHO; 2024. https://iris.who.int/server/api/core/bitstreams/e9e62c65-6045-481e-bd04-20e206bc5039/content
- Sharp G, Marini S, Tam E, Hu H. Body Image and Eating Disorder Education Chatbot, JEM, in Australia and Canada: First 6-Month Real-World Survey Evaluation. JMIR Hum Factors. 2026;13:e90783. https://humanfactors.jmir.org/2026/1/e90783
- Yim SH, Yoo DW, Polymerou A, Liu Y, Saha K. Generative AI for Eating Disorders: Linguistic Comparison With Online Support and Qualitative Analysis of Harms. Int J Eat Disord. 2026. https://onlinelibrary.wiley.com/doi/10.1002/eat.24604
This resource has been developed by Consortium for Research in Eating Disorders in consultation with, and endorsed by, the following international organisations: