Why celebrity children facts lose their source

A fact is a claim plus its basis. Remove the basis and what is left is a value: a word in a box, true or false with no way to tell which.

Almost everything that goes wrong with celebrity children facts happens at that moment of removal, and it does not happen in prose. It happens when a sentence is turned into a field.

What to take away

Two checks, habitually merged

"We verified it" describes two operations that share no method. One asks whether you have reached a primary source; the other asks whether the thing it records is so.

An attribution check asks whether the words were really said, by the person credited, on the date given. It is done against the original: the account, the transcript, the filing. It can succeed completely while the underlying statement is wrong.

A substance check asks whether the thing asserted is so. It needs evidence independent of the person asserting it, which for a private family matter usually does not exist and should not be sought.

For most claims about a public figure's family, only the first check is available. That is not a failure, provided the page says which check it ran. "She said this in a published interview" is an attribution finding and an honest sentence. "She has three children" presents a substance finding you did not make.

Write the first one, and the reader knows exactly what you are standing on. The chain of repetition that makes the second one feel safe is dismantled in the guide to working a claim backwards.

What the documents actually show

Take four celebrity children whose names are on the public record and ask what each document class can prove.

None of these rows is a substance check. Each is a note about which piece of paper exists, where it was filed, and what it does and does not say.

A birth announcement tells you a publicist spoke. A civil registration record tells you a jurisdiction accepted a filing, and which office holds that filing depends on where the birth happened: the New York City Department of Health and Mental Hygiene's Bureau of Vital Records for a Manhattan birth, the California Department of Public Health for one in Los Angeles, Namibia's Ministry of Home Affairs and Immigration for a birth registered there, the General Register Office for England and Wales. A court filing tells you a lawyer wrote something down. Those are three different claims.

What happens when a claim becomes a field

Databases, tables, article summaries, and the tagging systems underneath a publishing platform all force a claim into a shape. Three things happen in the process, none of them visible: the qualifier is dropped, the sentence that carried the basis is cut away, and the value is left standing on its own as a bare name or date.

The result reads as more certain than the sentence it came from, which is the opposite of what a format should do to evidence. It is the same degradation that quietly rewrites an announced spelling into a familiar one, operating on relationships rather than on letters.

The box with no author

Extraction is now automatic. Summaries, answer panels, and assistant responses lift values out of pages and present them without the sentence around them.

Two consequences follow for anyone writing in this area. A hedge you placed carefully will not appear beside the extracted value, so a claim you framed as uncertain is displayed as settled. And the presentation carries no author, which strips the reader's main tool for weighing it.

The loop then closes. Your page feeds the box, the box informs the next writer, the next writer cites the box. A claim that began as one deleted post is now confirmed by three independent-looking places, all of which are you.

The practical defense is unglamorous. Do not put a claim in a summary line, a table cell, or a tag that you are not prepared to see quoted alone, stripped of everything qualifying it, next to somebody's name.

The fact that is really an inference

A large share of what circulates as fact in this subject was calculated rather than observed, and the calculation is never shown. An age derived from a birth year, and a birth year derived from an announcement date, are two steps of arithmetic on records whose own construction is described under civil registration.

An age is arithmetic on a date, and it inherits every error in that date, including the case where the date was itself a guess. A birth order is arithmetic on two dates, at least one of which is usually unavailable.

A count of children is a claim about which sense of parentage you are counting in, and about completeness, which is exactly what nobody outside a family can establish. "Only child" is the worst of them, since it is derived from absence, and absence of a mention is not evidence of absence.

Store the inputs, not the output. A page that records what was said and when lets a reader do their own arithmetic and see the uncertainty in it. A page that records the answer has hidden the working, which is also how the same class of error propagates through a chain of parentage links.

Writing a fact so it survives being taken apart

The record behind a published fact

The last row does more work than the rest combined. It is what stops the next editor treating your careful claim as authority for the larger one nearby, and it is the difference between a fact file and a list that fails on every row.

Common questions

A widely used reference entry states it. Is that a check?

It is a place the claim is written down. Open its citation. Reference entries about family details commonly cite news articles, which cite each other, and the entry inherits whatever sat at the bottom of that chain, including nothing at all.

Our page was right and the extracted box is wrong. What can we do?

Fix your end and make the correction machine-visible rather than a quiet edit, so anyone comparing versions can see something moved. Beyond that, publish fewer bare values, because a value with no sentence attached is what gets lifted.

Is a hedge enough when we cannot run a substance check?

No. A hedge protects the publisher and does not change what a reader retains. It is the first thing removed on the way into a field.

If only an attribution check is possible, report the attribution as the finding, or report nothing. Repeating an unsourced claim with a qualifier is exactly the mechanism in the guide to claims with no accountable source.

Should any fact about a child be published at all?

Rarely, and never more than the relationship itself where a parent has made it public. Everything else about a person with no public role, however well attested, is a detail that gathers with other details into something no single source ever contained.