How AI Search Behaves: Why It Doesn't Give You Clicks
AI search is not "search with a summary added on." Its goal is to end the query outright. Understanding this is the only way to see why measuring GEO by click-through rate is bound to disappoint.
Bottom line first: AI search is designed to "let users stop clicking." It delivers the answer directly; clicks are a byproduct, not the goal. So in this chain, what you compete for is not clicks but the right to be written into that answer.
1. Three entry points, three different behaviors
- Built-in summaries in search engines: The user is still in the search box, but the answer is already given above. Clicks are intercepted, yet users still have the habit of "clicking a few more to see."
- Standalone Q&A apps: Users don't search at all; they converse directly. There is no "ranking" here, only whether you get cited.
- Summaries on vertical platforms and communities: These synthesize within a specific domain. They rely more on content already on the platform; your official website is just one source among them.
2. Zero-click is not an accident; it is the default outcome
A user asks "which one is good" or "how do I do this," gets a complete answer, and leaves. That is a product design success, not your failure — but your presence must be inside that answer.
So there is a metric more worth watching than clicks: zero-click presence rate — the share of questions relevant to you where the answer mentions you.
3. The three "nots" of AI search
| Characteristic | Meaning | Impact on you |
|---|---|---|
| No guaranteed links | Sometimes only a conclusion, sometimes only one or two sources | You may still be used even if not listed |
| No repeated identical answer | Ask the same question twice and the wording may differ | A single observation doesn't count; look at trends |
| No explanation of trade-offs | It doesn't say why these sources were chosen | You can only infer from signals, not guess |
4. The measurement approach you should switch to
- Fix 20 core questions, ask them once a week, and record: whether you are mentioned, and how you are described when mentioned.
- Look at tone, not just frequency — "Company X also offers similar services" and "recommend this one" are not on the same level.
- Track trends alongside branded search volume and direct traffic; these two are downstream reactions to answer exposure.
Measure GEO by click-through rate and the conclusion will always be "no effect." That is not GEO being ineffective; it is the wrong ruler.
Related reading: Four metrics that reflect change earlier than traffic, How to do the math in the zero-click era.