Cite, or don't claim.

Most public peptide writing is marketing dressed as science. The rule here is the opposite: evidence over opinion, citation over claim, published over promised — and where the evidence isn't in yet, we say so rather than inventing a number.

1 · The Evidence Maturity Profile

Instead of a single letter grade, every compound is profiled across seven independent axes. A single grade averages strong and weak evidence into one number that hides the detail. Seven axes show where a compound is actually backed — and where it isn't.

Literature VolumecomputedHow much candidate research exists, retrieved by compound name. A count, not a quality claim.
Human EvidenceeditorialStrength of human clinical evidence. Requires an editor to confirm each paper actually studies the compound in people.
Preclinical EvidenceeditorialDepth of animal, cell, and mechanistic work. Requires study-level classification (PubMed tags don't flag it).
Mechanism ConfidenceeditorialHow well the mechanism is characterized — hypothesized vs. receptor-validated.
Safety ClarityeditorialHow well short- and long-term safety is characterized in the literature.
Regulatory ClarityeditorialApproval and scheduling status. The factual status badge is shown immediately; the graded score is editorial.
Practical RelevanceeditorialWhether observed effects are meaningful in practice — a judgment, not a citation count.
Why the overall score can be withheld
A maturity score is shown only once every editorial axis is reviewed. A half-computed profile must never imply a finished grade — an empty axis is shown as pending, not a zero.

2 · From axes to an S–F tier

The public tier comes from a weighted 0–100 score across eight dimensions: the six evidence axes above plus Evidence Consistency (do studies agree?) and Claim Risk (how far do popular claims outrun the data?). Human evidence carries the most weight.

S90–100

Approved or trial-grade human evidence

A78–89

Strong, consistent human data

B65–77

Real human evidence, gaps remain

C50–64

Mostly preclinical / early human

D35–49

Thin — animal or anecdote-led

F0–34

Little credible evidence

Claim Risk is a penalty, not a boost
Heavy hype on thin evidence lowers a tier — it never raises one. Popularity is shown separately in a hype-vs-evidence panel, so a loud compound with little data lands where the data puts it.

3 · Three honesty states

Every number on the site carries one of three states, so you always know how solid it is:

Computed

Derived deterministically from the real corpus (e.g. Literature Volume from indexed source counts).

Editorial (data-grounded)

A first-pass score grounded in documented public facts — FDA approval, registered-trial activity, and the established state of the evidence — attributed to a reviewer and revisable. Not invented figures.

Pending review

Specific claim verdicts and other judgments not yet authored are shown as pending — never guessed, never a placeholder zero.

4 · Why we don't auto-grade evidence quality

The citation corpus is built by searching PubMed for each compound by name. That retrieval is broad on purpose — but a returned paper may only mention a compound, or match its name by coincidence. So a raw count of trials in the corpus is not proof a compound was tested in humans.

A name-matched PubMed count shows
  • How much literature mentions the compound
  • Rough research activity and momentum
  • A starting list for an editor to review
It does NOT show
  • That those papers actually study the compound
  • That any of it was done in humans
  • That the results were positive — or replicated
A real failure mode we designed against
An early build briefly showed a popular research peptide as “clinically studied” because one trial of an unrelated drug matched its name search. That is exactly the inference we refuse to make. Evidence-quality axes stay pending until an editor confirms each paper genuinely studies the compound — machine-counted publication types start that review, they never replace it.

5 · The source hierarchy

Not all citations are equal. Every source is graded A–F on its own merits and slotted on this ladder — strongest at the top:

  • Meta-analyses & systematic reviews. The strongest synthesis of human evidence.
  • Randomized controlled trials. Phase 3 down to phase 1 — the closest thing to causal human evidence.
  • Observational human studies. Cohorts and case-control work. Real people, weaker causal claims.
  • Animal & mechanistic studies. Useful for plausibility, never treated as human evidence.
  • In-vitro & case reports. Hypothesis-generating only.
  • Regulatory documents. FDA labels, EMA and WADA records — authoritative for status, not efficacy.
  • Press releases. Market context only, always labeled. Never evidence.

6 · How citations stay clean (three layers)

“Citation-backed” is only as good as what stops a bad citation from landing. A source has to beat three independent checks:

  1. 1
    At save
    When an editor adds a source, its identifier is re-resolved live against NCBI, Crossref, or ClinicalTrials.gov. The editor must write an original summary and assign a quality grade before it can be marked verified.
  2. 2
    At publish
    A database gate refuses to publish a profile or claim that's missing required fields or its minimum of verified sources — and refuses to flip a source to verified without a resolvable identifier.
  3. 3
    At commit
    A pre-commit audit re-resolves every citation in the changed data files and blocks the commit if a stored title doesn't match the real paper.

Candidate citations ingested from PubMed are shown as found and labelled unverified until they clear this review. The verified-citation count shown across the site is computed from records, never set by hand.

7 · Uncertainty, conflicts & scope

The remaining rules
  • Uncertainty is shown, not smoothed. Where human and animal evidence diverge, the page says so. Animal findings are never written up as human findings; mixed claims keep their verdict and a confidence rating.
  • No pay-to-play. Vendor rankings are editorial; any affiliate relationship is disclosed on that vendor's page. User reports are anecdotes, labeled as such — never evidence.
  • Scope. Coverage spans weight loss, healing, growth hormone, skin, cognition, longevity, libido, and immune categories. Many compounds are not approved for human use.
What this is — and isn't
A research reference. Not medical advice, not dosing guidance, not a purchase recommendation.
Keep reading
The tier board

See the scoring model applied across all compounds.

Claim checks

Specific claims graded against the evidence.

Safety framework

How risk is categorised and communicated.

Glossary

Every technical term used in the model.

Editorial policy

Independence, conflicts, and corrections.

The all-in-one peptide app

Stop reading, start tracking.

PepCue logs your doses, runs the vial math, counts your vials, and keeps the whole protocol in one place. It replaces the spreadsheet, the calculator, and the sticky notes.

  • Dose logging
  • Reconstitution math
  • Smart reminders
  • Vial & cost tracking
iPhone · free to start