Cite Where the Number Is

01Footnotes were built for readers who turn pages

The reference list at the end of an article is a print habit. It assumed a reader who reads from top to bottom and can flip to the back when a number looks surprising. Very few people read a web article that way, and the machines that now quote articles never do. They take a passage, sometimes a single sentence, and repeat it somewhere else. When the source sits in a numbered list four screens below, the passage they take does not contain it. A number without its source in the same sentence is a rumour waiting to be quoted.

How quoting actually works

Retrieval systems split a page into passages before they match anything against a question, and they hand the model the passages that matched, not the page. A claim near the top and a reference list at the bottom usually end up in different passages. The research I wrote about in the first thirty percent found that 44.2% of citations come from the opening third of a page, which is exactly where the claims tend to be and exactly where the reference list is not. The list does its job for a careful human and almost no job for anything that quotes.

02Detached numbers get new owners

I have watched it happen

Earlier this month I found four posts on another site that credited our agency with a 23% figure and a 41% figure, linked to one of our articles, which contains neither. I wrote it up in quoted with numbers I never wrote. I do not know how those paragraphs were written, but I know the shape of the failure: a number separated from the sentence that said where it came from, then reattached to the nearest name. When a source lives far from its number, the number gets adopted by whoever is closest.

Attribution is part of the fact

A number on its own is half a fact. Who measured it, how, and when are the other half, and they change what the number means. A share of a survey of 1,500 hiring managers is not the same claim as the same share of a dozen replies on a forum. Put the measurer and the date in the sentence and the claim becomes checkable wherever it lands. Every sourced sentence should survive being copied out of the page alone. That is the test I apply before publishing anything: lift the sentence, read it with nothing around it, and see whether a stranger could check it.

03How I source now

In the sentence, then beside it

The number, the measurer and the date go in one sentence: a 2026 survey by a named firm of so many people found this. When the page has a design that allows it, a small source tag sits beside the number as well, so a reader can check it in one tap. First party numbers carry the method and the window instead: six months of Bing data, fourteen days of logs, read on a stated date. The habit grew out of the rule in why I publish claims with numbers, which said adjectives cannot travel. Numbers can, but only with their passport attached. A number travels; its credibility only travels if the source goes with it.

No list at the end

I no longer finish articles with a references section. It adds length that readers skip, it duplicates sources that are already in the sentences, and it creates a second place for a citation to drift out of date. If a page needs a machine readable list of sources, structured data can carry it without a visible block. The visible page should make each claim stand on its own.

The fair objection is academic practice, where reference lists exist for good reasons: precise editions, page numbers, long chains of prior work. Those are real needs, and a research paper should keep its references. A web article answering a question is a different object. Its readers and its quoters want the claim and its origin together, in one breath.

Tomorrow: back to the record, with the article that earns everything.