44% of Citations Come From the First 30% of the Page

01The top of the page does most of the work

Research into where AI assistants take their quoted sentences from keeps landing on the same shape: 44.2% of citations come from the first 30% of a page. The remaining 70% of the page, which is where most writers put their best material, shares what is left. Divide it out and the top of a page earns about 14.7% of citations for every tenth of its length, against about 8% for every tenth after that.

Why the opening gets quoted

I am not going to claim to know the internals of any model, and what can be observed is enough. A system assembling an answer wants a sentence that states the answer and can stand on its own. Opening paragraphs written for a hurried human reader tend to be exactly that: the claim, the number, the scope. Paragraph nine tends to be a qualification of something said in paragraph four, which is useless on its own. A sentence that needs the previous paragraph to make sense cannot be quoted, however true it is.

02We were taught to save the answer

Most of us learned to write the way school essays are marked: introduce the topic, build the case, arrive at a conclusion. On the web that habit became a genre, the recipe with a life story on top, the guide that explains why the question matters for four paragraphs before answering it. The conclusion ended up exactly where the quote is least likely to be taken from.

The twenty minute page

The clearest proof I have is an embarrassing one. The page that earned 1,712 citations in the mismatch story took twenty minutes to write, and its first line was the answer. The pages I laboured over for days opened with positioning. On a small information site of mine, citations from one assistant went from about 2 a day to 30 after a run of new pages that each answered one question in the title and the first two sentences. Other things changed in the same weeks, so that is correlation and not proof, but it points the same direction as the research. The answer at the top is not a summary of the page. It is the page, as far as a machine is concerned.

03An afternoon per article

Move the sentence, keep the page

The fix is not a rewrite. It is a move. Take the sentence that answers the page's question, which is usually sitting in the last third, and put it at the top. Add the number that makes it checkable, with the date it was true, the way I argued in publishing claims with numbers in them. Retitle the page as the question a person would actually type. Then read the new opening on its own, with nothing after it, and ask whether a stranger could repeat it and be right. If your first paragraph could be pasted into an answer unchanged, the rest of the page is doing its job.

The objection, and why it does not hold

The worry is always the same: if the answer is at the top, nobody will read on. That has not been my experience, and it misunderstands who reads on. The person who wanted only the answer was never going to read your fourth section. The person who stays is the one who wanted the method, the caveats, the context, and they still get all of it. Nobody ever left a page because it told them the answer too early. They leave because it did not tell them at all.

The other two numbers

Position is the cheapest lever on a page, because moving a sentence costs nothing. It is not the only one. The same research found structured data associated with roughly a 1.7 times lift, and fresh pages earning about 6 citations against 3.6 for stale ones. None of the three replaces the others. A page with the answer at the top, clean markup and a recent date is simply giving a machine less reason to look somewhere else.

If you do only one thing this week, open your three most important pages and read the first thirty percent of each as if it were all anyone would ever see. For most sites, it very nearly is.

Tomorrow: the page nobody wants to write, the one that names your competitors.