The first query: why the first question a customer asks AI decides the shortlist
Written by
Odin Lund
Category
AI search
Date
Share this article
What the first query is
The first query is the first question a customer puts to an AI model about a purchase: “what should I use for X”, “who handles Y near me”, “best option for Z”. It is the AI-search equivalent of the first search, with one structural difference. A search result offered ten links and left the choosing to the customer. An AI answer names two or three options and, in effect, makes the shortlist for them. Everything the customer asks next (“how does A compare to B”, “is A right for me”) is asked about that shortlist.
The audience for this moment is now mainstream. Google reported more than 2 billion monthly users of AI Overviews (Alphabet Q2 2025 earnings), and OpenAI reported 800 million weekly users of ChatGPT (OpenAI DevDay, October 2025). Pew found that when an AI summary appears, users are half as likely to click a result at all (8% versus 15%; Pew Research Center, July 2025). The first query is where the decision narrows, and increasingly where it ends.
What our data shows
In our legal data set (60 prompts, 14 firms, three models, daily), the businesses named in the first-query prompts were, with few exceptions, the businesses named in the comparison and decision prompts. Businesses that were absent from the first answer rarely re-entered later in the journey. For the client we manage, decision-stage visibility reached 37.8% while comparison-stage visibility was 6.3%: the model would recommend the firm when asked directly, but it was not yet being placed on the shortlist against named rivals. Closing that gap is now the work.
Where do first-query answers come from? From the sources that agree. In the same data set, the client’s site drew 4,319 retrievals in 30 days against 1,296 for the nearest competitor, and third-party sources carrying the client produced roughly four times the retrievals of its own site. The models built the shortlist from the domains they trust in the category, and they trusted the business whose facts were the same everywhere.
How a brand gets into the first answer
Be the consistent one. Pick the facts that define you (what you do, for whom, where, since when, with what proof) and make every source state them identically. Answer the first query directly on a page of your own: not a landing page about you, a page that answers the question a stranger would ask, with the facts underneath. Be present, with the same facts, in the directories, review sites and publications the models cite for your category. And add structured data so the facts are extractable. The controlled evidence is consistent with this: content with citations, quotations and statistics gained up to 40% visibility in generative engines, keyword stuffing gained nothing (Aggarwal et al., KDD 2024).
Measuring it, and sources
Track the first-query prompts as their own group, by model. Visibility on that group is the earliest indicator you have; share of voice on it tells you who else is on the shortlist; citations tell you which of your pages the models used to put you there. Then follow the customer: visits from AI answers, the page they landed on, and whether they became a customer. That is the number the first query is worth.
Sources
SourceNode tracking, aisearchmonitor.com, 60 prompts, 14 firms, June to September 2026.
Alphabet Q2 2025 earnings call, July 2025: AI Overviews monthly users.
OpenAI DevDay, October 2025: ChatGPT weekly active users.
Pew Research Center, 22 July 2025: click behaviour with and without AI summaries.
Aggarwal et al., “GEO: Generative Engine Optimization,” KDD 2024 (arXiv:2311.09735).
Share Blog
