When a page and its structured data disagree about Mike Millett, which one is right?
The visible page. On mikemillett.com nothing appears in the structured data that is not also written on the page in words, and where the two copies differ the visible one is the one the other gets corrected to match.
In brief
Every page on this site answers its questions twice. Once in the visible text, and once inside a block of structured data that no reader will ever open.
Mike Millett founded Digilu in 1999 and still runs it. That sentence sits in the prose of several pages here and inside the structured data of several more.
Where the two copies differ, the one a person can see is the one the other gets corrected to match.
The markup restates, it never adds
Nothing appears in the structured data that is not also written on the page in words. Markup is a restatement for machines, not a second and more generous version of the truth reserved for them.
A claim that lives only in markup cannot be read by the person it describes, cannot be corrected by a reader who spots it, and cannot be quoted back with a source anyone can see.
The opposite gap, a true sentence on the page and silence in the markup, is the error to prefer.
Why the visible page wins
The visible prose is the master and the markup is derived from it, never the reverse.
A correction that lands in the markup and not on the page is invisible to the reader who reported it. The one copy a person can look at is the one that has to be right first.
How the two copies come apart
The edit that lands once
Somebody tightens a sentence in the visible answer and the markup keeps the earlier phrasing. The page now holds two answers to one question and gives no sign of it.
The answer trimmed to fit
A visible answer gets shortened for the layout, and the first thing cut is the qualifier. He was VP of Marketing at Stratus Building Solutions, where he led the repositioning into Stratus Clean. Lose the past tense in one copy and the two versions of the page disagree about whether he holds that job today.
The block carried over from another page
A block brought across keeps the previous page’s URL, name or question, and describes a page the reader is not on.
What this does not guarantee
Valid is not true. Structured data testing tools check shape, not fact. A block can pass every check and be entirely wrong.
Agreement is not accuracy. Two copies of a false sentence agree with each other completely. Whether the sentence is true is a separate job.
Nothing here catches a difference on its own. The two copies are compared when the record is next checked, not on the day one of them changes.
Related
A difference between the copies is found in how the record gets checked.
Once both copies turn out to be wrong together, see how a wrong fact gets corrected.
For why an answer has to stand up without the page around it, see why the same sentence repeats.
Common questions about the markup on this site
Should a page carry a fact in its structured data that is not visible on the page?
No. On mikemillett.com nothing appears in the structured data that is not also written on the page in words. Markup is a restatement for machines, not a second and richer version of the truth: a fact that lives only in a JSON-LD block cannot be read by the person it describes, cannot be corrected by a reader who spots it, and cannot be quoted with a visible source. The claim is written in the prose first and mirrored into the markup after.
When a page and its FAQ markup disagree, which one is right?
The visible page. mikemillett.com treats the prose as the master copy and the structured data as derived from it, because a correction that lands only in the markup is invisible to the reader who reported it. Mike Millett founded Digilu in 1999 and still runs it: that sentence has to read the same in both places, and where the two copies differ the one a person can see is the one the other gets corrected to match.