Analyze Supplement Ingredients with AI and Opus 4.5

What you’ll get

A clear, evidence-weighted breakdown of what is actually in a supplement: for each ingredient, what it claims to do and how strong the science behind that claim really is. This helps you tell a scientifically backed product from one selling mostly marketing, so you spend money (or stock your shelves) wisely.

Tools you need

• Claude (Opus 4.5) (freemium): an AI chat app. Opus 4.5 is one of Claude’s most capable models and is on paid plans, but the free tier of Claude, ChatGPT, or Gemini works too. You will paste ingredients and research text and ask it to analyze them.

Steps

1. Open the AI chat app: go to claude.ai (or chatgpt.com or gemini.google.com) in your browser and sign in. You should see a message box ready for your first prompt.

2. Gather your inputs: copy the ingredient list off the supplement’s label (or its online product page). If you have access to any research, such as a paper’s abstract (the short summary at the top of a study), keep that handy too. You do not need to be a scientist; you just need the text.

3. Ask the AI to analyze the ingredients: paste the list into this prompt:

   Analyze the following supplement ingredients: [paste the ingredient list here].
   For each ingredient, give me: its main claimed benefit, how strong the
   scientific evidence for that claim is (strong, mixed, weak, or none), and any
   known side effects or interactions. Flag any ingredient where the evidence is
   weak or the dose is too low to matter.

   You should get a per-ingredient breakdown that separates well-supported claims from shaky ones.

4. Rank the claims by strength of evidence (optional): if you have a research paper’s abstract, paste it in to have the AI weigh it:

   Here is a research abstract: [paste the abstract here]. Rank the claims it makes
   by effect size (how big the benefit is) and quality of evidence (was it a large
   human trial, a small one, or just animal or lab work?). Tell me which claims are
   the most trustworthy.

   You should get a ranked list, with large, well-tested effects at the top.

5. Cross-check, then decide: compare the two answers. Does the research actually back up the benefits claimed on the label? Ask the AI directly: “Based on the research above, which of this product’s claims hold up and which do not?” Treat the result as a strong starting point, not gospel, and verify anything important against the original study.

Original source

Based on a Hacker News post by lilouartz describing a multi-year effort to analyze supplements by indexing products alongside scientific research and using AI to connect the two. The author discussed the challenges involved, including legal questions and the changing capabilities of AI models like Opus 4.5.

Notes & variations

• Free-tier alternative: no need for a paid plan. Claude’s free tier, ChatGPT, or Gemini can all do this. Just be extra careful to fact-check, since any model can “hallucinate” (state made-up information as if it were true).
• Common pitfall: trusting the AI blindly on a health topic. Always check its summaries against the original research, and never treat this as medical advice.
• Tip for better results: be specific about the evidence you want. Ask for “claims proven in human clinical trials” rather than just “benefits”, so the AI does not lump strong and weak evidence together.