The reader who arrives from an AI summary already has the answer

Someone asked an assistant about your category, read a competent three-sentence answer, and clicked your link anyway. That is not a search-result click. They already have the answer; they came for something the answer could not carry — and most pages greet them by carefully explaining the answer they just read.

The click means the summary was not enough

A visitor from a results page arrives with a question. A visitor from a summary arrives with a provisional answer and a specific doubt about it. Those are different states of mind, and they want different opening paragraphs.

The doubt is usually one of three kinds. Sometimes it is verification: who says this, and are they in a position to know. Sometimes it is particulars, because the summary said costs vary and this reader needs the range for a company of their size, in their market. Sometimes it is the exception — they suspect theirs is the case the general answer quietly excludes.

None of the three is served by an introduction that restates the general answer. By the third paragraph the reader has confirmed only that you know what everybody knows. Then they leave.

Put the particular first and the framing after

Most business writing is built as a funnel. Context, then narrowing, then finally the claim. That shape assumes a reader who needs bringing up to speed, and it was a reasonable assumption for as long as arriving cold was the normal way to arrive.

The summary visitor is not cold. They have been briefed by a machine that read a dozen pages on the subject, possibly including yours. So invert the funnel. The figure, the date, the named exception, the thing you learned running this yourself — those belong near the top, with the framing kept underneath for the reader who genuinely did need it. That order costs the ordinary visitor nothing and gives the briefed one a reason to stay.

Google's own guidance has pointed this way for years. Its self-assessment questions for content quality ask whether a page provides original information, reporting, research, or analysis, and whether it offers substantial value compared with other pages in the results. Those questions were written before assistants were summarising anything, and they have only got sharper since. If a model could assemble your page out of other pages, it already did, and your reader has read that version.

Attribution does more work than it used to

The summary took your name off. Whatever the assistant told them, it told them in its own voice — so the reader holds an answer with no author attached to it, and no particular reason to believe it. That is the gap your page exists to fill.

Say who you are and how you know, early and concretely. Not a credentials paragraph, which reads as noise. The actual basis: we run this every morning on our own blog, or we did the thing this post describes for eleven clients and here is where it broke. The same guidance asks writers to consider Who, How, and Why — who made the content, how it was produced, and why it exists. Sensible questions when the reader arrived cold. For a reader holding an unsourced answer, they are the entire reason the page is still open.

What we cannot tell you

We cannot measure this cleanly, and neither can anyone else yet. Assistant referrals reach analytics inconsistently: some pass a referrer, some pass nothing, some arrive through a redirect that launders the origin. So we cannot hand you a percentage for how much of your traffic came in pre-briefed, and we would treat a confident number from anyone else as a claim requiring working shown.

The volume is also probably smaller than the noise around it suggests. Most business research still runs through a results page, and this is not an argument for rewriting an archive.

It also cuts against something we have argued elsewhere. Writing clean, self-contained, quotable passages is how you get cited by an assistant in the first place — and it is also what makes you easy to summarise away. Both things are true at once. We would still take the citation.

The fix lives in the first hundred words

None of this asks for a different blog. It asks for a different opening.

Take a post you already have and read its first hundred words as somebody who has just heard a competent summary of the subject. If those words say nothing the summary did not, the page has not earned its click. The repair is usually not writing but moving: the specific paragraph, the one carrying the number and the caveat, already exists somewhere in the middle. It goes on top.

The machinery that produced the page does not change the test. Google's problem was never machine authorship — it was pages that added nothing to what already existed. A summary is now the thing that already exists.