How Selpho works,
in full.

1. How we curate the peptide library

The library is hand-curated, not scraped. A compound is only included if it meets all of the following:

• It is real and documented. There is published literature — human or animal — that we can point you to. We do not list compounds that exist only in forum folklore.
• Its mechanism can be described. If we cannot explain in plain English what a peptide is believed to do and why, it does not go in.
• Its research stage can be stated honestly. Every entry carries a research-status label written by a human, ranging from preclinical / animal studies only through investigational, in clinical trials to FDA-approved. We do not round that up.
• Its caveats are written down. Every entry has a considerations field covering what the research does not show, populations who should avoid it, or endpoints a trial failed to meet.

Blended products are listed by their components, so you can research each part separately rather than take a marketing name at face value. Every peptide page links out to a live PubMed search for that compound — the point is for you to read the primary literature, not to take our summary as the last word.

2. How the AI advisor works

The Peptide Advisor runs on a large language model (Google Gemini) that we constrain heavily rather than let answer freely. On every single request:

• Our curated catalog is injected into the prompt. The model receives the full list of the peptides we cover, with their categories and researched-for uses, and is instructed to prefer those entries and to use their exact canonical names.
• Inventing compounds is forbidden. The model may only name a peptide outside our catalog when nothing in it fits, and only if that peptide is itself well-documented. It is explicitly instructed never to fabricate one.
• Off-topic questions are declined, not answered. If you ask about something that is not a peptide or a health goal peptides are researched for, the advisor says so rather than improvising.
• The output shape is fixed. Responses come back in a defined structure — relevance, commonly researched for, typically stacked with, research status, considerations — so a suggestion cannot arrive without its caveats attached.

To be precise about what this is: the advisor is a constrained language model working from our curated catalog — it is not a live search of PubMed, and it does not retrieve or quote individual studies. The research-status labels and considerations you see on peptide pages are written by us, not generated per-request. The PubMed links we give you are searches, so you can go read the actual papers.

3. What the AI can get wrong

We would rather say this plainly than let you find out on your own. Language models produce fluent text whether or not they are correct, and the constraints above reduce that risk without eliminating it. Specifically:

• A suggestion may be plausible but poorly matched to your actual situation, because the advisor only knows what you typed into a text box.
• It has no knowledge of your medical history, medications, or interactions. It cannot know what would be unsafe for you specifically.
• Its sense of how strong the evidence is can be softer than the literature warrants. When our human-written research-status label and the model's phrasing disagree, trust the label.
• Peptide research moves. A page reviewed months ago may not reflect a trial that reported last week.

This is why every result ends by pointing you to a clinician, and why we describe Selpho as preparation for a medical conversation rather than a substitute for one.

4. What Selpho will never do

These are enforced in the system prompt itself, not left to the model's discretion, and they apply to the free plan, Pro, the provider portal, and the developer API alike.

5. How we handle safety

Selpho is an educational research tool. It is not a medical device, it is not regulated as one, and it does not practice medicine. In line with that:

• Every advisor result carries a disclaimer and a prompt to consult a licensed healthcare professional.
• Considerations and contraindications from the research travel with each suggestion rather than being buried in a footer.
• Where a compound is investigational or its trials disappointed, the research-status label says exactly that.
• If you use our lab tools, out-of-range markers are surfaced as things to raise with a clinician — never as a diagnosis or a treatment plan.

6. How our comparison and goal pages are built

Our peptide comparison pages are generated programmatically from the same curated library — but only for pairs within the same therapeutic category. Comparing two compounds researched for unrelated purposes would produce a page that ranks but does not inform anyone, so we do not generate those. Goal pages and guides draw on the same curated data, which is why a peptide's research status reads identically wherever you encounter it.

7. How we make money

Plainly, because incentives are the thing you should interrogate hardest on a site like this:

• Subscriptions. Selpho Pro is $4.95/month ($1.99 for your first week) or $49.50/year. There is a free plan that does not expire.
• Developer API credits. Businesses building on our API and widgets buy credits.
• Practitioner accounts. Clinicians pay for the provider portal to review their own patients' results.

That is the complete list. We do not sell peptides, we take no commission or affiliate revenue from anyone who does, and no vendor can pay to appear in the library or to be suggested by the advisor. A peptide is in our catalog because the research justifies covering it, or it is not in it at all. If we ever add another revenue source, it will be listed here before it launches.

8. Corrections

We will get things wrong. If you find an inaccurate claim, a research-status label that no longer matches the literature, or a study we have mischaracterised, tell us and we will fix it and update the review date on this page. Send us a correction →

Want to know who is behind this? Read about Selpho and its founder.