On this page
The short version
I spent 15 years planning content from keyword volume. Zero search volume questions were the ones we skipped. Then we compared Google's search data with what five AI systems actually answered, on 50 real buying questions for four businesses.
- Not one of the 50 questions had Google search data in the words a person would actually use. For 20 of them, Google had no data even for a broad version. (What Google knows)
- AI answered every one of those questions and named businesses in them. Without their own Google data: 52% of answers named a business. With it: 64%. (AI answers anyway)
- A dental training institute had no Google data for any of its 17 questions, and AI named it in 28% of answers. (Who wins)
- An electrician named in 35% of answers on his searchable questions dropped to 4% on the rest. (Who wins)
To be clear: nobody publishes how often people ask AI a particular question. I won't guess at "AI search volume". What the data does show is that these questions have no recorded Google demand, they read the way people talk to AI assistants, and AI is recommending businesses for them every day.
How we measured
The short answer: Google's search data for each question, next to every AI answer to it.
Our tracker, LLM Visibility Lens, asks each business's buying questions every morning in ChatGPT, Gemini, Perplexity, Google AI Mode and Google AI Overviews, logged out, from the business's area. It also looks up Google's monthly search demand for each question.
Nobody searches Google with a 25-word sentence, so the tool first shortens each question to a keyword. "My power is out in Coquitlam tonight who can come quickly" becomes "emergency electrician coquitlam". If that has no data, it tries a broader one, such as "electrician coquitlam".
This study covers four businesses and their 50 questions, with 6,010 AI answers from 30 September 2026 onwards:
- Kato Electrical, a Vancouver electrician and our public demo.
- A Miami dental training institute (anonymous).
- An online mobility store (anonymous).
- An industrial software company (anonymous).
What this can't tell you. "No data" means Google's keyword data has nothing recorded, not that nobody ever searches it. Ahrefs found that most "zero-volume" keywords still get a few searches a month. And we can't measure how many people ask AI each question. We can only see what AI says when they do.
I sell this tracker, so weigh my view. Kato's questions and answers are in the demo.
What Google's data knows about real buying questions
The short answer: almost nothing about the questions as people actually ask them.
| What Google's data had | Questions |
|---|---|
| The full question, as typed | 0 of 50 |
| Its shortened keyword had data | 7 |
| No data for the keyword, fell back to a broader one | 23 |
| No data even for the broader keyword | 20 |
The only 7 with their own keyword data were Kato's short "best electrician in…" questions, which shorten to keywords like "electrician vancouver" (720 searches a month) or "electrician burnaby" (70).
The rest were real, specific situations:
- "Is there an electrician open right now in Port Coquitlam that can come out tonight"
- "What is the best lift recliner chair I can order online for my dad who has trouble getting up from a seated position on his own"
- "Is there dedicated software for tracking shutdown milestones and punch lists?"
- "Where can I find a hands-on dental implant training course in Miami that actually lets me work on live patients and counts toward my CE credits?"
That last one is 26 words. A keyword tool would tell you to forget it.
AI answers them anyway
The short answer: AI recommended businesses for questions with no Google data almost as often as for those with data.
| Google data for the question | AI answers naming at least one business |
|---|---|
| Short keyword had data (7 questions) | 64% (571 of 899) |
| Fell back to a broader keyword (23) | 52% (1,366 of 2,631) |
| No data at all (20) | 52% (1,291 of 2,480) |
Every one of the 43 questions without their own keyword data got at least one AI answer that named a business. For 24 of them, AI named a business in at least half its answers.
Length mattered a little, not a lot:
- 8 words or fewer: a business was named in 59% of answers.
- 9 to 14 words: 57%.
- 15 words or more: 46%.
The longest questions still got a recommendation in nearly half the answers. On Google, some of these would barely trigger an AI Overview; we showed how wording changes that in AI Overview tracking for near me searches.
Who wins the questions keyword tools can't see
The short answer: whoever AI already trusts for that specific situation, which may well not be you.
The electrician is strong where keywords exist and weak where they don't:
- The 7 questions with keyword data: named in 35% of AI answers (312 of 899).
- The 9 without: 4% (53 of 1,189).
His SEO has been built around keywords like "electrician vancouver", and it shows. But "my power is out in Coquitlam tonight who can come quickly" had a business named in 55 of 135 answers, and none of them was Kato. We saw the same in spoken questions.
The dental training institute shows the opposite. None of its 17 questions had any Google data, even broad. Yet AI named it in 28% of answers across them, and in 43% for the 26-word question above (54 of 125). Its course pages answer those exact situations, so AI picks them up. We looked at which pages do that in which pages AI reads.
The software company and the online store show the risk. AI named a business in 60% of answers to the software company's questions, none of which had their own keyword data, and in 80 of 126 answers to the store's lift chair delivery question. Neither business was named once. Someone is winning those questions; it just isn't them.
Why keyword tools miss them
The short answer: keyword tools count what people type into Google, and people talk to AI differently.
- People search Google in fragments, like "emergency electrician coquitlam".
- They ask AI in situations, like "my power is out in Coquitlam tonight, who can come quickly". The extra words carry what really matters to them: tonight, quickly, my dad can't stand up, I want to work on live patients.
- Keyword tools also round small numbers down to nothing, and they don't record every phrasing.
So a content plan built only from keyword volume is built from the fragments, and misses the situations AI answers.
Google's own guidance on AI features and your website says its AI can run several related searches behind the scenes to answer one complex question. A long question is answered from many short ones. The page that wins is the one that fits the whole situation.
How to find these questions
- Write down what customers actually ask. Phone calls, emails, booking forms, sales conversations and reviews are full of full-sentence questions.
- Look at the long queries in Search Console. The rare, long ones hint at the situations people describe.
- Use a question generator built for AI. Our free question finder writes the buying questions to track from your website, including long, spoken-style ones.
- Ask them in each AI system, over several days, and note who gets named. Answers change daily, so one check proves little; see what one free check can and can't do.
How to win them
- Answer situations, not just keywords. Have pages that cover the service, the place and the circumstance: same-day, out of hours, for a particular customer type. See how to get recommended by ChatGPT.
- Use your customers' words in headings and FAQs. If people ask "can you come tonight", say "yes, we come out tonight in Coquitlam" if it's true.
- Make sure the page names you clearly. A page that explains the situation without making you the obvious answer helps others; see when your own web page helps a competitor.
- Be on the sources AI reads for that situation: directories, lists and reviews. See the directories AI reads.
- Track them. Questions with no keyword data never show up in a rank tracker. Add them to an AI tracker and watch who wins.
Questions to keep tracking
LLM Visibility Lens tracks your buying questions in all five AI systems daily, shows Google's search demand next to each one, and tells you which ones nobody is winning yet. See Kato's in the live demo. Plans start free, and agencies can report under their own name; see the agency plan.
More reading:
- Which AI system recommends local businesses most
- Invisible on Google, recommended by AI
- Why a single AI visibility score can mislead you
- How to report AI visibility to clients
- Which websites ChatGPT trusts
- 5 AIs, same questions, 1,814 answers
- 5 AI systems, 35 questions
- Google Business Profile and AI answers
- Google AI Mode for local businesses
- How to rank in AI Overviews
- Perplexity vs ChatGPT
- ChatGPT citations
- How to rank in ChatGPT
- Why there's no rank #1 in ChatGPT
- We asked AI who to hire 80 times
The method is on the how it works page. There's also a comparison with Semrush's AI visibility tool, the changelog, the FAQ page and the security page. If you'd like help finding the questions your customers ask AI, get in touch.
Questions people ask
Are zero search volume keywords worth targeting for AI?
In our data, yes. Of 50 real buying questions, none had Google search data in the words people use, and 20 had none even for a broad version. AI still named a business in 52% of answers to the questions without their own data, so someone is being recommended for them.
Do people ask AI questions that nobody searches on Google?
Nobody publishes how often people ask AI a particular question, so we can't count it. What we can show is that real buying questions, especially long situational ones, have no recorded Google demand, and AI answers them and names businesses every day.
Why don't keyword tools show conversational questions?
Keyword tools count what people type into Google, which is usually short fragments. Full situations, like needing an electrician tonight or a course that lets you work on live patients, are rare as exact searches, so tools show no data or round them down to zero.
How do I find the questions people ask AI about my business?
Start from what customers really ask: calls, emails, booking forms, reviews and the longest queries in Search Console. A question finder built for AI can draft them from your website. Then ask them in each AI system over several days and see who gets named.
Can a business win AI answers without search volume?
Yes. A dental training institute had no Google data for any of its 17 tracked questions and was named in 28% of AI answers across them, and 43% for one 26-word question. Its pages answered those exact situations.
Should I stop using keyword research?
No. Keyword data still shows what people search on Google, and an electrician in our data was named most on his searchable questions. But keyword research alone misses the situational questions AI answers, so add a list of real customer questions and track them separately.
Harjeet Singh Bedi Founder, LLM Visibility Lens · SEO consultant, Bedi SEO
Harjeet has worked in SEO since 2010, first in agencies and now running Bedi SEO for local businesses, online stores and agencies. He built LLM Visibility Lens after clients kept asking whether they showed up in ChatGPT and other AI tools and nobody could give them an honest, measured answer. Every figure in this article comes from answers his tool collected.