Where construction facts come from

Search for almost any construction fact and you will get a confident answer within three results. Some fraction of those answers are correct. Nothing on the page tells you which fraction.

This is not a general complaint about the internet. Construction has a specific version of the problem: the questions are technical enough that a wrong answer looks exactly like a right one, and consequential enough that the difference eventually shows up in a building.

The hierarchy, and why it is ordered this way

Sources are ranked by proximity to the thing being claimed:

1. Government bodies and public authorities [1] [2] [3] [4]
2. Building-code and standards bodies, where their material is publicly accessible [5]
3. Official research institutions [6] [7]
4. Peer-reviewed literature
5. Universities
6. Recognised architecture and construction institutes
7. Official manufacturer technical documentation — for product-specific facts only
8. Industry organisations
9. First-party company announcements — for company and product news only

The ordering is not about prestige. It is about who had to be right.

A regulator publishing a requirement has to defend it. A research institute publishing a method has to describe the method. A manufacturer publishing a datasheet has to stand behind the numbers for their product under their stated test conditions — which makes them an excellent source for exactly that, and a poor source for whether their product is the right choice.

The manufacturer datasheet is the interesting case

Item seven is where most technical writing goes wrong, and it goes wrong in both directions.

Dismissing manufacturer documentation is a mistake: for a specific product's tested performance, it is often the only primary source that exists. Nobody else has run the test.

Treating it as neutral is the opposite mistake. A datasheet reports performance under defined conditions, and the conditions are the part people skip. A thermal conductivity figure measured dry is a true figure that becomes misleading the moment the material is installed somewhere it can get damp. The number was never wrong; the reader supplied a context the test did not cover.

So a manufacturer document is cited for what it measured, under the conditions it states, and not for the conclusion the marketing copy draws from it.

What does not count

Not SEO content. Not anonymous trade blogs. Not affiliate sites. Not scraped summaries or AI-generated summaries of sources nobody checked. Not image boards.

Secondary journalism has a real role — it is often how you find out that something happened at all — but "a trade magazine reported it" is a lead, not a citation. When a secondary source is genuinely the only one available, it is cited as secondary analysis and the limitation is stated in the text rather than smoothed over.

The paywall problem is real

A large share of the standards literature that actually governs construction is not publicly readable. This is a genuine constraint on any construction publication, and there are two dishonest ways to handle it.

The first is to paraphrase a standard you have not read, usually via somebody else's summary of it. The second is to cite it in a way that implies you have.

Both produce the same failure: a reader believes a claim is backed by a standard, when it is backed by a chain of summaries whose fidelity nobody checked.

The honest handling is unglamorous. Cite the standard by number and issuing body so a reader with access can check it. Describe what is publicly documented. Say when the full text was not consulted. An article that admits a gap is more useful than one that papers over it, because the reader can see exactly where to go next.

Access dates

Where a reference records an access date, it is the date the source was actually consulted. Access dates are never approximated to look diligent.

This sounds like a trivial point. It is not: a fabricated access date is undetectable, costless, and makes every other date on the page worth slightly less. The value of a citation apparatus comes entirely from the assumption that none of it is decorative.

What this is for

None of this makes a publication right. It makes it checkable, which is a different and more achievable goal.

Every factual claim here should trace to something a reader can go and look at. Where that chain is short and public, the claim is strong. Where it is long, or private, or missing, the article should say so — and the reader gets to decide what to do with that.