How we read evidence and review products.

• Human clinical evidence — Controlled trials in people, for this specific claim.
• Early human evidence — Small, short, or uncontrolled human studies.
• Preclinical — Cell or animal studies only.
• Mixed / insufficient — Conflicting results or too little to judge.

Guides that explain categories, methods or documents, rather than summarizing research on a claim, carry the label Literacy guide · no evidence rating.

A rating always combines three things: a claim (for example, “visible signs of skin aging”), a product form and route (for example, a topical serum), and the level of evidence for that pairing.

• We rate each claim and form separately. A guide can show a higher level for one claim and a lower level for another.
• Evidence is never moved from one form or route to another. A trial of an oral supplement does not support a cream, and an injected drug trial does not support a cosmetic with a similar name.
• Results from cells, animals and people are kept apart and labeled in evidence tables. A strong cell result does not upgrade weak human evidence.
• A study of one finished product tells us about that product. We do not apply it to other products that share an ingredient.
• When the evidence could fit two levels, we choose the lower one.
• The main evidence level shown at the top of a guide reflects its central claim; the evidence snapshot and table give the detail behind it.

Ratings describe how much research supports a claim. They are not safety ratings, and they are not recommendations to use anything.

• We prefer primary sources: PubMed records and journal articles, clinical trial registry entries, regulator documents, and official product labels.
• Every source is opened and checked against the specific claim it supports before it is cited. PubMed IDs, DOIs and registry numbers are copied from the source itself.
• Each source is listed with its type (for example, journal article, trial registry record or regulator document) and, for studies, its design.
• When we learn that a cited study has been corrected, flagged with an expression of concern or retracted, we mark the source and re-check the claims that rely on it.
• Sources are numbered in the text in the order they are first cited, and listed in full at the end of each page.

AI tools help find and summarize candidate sources and draft text. They are never treated as a source. See the editorial policy.

Regulatory status is stated for a specific product, form, route and country, together with the date we checked it. A cosmetic's status says nothing about an injected or oral product that uses a similar name.

When we report a regulatory change, we keep three dates apart where they exist: when something happened, when it was published, and when it takes effect. Every guide also shows the date it was last checked against its sources.

• Guides in Therapeutics Explained and Research Peptides are scheduled for a re-check every 30 days, because regulatory status and research in those areas change quickly.
• All other guides and reviews are scheduled for a re-check every 90 days.
• We re-check sooner when we learn of a relevant regulatory action, a new study or a correction.

A new date means something was actually re-checked. Each page keeps an update log, and each entry has one of four labels: Checked (a routine check found nothing to change), Updated (content was added or revised), Corrected (an error was fixed) or Status re-checked (regulatory status was verified again). Entries for guides, reviews and comparisons are also collected in research updates.

Our reviews are documentation reviews: we check labels, ingredient lists, testing documents and supplier policies, and record what we could and could not confirm. We did not independently analyze product samples. Findings are shown criterion by criterion, with no star ratings or overall scores, and “Not provided” is never counted as a zero.

The full method, including what we check and what we do not do, is in How We Evaluate Products and Suppliers.

Our research price pages record what suppliers of research materials advertise on their own websites. They are not an endorsement of any supplier or product, and nothing on them is advice about personal use.

• How prices are collected. Once a week an automated check reads each listed supplier’s public catalog — the store’s own product feed where it has one, otherwise its product pages. It identifies itself as PeptideAtlasBot, follows each site’s robots.txt rules and makes no more than one request per second to a site.
• Which suppliers are listed. An editor first checks that the supplier has an official HTTPS website, publishes a laboratory report for each lot, names the testing laboratory and sells under research-use terms. We also record whether its product pages include use or preparation wording; where they do, its listings say so. No supplier pays to be listed, and we may choose not to list a supplier that passes these checks.
• Matching a listing to a substance. A listing is shown only after its contents — which substance, how much, in what form and how many units — have been matched to our catalog and approved by an editor. An AI model (Claude, by Anthropic) suggests matches; it never approves them. Listings we cannot match with confidence are not shown.
• What “current” means. Every price shows when we last saw it. A price we have not seen again within 8 days moves to Not current and cannot be shown as the lowest price. Out-of-stock listings and listings whose quantity we could not confirm are shown in separate groups and left out of the ranking.
• Changes we hold back. If a price changes by more than half, its currency changes or the listing’s title changes, the new value waits for an editor, and the previous price stays on the page until then. It still moves to Not current after 8 days.
• How listings are ordered. Single-substance vials and powders are ordered by price per milligram (or per IU where a substance is labelled in IU); blends, capsules and other forms are compared by package price. Ties are broken by supplier name. Commission never changes the order, and suppliers that pay us nothing are listed in exactly the same way. Prices exclude shipping, taxes and discount codes.
• Laboratory reports. Where a supplier publishes a laboratory report, we record which of seven checks it covers: identity by mass spectrometry, purity by HPLC, net content, endotoxin, heavy metals, sterility and testing across several vials. An AI model reads the report and an editor approves the summary. Whether the issuing laboratory has confirmed a report is recorded separately, and only once we have checked. Reports are vendor-reported and not independently authenticated.

What these pages cannot tell you: a matching name or a laboratory report does not verify the contents of the lot you would receive, and a low price says nothing about quality. We do not buy or test products ourselves. How supplier links work, and when a link would be paid, is set out in our affiliate disclosure.

If you find an error, tell us. We check every report against the sources. Material corrections are made on the page and recorded in its update log with the date. The corrections page explains the process.