How we edit, cite, and review

Why publish this

Most affiliate-funded health content signals quality through tone and typography. BiologicalX signals it through deterministic, automated gates in the build pipeline. This page makes those gates visible. If you see an article on our site, it passed every rule below at the moment it was committed.

We are affiliate-monetized and vendor-agnostic. Our editorial output is not influenced by vendor relationships. We do not sell first-party products, do not accept sponsorships for individual articles, and do not publish guest posts. Vendor affiliate URLs are labelled as such inline and surfaced with standard FTC disclosure on every page that carries them.

Evidence tiers

Every article declares one of four tiers in its frontmatter. Tier assignment is editorial judgement but is accompanied by a required justification sentence that cites the tier driver. The tier also sets the freshness review window: weaker evidence reviewed sooner, since a single new trial can shift the verdict on a preliminary article more than it can shift one with a robust base.

robust (180-day review)

Multiple large RCTs or meta-analyses with consistent direction of effect. Clinical recommendations acceptable with standard medical-advice disclaimer.

moderate (120-day review)

At least one well-powered trial plus supporting cohort or mechanistic evidence. Direction of effect is plausible; magnitude is uncertain.

preliminary (60-day review)

Small human trials, open-label studies, or strong animal data. Effect direction plausible; claims remain provisional.

insufficient (30-day review)

Mechanistic or anecdotal evidence only. Article exists to catalogue what is claimed and what is not yet known.

Required structure per article

1. Thesis in frontmatter, under 200 characters. Rendered below the H1 in italics.
2. TL;DR Callout at the top of the body. Flagged to AI voice assistants + AI Overviews via SpeakableSpecification.
3. Verdict block near the end. Equivalent of "the bottom line": three to five sentences, practical takeaway.
4. At least two StudyCite references inline. Each resolves against src/data/studies.ts; build fails if the registry key is unknown.
5. Numeric density of at least three quantified claims per 500 words. Ballast against vague prose.
6. Banned-phrase regex lint. AI-tell phrases (the canonical "it is important to note"), unqualified marketing shortcuts (assertions of clinical proof, blanket FDA-approval claims), and hedge language are rejected at commit.
7. No em-dashes in source or rendered HTML. House rule. Smartypants is disabled globally to prevent silent conversion.

Citation integrity

Primary-source citations live in a single typed registry (src/data/studies.ts), currently 98 entries and growing with every article. Every entry carries at least one of: DOI, ClinicalTrials.gov NCT identifier, or permanent URL. Three gate tiers enforce integrity:

• Static (every commit): every StudyCite resolves to a registry entry; every registry entry has a resolvable locator; orphan entries emit warnings. Build fails on a missing key.
• Online (weekly): HEAD-check every DOI, NCT, and URL for 200 response. Broken locators are queued for manual reroute or removal.
• Semantic (monthly): a Claude pass reads every article and flags quantified clinical claims that lack an inline primary-source citation. Flagged claims enter an editorial queue and the article is re-edited or annotated.

Source policy

Primary sources only: peer-reviewed journals, registered trials with named principal investigators, regulatory dossiers (FDA SBA, EMA EPAR, MHRA), and named position papers from established societies (ISSN, AHA, ESC, ADA). We do not cite anonymous blog posts, supplement-brand whitepapers, or single-tweet claims as evidence. Where we use a clinician interview or expert quote, the source is named and the affiliation disclosed. Anonymous sourcing is reserved for the rare case where a person flags a safety issue and asks for off-record protection; in that case, we publish the underlying primary data and note that an anonymous tipster directed us to it.

Tier-specific freshness review

Every article carries a reviewedAt date, defaulting to its publication date. Robust articles are reviewed every 180 days, moderate every 120, preliminary every 60, and insufficient every 30. Articles past their tier window are flagged visibly in the header and surfaced at our internal research index.

The review workflow is: re-read against the current literature, either rewrite the affected sections or bump reviewedAt. No silent republishing. If a referenced trial has updated endpoints, the review is the trigger to update. The shorter window for weaker tiers reflects that a single new readout can shift the verdict on a preliminary article more than on one anchored in dozens of consistent trials.

Numeric density

The prose linter requires at least three quantified claims per 500 words of body copy. A quantified claim is a number with a unit and a referent: a dose, an effect size, a percent change, a sample size, a half-life, a participant count. The reason is operational: vague prose hides the absence of evidence, but a paragraph that has to land a number can only do so by referencing a specific trial. The rule is enforced at commit time and gates the build.

Accessibility

Every indexable page passes axe-core WCAG 2AA on every commit. Zero serious or critical violations is the gate; minor violations are tracked as tech debt. The skip-to-content link is keyboard-reachable on every page, color-contrast ratios meet AA across the dark theme, hero images carry required alt text enforced at the schema layer, and form controls on the calculator pages are labelled with both visible text and accessible names. The full ruleset is documented at docs/A11Y.md in the repository.

What we will not do

The shape of a content business signals as much as its claims. The list below is what we have decided not to build, and why.

• No first-party stack or branded supplement. Affiliate-only monetization removes the largest editorial conflict in this category. We have no SKU to push.
• No premium tier or paywalled protocols. Every article and tool is free and ad-free.
• No podcast, no video. The format rewards charisma over rigor. We are not entering it.
• No inflated word counts. Articles end when they end. We do not pad to a 2,500-word SEO target. Average article is shorter than the category norm by design.
• No fictional research team or invented credentials. Articles are published under the BiologicalX Research byline, not under made-up doctor avatars. The team is small and accountable; the byline is the organization.
• No cult-of-personality framing. No founder-as-protagonist, no "what I take" stack pages designed to sell vibes. The reader is the protagonist.
• No sponsored articles or guest posts. Vendor relationships are limited to affiliate links labelled inline. Sponsorships introduce article-level bias we cannot audit away.

Disclosure and independence

BiologicalX is exclusively affiliate-monetized. We earn commission when a reader clicks through to a vendor and purchases. Commission does not alter the editorial tier, trust score, or inclusion decision for a vendor. We do not take sponsorships, advisory fees, or gifted product credits from any vendor we rank or any compound manufacturer we cover.

Vendor trust scores are published with a full breakdown at the vendor index. Affiliate links are disclosed with an inline rel="sponsored noopener" attribute and surfaced via the AffiliateDisclosure component on every page that contains them. The canonical FTC disclosure lives at /legal/disclosure/.

Entity grounding for AI search

Every compound in our directory carries a Wikidata Q-identifier and, where applicable, a PubChem CID and CAS registry number. This grounding lets AI search (ChatGPT, Perplexity, Claude, Bing Copilot, Gemini) resolve our content to a canonical entity, which materially improves citation accuracy. We also publish an llms.txt, llms-full.txt, and llms.json following the llmstxt.org spec.

Not medical advice

Content on BiologicalX is educational and reflects editorial interpretation of the cited evidence base. It is not medical advice. It is not a substitute for consultation with a qualified clinician. Always discuss starting, stopping, or changing any intervention with your physician, especially for Rx medications, compounds with narrow therapeutic indices, and during pregnancy or lactation.