For three years chatbots have been sold as the end of the search box. According to Golem, they are now being sold to the advertising industry as its replacement — a surface where product recommendations arrive inside a conversation rather than beneath a list of links.
The commercial logic is easy to see. A search query is a handful of words and an inference about intent. A chat message is the intent itself, spelled out, with budget, constraints and context attached. It is the most qualified audience signal the advertising business has ever been handed, and it arrives in a window users already trust enough to type their medical symptoms into.
The mechanics are where it gets uncomfortable. In search advertising the border between a result and an ad is spatial: a shaded block, a small label, a different background. In a conversation there is no block. An assistant that recommends a specific pair of headphones in a fluent sentence looks exactly like an assistant that recommends them because someone paid for it. Every honest design has to answer a question the ad industry has never had to answer before: how do you mark a recommendation that has no slot to mark?
Measurement is the second unsolved problem. Impressions and click-through rates describe rectangles on a page, not sentences in a dialogue. Attribution inside a chat may end up resting on the assistant operator's own logs — the seller grading its own homework.
Europe has a partial frame for this. Paid placement must be identifiable as advertising, and the EU AI Act adds transparency duties for systems that interact with people. Neither has been tested on a conversational answer that quietly ranks a product first.
The interesting question is no longer whether advertising reaches chatbots — Golem's reporting suggests that conversation has already started. It is whether the assistant stays worth asking once its answer has a price.
