Why Relying on Search Volume Is Costing You Your Best Content Opportunities

Most marketing teams have changed how they write content for AI, but have barely changed how they decide what to write about. The result is an odd mismatch: content that's structurally and technically ready for AI, while the topic list is still pulled from a spreadsheet sorted by search volume. Writing style has evolved, but the topic-selection criterion hasn't.

The reason lies in a fundamental shift in how AI assistants operate. According to Google's own documentation, AI Overviews and AI Mode don't rely on a single query to shape a response — they use "query fan-out," triggering multiple related searches across different topics and sources. ChatGPT works similarly, rewriting a user's prompt into several search queries before retrieving information. In other words, search hasn't disappeared — it's simply being carried out by the model on the user's behalf.

That shift is exactly what makes search volume unreliable as a prioritization metric. Competition no longer plays out on a single keyword's results page — it plays out in the list of sources the model draws on to build its answer. Your page might get cited for a supporting question you never targeted, or, just as easily, stay invisible across an entire topic simply because it only covers the headline question.

The author recommends four practical shifts: first, fully answer the eight to ten supporting questions that branch off the main query, rather than the main query alone; second, cover every concept and entity tied to the topic consistently, instead of repeating exact-match phrases; third, put comparisons and decision criteria — not definitions — at the center of the content, since those are what actually help a buyer decide; and fourth, keep search volume as the primary metric only for short, transactional queries, such as branded or local searches.

The bottom line: keyword research hasn't lost its value — what's dead is blind faith in the search-volume column alone. Understanding what your audience wants to know, in what order, and with what intent matters more than ever, because the model assembling the answer is far less forgiving of a page that covers a topic only halfway than legacy search engines ever were.