Keyword research after AI search: map intent, not volume
Search volume has become a poor guide to opportunity, because the highest-volume informational terms are exactly the ones AI answers absorb. Intent is the better axis.
Key takeaways
- Sort keywords by intent rather than volume; the highest-volume informational terms are the most absorbed by AI answers.
- Commercial investigation queries retain click-through best and carry the highest return in 2026.
- Search Console's high-impression, low-click queries are usually the fastest available win.
- Cluster by the answer a query needs, not by string similarity, or your own pages will compete with each other.
Keyword research in 2026 should sort terms by what the searcher intends to do rather than by how many people search them. High-volume informational queries are precisely the ones an AI Overview answers without a click, while lower-volume commercial and local queries retain their click-through and their value. Volume tells you the size of the room; intent tells you whether anyone in it will buy.
Four intents and what each is now worth
- Informational. Largest volume, heaviest AI absorption, lowest direct value. Still worth publishing, because topical depth is what makes your commercial pages credible, but do not forecast revenue against it.
- Commercial investigation. Comparisons, alternatives, pricing, reviews. Holds click-through well because the user wants to weigh options. This is where the highest return sits in 2026.
- Transactional. Explicit purchase or enquiry intent. Lowest volume, highest conversion, most competitive. Deserves the best page and the fastest template you have.
- Navigational. Someone looking for you by name. Cheap to win, and it is the surface where AI-driven awareness turns into a measurable visit.
Where to get the terms
Search Console is the most underused source, because it reports the actual strings that produced impressions for your site rather than a tool's model of demand. Filter by country and by page, sort by impressions with low click-through, and you have a list of queries where you are already visible and failing to earn the click. That is usually the fastest available win.
Beyond that, read the questions your sales and support teams answer repeatedly. Those are real queries phrased the way real customers phrase them, and they map directly to commercial investigation intent. For multilingual markets, run this exercise separately per language rather than translating a single list.
Clustering without inflating the plan
Group terms by the answer they need, not by string similarity. If five queries would be satisfied by the same page, they are one target, and splitting them across five thin pages will cause those pages to compete with each other. The opposite error is equally common: forcing genuinely different intents onto one page because the words look alike. Ask what a satisfied reader would do next; if the answers differ, the pages should too.
Judging opportunity honestly
For each cluster, estimate three things: whether an AI Overview already answers it completely, whether the current top results are stronger than anything you can realistically produce, and whether a person searching it could plausibly become a customer. A term that fails all three is not an opportunity regardless of its volume, and a term that passes all three is worth pursuing even at a hundred searches a month.
Volume measures the size of the room. Intent tells you whether anyone in it is buying.
Sources