Key takeaways
Pair each visible query with its landing page before drafting a buyer question.
A rewritten prompt is a research hypothesis, not proof of an exact AI-assistant query.
Keep Search Console metrics attached to the original query, not the expanded prompt.
Query and URL filters can change totals because of truncation and excluded anonymized queries.
Google includes AI Overviews and AI Mode traffic in the Performance report’s Web search type.
Start with a Search Console query and the landing page associated with it. Use that pair to identify the buyer’s likely task, then rewrite the query as a natural question worth tracking in an AI answer. Keep the original query and page as evidence, but treat the new prompt as a hypothesis—not proof that someone asked an AI assistant that exact question.
This is useful because Google’s AI experiences can support nuanced questions and comparisons, while Search Console reports the search activity associated with your site. Google includes traffic from AI Overviews and AI Mode in the Performance report’s Web search type; that report is not a standalone list of AI-assistant prompts. Source
Build a list of query and landing-page pairs
Choose a scope. Open the Search results Performance report and set a date range. Add country or device filters if they reflect the audience you want to study. Multiple filters can be applied together. Source
Start with a relevant page. Filter for a product, comparison or educational page, then review its queries. Record the page and each useful query together, along with the date range and filters. Repeat for other pages.
Keep the source metrics with the query. Record available impressions and clicks as context for the original search term. Do not assign those numbers to your rewritten AI prompt.
Identify the likely task. Ask what someone using that query might be trying to learn, compare or decide. The page provides context, but neither the query nor the page proves the searcher’s intent.
Draft and review questions. Write natural questions that preserve the query’s topic while adding a plausible decision or constraint. Remove duplicates and questions that do not fit your audience.
Five examples: from search query to buyer question
The following pairs and prompts are fictional illustrations, not Search Console data or documented customer questions. Page descriptions stand in for landing-page URLs.
Search query | Landing page | Possible buyer task | Prompt to consider tracking |
|---|---|---|---|
project management software | Product overview | Understand options | What should a small team look for in project management software? |
project management tools comparison | Comparison guide | Compare options | How should I compare project management tools for a growing team? |
project management software for agencies | Agency use-case page | Assess fit | What features matter most when an agency chooses project management software? |
how to manage project deadlines | Deadline planning guide | Solve a problem | How can a team organize and monitor project deadlines? |
switch project management tools | Migration guide | Plan a change | What should a team consider before switching project management tools? |
Each prompt adds context that is plausible, not confirmed. For example, “small team” is a research choice; the original broad query does not establish team size. Review those additions against what your team knows about its audience before tracking them.
Expand each pair without losing buyer intent
For a promising pair, draft one question in the searcher’s apparent frame. Then, where relevant, test a second angle rather than generating a long list of near-duplicates:
Problem: What is the person trying to accomplish?
Comparison: What options or criteria might they need to weigh?
Fit: What type of team, use case or constraint might change the answer?
Implementation: What would the person need to know before acting?
Label each question with its intended audience, topic and source query. Favor questions that a buyer could reasonably ask over prompts that force a brand name into every sentence. Google describes AI Mode as useful for exploration and complex comparisons, and says its AI features may use multiple related searches to develop a response. That makes related questions useful research candidates, but it does not validate any particular prompt. Source
Filter carefully and account for missing queries
Search Console supports query and page filters, including substring matches and custom regular expressions. For example, a query regex such as (comparison|compare|versus) can help find comparison-related terms. Regex matches are partial and case-insensitive by default; use anchors when you need a stricter match. Source
Filtered totals may not match unfiltered totals. Google says filtering by query or URL can change report totals because of data truncation and the exclusion of anonymized queries. Do not interpret the visible query list as every search behind a page’s traffic, or use a filtered total as a complete inventory of buyer demand. Source
Filters also narrow the research question: a country or device filter can make a list more relevant to a specific audience, but it leaves out activity outside that scope. Record the filters with your notes so colleagues understand what the list does—and does not—cover.
Choose a manageable tracking set
Review each candidate for relevance to your audience, clarity and distinctness. Keep a mix of problem, comparison and fit questions where those angles matter to your business. For each selected prompt, save the original query, landing page, reporting scope and reason for selection.
Use the resulting list to guide AI visibility tracking and content research, not to claim search demand for the rewritten wording. If a question reveals a content gap, review whether an existing page answers it clearly. For eligibility as a supporting link in Google’s AI Overviews or AI Mode, Google says a page must be indexed and eligible to appear in Search with a snippet; its existing SEO fundamentals still apply. Source
Prepare your prompt research list
Select a date range and audience filters, and record them with your research.
Collect relevant query and landing-page pairs with their original metrics.
Draft questions around the buyer tasks each pair suggests.
Review added assumptions, remove duplicates and select distinct prompts.
Save the source query and page alongside each prompt you choose to track.
FAQs
Can Search Console tell me which exact prompts people ask AI assistants?
No. Search Console reports search activity for your site. Google says traffic from AI Overviews and AI Mode is included in the Performance report’s Web search type, rather than described as a standalone list of AI-assistant prompts. Use visible queries as research inputs, not proof of exact assistant wording. Source
Why do totals change when I filter by query or page?
Google says query or URL filtering can change report totals because of data truncation and the exclusion of anonymized queries. Record your filters and avoid treating filtered results as a complete list of searches. Source
Can I use regular expressions to find buyer-intent queries?
Yes. In the Search results Performance report, add a query filter and choose a custom regular expression. For example, (compare|comparison|versus) can surface terms containing those words. Matches are partial and case-insensitive by default. The matches are candidates for review, not proof of intent. Source
Does a page need special optimization to appear in Google’s AI features?
Google says there are no additional technical requirements for AI Overviews or AI Mode beyond being indexed and eligible to appear in Google Search with a snippet. It recommends continuing to follow existing SEO fundamentals and create helpful, reliable, people-first content. Source
Discussion
Join the discussion