Editorial policy, version 2.3
The rules we hold ourselves to
This page exists so that our standards are checkable rather than asserted. Every rule below is one you can hold us to in a specific, falsifiable way, and the corrections log shows the times we did not manage it.
In force since 1 June 2026. Owner: Marta Lindqvist. Next review: January 2027.
Sourcing standards
The research base on mad honey is small, geographically lopsided and mostly clinical. We say so constantly, because summaries that hide the thinness of the evidence are the main thing wrong with existing writing on this subject.
Our ordering of sources is fixed. Peer-reviewed clinical literature and toxicology outrank everything. Government agency guidance and statute text come next, and are the only acceptable source for anything on our Law pages. Ethnographic and historical scholarship carries Culture and History. Field reporting, our own interviews, named or anonymised with a stated reason, is used for practice and process, never for a health claim. Trade press and producer material is not a source; at most it is evidence that a claim is circulating.
Fixed rules
- Every health or pharmacology claim cites a primary source. A review article is acceptable only where we also name what it reviewed.
- We read what we cite. Citing a study on the strength of its abstract is a firing-level error here, not a style issue.
- Where the only evidence is a case report, the article says "case report" in the sentence, not in a footnote.
- Sample sizes and species are stated inline. "Studies show" without a number does not pass the edit.
- Where there is no evidence, the article says there is no evidence. It does not reach for a weaker source to fill the hole.
- No article cites another of our own articles as evidence for a claim. Internal links are navigation, not support.
Review cadence
Nothing publishes on one person's say-so. Every draft goes through a subject edit and a source check, and anything touching health goes to a clinical reviewer who does not write for the site. Published articles then re-enter the queue on a fixed schedule, whether or not anyone has complained about them.
| Section | Pre-publication | Re-review | Out-of-cycle trigger |
|---|---|---|---|
| Safety | Subject edit, source check, clinical review | Every 6 months | Any new clinical literature; any reader report of a factual error |
| Science | Subject edit, source check, clinical review | Every 12 months | New primary study, retraction, or correction to a cited paper |
| Law | Subject edit, statute check against primary text | Every 6 months | Statutory amendment, new agency guidance, court decision |
| Basics, Comparisons, Guides | Subject edit, source check | Every 12 months | Change in an underlying Safety or Science article |
| Culture, History | Subject edit, source check | Every 24 months | New scholarship; correction from a person or community described |
Both dates appear at the top of every article, published and last reviewed, with the reviewer named. If an article is overdue for review, the stamp says so in plain words rather than quietly showing the old date.
How claims are graded
Readers cannot be expected to weigh a Turkish cardiology case series against a rodent study against a nineteenth-century travel account. So we grade the evidence behind each significant claim and put the grade next to the claim, as a chip, in the article body. Four grades, no others.
| Grade | Test it must pass | How it is written | Worked example |
|---|---|---|---|
| Established | Supported by more than one independent peer-reviewed source, with no substantial contradicting literature, and a plausible mechanism. | Stated plainly in the indicative. No hedge. The citation carries the weight. | Grayanotoxins bind sodium channels and delay their inactivation. |
| Supported | Good evidence, limited in scale or setting. Clinical case series, or repeated animal work with a plausible read-across, stated as such. | The limitation travels with the claim in the same sentence. | Bradycardia and hypotension are the typical presentation in reported poisonings. |
| Contested | Real evidence exists on more than one side, or a single study whose method is disputed in print. | The disagreement is described, not averaged. Name what each side rests on. | How much grayanotoxin content varies by rhododendron species alone. |
| Unsupported | Widely repeated, no evidence we can find. | State the claim, state that it is unsupported, and say where it came from. | Aphrodisiac claims in marketing and folklore. |
Grades are assigned by the writer, checked at source-check stage, and revisited at every re-review. A grade moving down is treated as an update and logged, because a reader who acted on "Supported" deserves to know it is now "Contested".
Corrections policy
We correct in the open. When a factual error is confirmed, the article is fixed within two working days, a dated note is appended at the foot of the article explaining what was wrong, and a permanent row is added to the log. We do not silently edit, we do not delete an article to make an error disappear, and we do not remove old log rows.
Correction
The article said something untrue. The text is fixed, a note is appended, and the row is logged. If the claim was a safety claim, the note stays at the top of the article for 30 days.
Clarification
The fact was right but readable the wrong way. Rewritten, noted, logged. Most of these come from readers telling us how they read a sentence.
Update
New evidence, a new statute, a retraction of something we cited. The article is revised and the check date advanced. Logged like any other change. Target turnaround is two working days for anything on a Safety or Law page, five for everything else.
Conflict of interest
Some members of this team have professional ties to honey production and trading companies. We cover why we publish anyway on the About page. Here is the machinery that makes the claim mean something.
Binding rules
- Written declaration on joining, covering employment, consulting, equity, family interests and unpaid positions in anything connected to honey production, export or retail.
- Re-confirmation twice a year, every January and July, in writing, by every contributor including the clinical reviewer. Last round closed 14 July 2026.
- Seven-day rule. A new or changed interest must be reported within seven days. Failure to report is grounds for removal of a byline.
- Recusal. Anyone with a current commercial interest in a subject cannot write, edit or review that subject. Recusals are recorded in the article's revision history and are visible to readers.
- Public disclosure lines. Every contributor's disclosure appears on their author page and in the byline block of every article they write. It is not optional and not abbreviated.
- No payment for coverage, ever. We hold no advertising, no sponsorship, no affiliate accounts and no paid-placement arrangements. Offers received are logged and refused; we have refused eleven since launch.
- No gifts. Product sent to us unsolicited is not opened, not photographed and not written about. It is disposed of and the sender is told once.
- No named sellers. Because we never name a company, an undisclosed tie cannot turn into free promotion even if a rule above fails.
Use of AI
We use language models for a narrow set of tasks and not for others, and we would rather state the boundary than let readers guess it.
What we use it for
- Transcribing and translating interview audio, always checked against the recording by the reporter.
- First-pass copy-editing suggestions on a human-written draft.
- Finding candidate literature, which is then read in full by a human before it is cited.
- Generating alt text drafts, which an editor rewrites.
What we never use it for
- Writing article text. No published sentence is model-generated.
- Summarising a study we have not read.
- Producing or extending citations. Every reference is verified against the source record.
- Any image. Every photograph on this site is a real photograph, shot or licensed, credited in the caption.
If we ever change this, the change will be announced on this page with a date, and the old version will remain accessible. Policy versions are numbered for exactly this reason.
How to challenge a claim
A challenge is a specific objection to a specific sentence. It is the most useful thing a reader can send us, and it is handled on a defined track rather than in a general inbox.
| Stage | What happens | Time |
|---|---|---|
| 1. Received | An editor reads it and acknowledges by email. Nothing is auto-replied. | 2 working days |
| 2. Assessed | The section editor checks the claim against its cited source. If the claim came from us rather than a source, that is noted. | 5 working days |
| 3. Decided | One of: correction, clarification, update, regrade, or a reasoned refusal that quotes the evidence we are relying on. | 10 working days |
| 4. Published | If anything changes, the article is amended, a note is appended, and a row is added to the log. | Same day as decision |
DOIs, PubMed IDs and statute references get read first, because they are fastest to verify.
Policy history
One email a month, when there is something worth saying.
New articles, changes to any of the nine legal-status pages, and a short note on any study worth reading. If a month passes with nothing substantial, you get nothing: we would rather skip than pad. No sponsors, no seller mentions, no third-party sharing.