A growing share of the questions that decide whether a customer picks you are now asked in a conversation with an AI assistant rather than typed into a search box. Those questions rarely show up in a keyword tool. Plenty of them are already sitting in your inbox and your call recordings, and you can pull them out this week.
Here is what one looks like. A furniture retailer opened a call with this: “We moved to Shopify a couple of years ago and lost a big chunk of our rankings. Some of it came back, and then we dropped again this year. Is that still the migration, or are our own collection pages competing with each other?” The closest a keyword tool gets is something like “shopify migration SEO”, which tells you almost nothing. The real question names a platform, a timeline, two possible causes and a decision to make. That is a content brief.
Getting the questions out
- Export transcripts from the last three months of sales and discovery calls. Add support tickets and pre-sale inquiries, because sales calls skew toward people who are already close to buying.
- Strip out names, company details and anything identifying before transcripts go into a third party tool, and check that your call recording consent covers this kind of use.
- Run them through an AI assistant with a prompt like this: “From these call transcripts, list every question the prospect asked or implied. Rewrite each one the way they would phrase it to an AI assistant, and keep their constraints, such as company size, budget, timeline, platform, and what they’ve already tried. Group similar questions together and note how many calls each group came up in.”
- Do the same with forums, because your calls only cover people who already found you. A search like site:reddit.com shopify migration rankings dropped brings up threads where owners describe the same problem in their own words. Copy the relevant ones into a document and run the same prompt.
That last instruction in the prompt is the important one. If the same scenario comes up in five separate calls, that is the closest thing to demand data you will get for a question with no search volume, and it is far more persuasive in a meeting than “I think people are asking this.”
Checking what already reaches you
Search Console works here as validation, not discovery. In the Performance report, use the Custom (regex) option in the Query filter to pull queries of eight words or more, and queries that start with a question word. If your scenario shows up there, real people are bringing it to Google too. Google anonymizes rare queries, and conversational queries are rare almost by definition, so treat what you see as an undercount.
Then put everything in one sheet: the question, where it came from, how many times it came up, and a priority based on business value, evidence the question is being asked, the size of your coverage gap, and whether you can answer it with real specifics. The questions tend to cluster around a handful of situations, and those clusters are your content plan. Often a cluster can be handled by strengthening a page you already have, which is quicker than publishing something new.
If you want to go further, the Gemini API’s Grounding with Google Search returns a field called webSearchQueries, listing the searches the model ran to answer your prompt. Feed in one of your questions and each sub query is effectively a section your page needs to cover. It is an approximation of how a Gemini model expands a question, not a record of what ran for a real user, so use it as a coverage checklist and nothing more definitive.
One honest limit: none of these questions comes with a volume figure. You are prioritizing on judgment and frequency, so present the result as a well evidenced shortlist, especially to anyone used to seeing a number next to every target.



