Writing for buyers who start in a chat window

A buyer evaluating inventory software used to open five tabs. Now they open one, describe their problem in a sentence, and read a synthesised answer assembled from whichever sources the model found most legible. The underlying question is identical. The surface where it gets answered is not.

The page still has to be worth citing

Content that gets pulled into an AI-assisted answer tends to share a few traits. It states a position early. It attributes claims. It defines terms rather than assuming them. None of this is new advice — it is what good editing has always produced — but the penalty for ignoring it is now immediate and invisible.

What I check before publishing

• Does the opening paragraph answer the question, or set up answering it later?
• Is every number attached to a source a reader can verify?
• Would a section heading make sense lifted out of the page entirely?
• Is there one claim here a competitor could not also make?

Most B2B content does not fail because it reads like AI slop. It fails because there is no compelling reason for anybody to read it.

Structure is the cheapest advantage left

Two pages can contain the same information and perform very differently. The one that wins is usually the one where a reader — or a model — can find the answer without reading to the end. Clear headings, short declarative openings and tables where a table is genuinely the right shape for the data.

Where this leaves the brief

I have not changed how I brief a piece. I have changed what I check at the end of it. The research, the product angle and the objection handling all happen exactly as before; the difference is a final pass asking whether the argument survives being summarised by something that has never met your company.