For years, one of the basic assumptions behind SEO analysis has been straightforward: more search volume or more impressions usually means a more important query. If one query generates 1,500 impressions and another generates 10, we naturally pay more attention to the first one.
I am starting to wonder whether AI search makes that assumption incomplete. As I have been digging through Google Search Console data, including the new Generative AI performance report, I keep coming back to one question: what if some of the highest-impression questions are not where the buyer’s search journey starts, but where multiple journeys converge?
I do not know yet if that is true, but I think the pattern is worth investigating.
Traditional search is fairly easy to conceptualize. Someone searches for something like “generative engine optimization tools,” Google returns results, and the user may refine the search from there. Historically, we have tended to analyze those searches as relatively independent events.
AI Mode changes that behavior because it is conversational. A user can ask an initial question, get an answer, and continue asking follow-up questions while maintaining the context of the conversation. Google also counts those follow-up questions as new queries in Search Console.
That creates an interesting measurement problem. A query in GSC may represent someone’s original question, or it may be question two, three, or five in a much longer exploration. From the query alone, we do not know.
Here are four actual queries showing impressions for my site:
“what is the best tool for showing the performance of generative engine optimization?”
1,454 impressions
“how do ai search optimization/geo platforms present benchmarks to leadership?”
467 impressions
“which ai search optimization/geo platforms provide the most credible competitive benchmarking?”
151 impressions
“how can i track my ai visibility”
1 impression
The traditional interpretation is obvious: the first query is the opportunity, and the last one barely matters.
But if these questions are part of conversations rather than isolated searches, I think there are at least two possible explanations.
The first possibility is the intuitive one. A large number of people begin with the highest-impression question, get an answer, and then branch into increasingly specific follow-ups. One person asks about benchmarking. Another asks about reporting to leadership. Someone else asks about price or integrations.
In that model, the journey looks like this:
Common starting question → many different follow-up questions
That may absolutely be what is happening.
But there is another possibility. What if users begin with highly individualized questions based on their own situation, then gradually move toward a smaller set of similar evaluation questions?
One person might start with “how can I track my AI visibility?” while another asks how to know whether ChatGPT recommends their company. A third may want to understand how to show GEO performance to a CMO.
Those starting questions could each have very low impression counts because they reflect the specific language and context of individual users. But as Google responds and the conversation progresses, multiple journeys could begin to converge around a question like: “What is the best tool for showing the performance of generative engine optimization?“
In that model, the pattern is reversed:
Many individualized starting questions → smaller set of common questions
That is the theory I find more interesting right now.
There are a few things we know. AI Mode supports follow-up questions and maintains conversational context. Google uses query fan-out to explore related subtopics, and each follow-up question in AI Mode is counted as a new query in Search Console. AI Mode activity is also included within the broader Search Performance data.
There is an important limitation, though. Google’s Generative AI report does not currently provide query-level data, so I cannot take one of the queries above and say with certainty that it came from AI Mode. The broader GSC query data includes both traditional search and AI activity.
So the pattern itself is not proof.
What I am inferring is that impression distribution across some long, conversational queries may contain clues about where those questions occur within a broader journey. Specifically, low-impression questions may sometimes represent individualized entry points, while higher-impression questions may represent places where multiple journeys begin to converge.
There are other explanations. Search demand may simply be higher for one phrasing. Google’s suggested follow-ups may influence what people ask next. Some of the queries may have nothing to do with AI Mode at all.
That is why I would not look at a query with one impression and declare it a starter question. But I also do not think we should automatically conclude that it is strategically unimportant.
This gets more interesting if where a question occurs in the journey affects which sources influence later answers.
AI systems do not necessarily approach every follow-up as an entirely isolated request. Context can carry forward, and the information retrieved earlier in a conversation can influence what comes next. Additional retrieval can still happen later, so I am not suggesting that grounding only happens in the first one or two questions.
My working hypothesis is narrower: if a source helps establish the context early in a conversation, it may have an advantage in remaining part of that conversation later.
If that is true, even some of the time, then identifying early-stage questions could matter a lot. A query with 10 impressions might deserve more attention than one with 1,000 if those 10 impressions represent the questions that shape the context of the journey.
That does not mean volume stops mattering. It means journey position may matter too.
Rather than treating this as a new GEO rule, I’m approaching it as a hypothesis to test.
The next step is to group related queries and map likely journeys. I want to look at which questions seem broad or individualized, which look like natural follow-ups, where multiple paths appear to converge, and which pages are being surfaced along the way.
From there, the next question becomes more practical: once I identify a strategically important question, why is my brand not showing up for it?
That is what I explore in the next post, where I look at three possible explanations: discoverability, content, and authority, and why each one requires a different response.
I am not ready to say we should stop prioritizing high-impression queries. But conversational search introduces another variable that traditional keyword research rarely had to consider: where the question sits within the conversation.
For years, we have analyzed keywords based on volume, intent, competition, and conversion potential. We may now need to add another dimension: journey position.
If AI search increasingly involves a series of connected questions rather than isolated searches, understanding how those questions relate to one another may become just as important as understanding how many impressions each one receives.
Which leaves me with the question I am testing now:
What if some of the smallest numbers are actually where the most important journeys begin?