Corrections
If you have found an error
Tell us what is wrong and where. A reference nobody can correct is not a reference, and structured data makes errors easier to find and easier to fix than prose does.
What we treat as an error
Some of these will be more useful to us than you might expect.
- A factual claim that is wrong — a botanical name, an origin, a production detail, a historical statement.
- A claim graded more confidently than its evidence supports. Something published as established fact that is really a convention is an error even if the underlying statement is defensible.
- A source that does not support the claim it is attached to, or is being used outside what it is competent to say.
- A cultural or historical description that flattens, misattributes or overstates. These matter and are easy for us to get wrong.
- A substitution or pairing presented without the context where it fails.
- A missing “we do not know”. Silence where a gap should be declared is a defect here.
What we will do
What we commit to, rather than what we aspire to.
- Assess the report against the source, not against how confident the current text sounds.
- Where a claim cannot be supported, remove or downgrade it. Downgrading to a lower grade, or to an explicit unknown, is a normal outcome and not a failure.
- Where sources genuinely disagree, record the disagreement rather than picking a winner. Several claims on the site are already published as disputed for this reason.
- Record the review date on the affected record, so a reader can see when it was last looked at.
What we will not do is quietly change a claim and leave the page looking as though it always said that. Records carry their evidence and their review dates, and both move when a correction lands.
Report a correction
Fill this in and it composes a structured report. It takes a minute and it is genuinely useful.
A source that supports the correction is the difference between a report we can act on and one we cannot.
There is no address to send this to yet. SpiceHQ is not public, so no monitored inbox exists. Copy the report and keep it — we would rather tell you that than print an address nobody reads.
Errors we already know about
Gaps are published rather than hidden.
Incompleteness is measured and published. Profiles show which fields are explicitly unknown, and the coverage report shows how much of modelled search demand the site can actually answer — including the families where the answer is very little. Neither figure is presented more favourably than it deserves.