How CleaningHQ works
CleaningHQ is a structured body of cleaning knowledge rather than a collection of articles. This page explains how it is built, what its verdicts mean, and where it refuses to answer.
It is a graph, not a set of articles
Most cleaning advice is written page by page: somebody writes “how to clean a marble worktop”, and the knowledge inside it is trapped there. CleaningHQ instead records facts separately and connects them — that marble is calcium carbonate, that acids dissolve carbonates, that citric acid is an acid, that limescale is also calcium carbonate.
Four facts, written by different people for different reasons. Together they answer a question nobody wrote a page for: whether the descaler under your sink is safe on your worktop. Today that is 291 records connected by 483 relationships. The connections are the product; the pages are a view of them.
Two questions, never collapsed into one word
Almost every mistake a cleaning publication makes comes from answering “is it good?”. CleaningHQ splits that permanently in two, because the answers routinely disagree:
- Does it work? Recorded as an evidence level, from strongly supported down through established practitioner practice and plausible mechanism to disputed, unresolved and does not work.
- Is it safe here? Recorded separately, per material, as a safety state.
Vinegar descales a kettle and etches a marble worktop. Same substance, same mechanism, opposite answers — and any model with a single “rating” has to be wrong about one of them. There is no numeric confidence score anywhere in CleaningHQ and none may be added. A reader can act on “established practitioner practice, mechanism plausible, not experimentally isolated”. Nobody can act on “0.64”.
Sources are ranked, and read
Sources sit in tiers: regulators and poison authorities, standards bodies, scientific literature, manufacturers, trade and professional bodies, then reference works and credible secondary material. A safety claim resting only on the lower tiers is not published with a caveat — it is not published.
The distinction that matters most is between a source that has been cited and one that has been read. Every source here records which it is, and 106 of 116 have had the specific passage read by a person. The rest are sold rather than published, behind subscriptions, or served by hosts that refuse automated retrieval, and they say so.
Reading changes things, which is the point of doing it. It has widened claims, narrowed them, corrected an attribution, and in one case withdrawn a statement entirely. The source register records each of those.
A clearance costs more than a warning, deliberately
Telling you something is dangerous requires one authoritative source. Telling you something is safe requires an authoritative source somebody has actually opened, plus the conditions the clearance holds within.
The asymmetry is not fussiness. An over-cautious warning costs you an unnecessary precaution; a wrong clearance costs you the worktop, or the injury. So there is no plain “safe” in this model at all — the strongest available answer is “safe within these stated conditions”.
Manufacturer instructions outrank us
Where your item has instructions, those instructions govern. CleaningHQ records manufacturer positions as their own kind of evidence and never generalises one manufacturer’s guidance to another’s product. Two stove makers give different answers about the same technique; both are right about their own appliance, and the record says so rather than picking one.
Where we stop
CleaningHQ does not infer. If no record exists for a combination, the answer is “we have not established this” — not a cautious yes, not silence, and never a guess assembled from adjacent chemistry. That refusal is enforced by the type system rather than by editorial discipline, because discipline erodes and types do not.
How the safety model works · How the writing is governed · How hacks are classified