A client told me once: "We checked every query we set up in Genview, and we showed up on all of them."
When I actually looked at what they'd registered, the queries were these:
"What is [Company]"
"[Company] pricing"
"[Company] reviews"
Every single one included their company name. I couldn't help but ask: "What happens if someone who doesn't know your company asks AI for help?"
There was a pause. They looked back at the query list. Not a single one of those queries was there.
This isn't a rare case. Most companies that start using Genview hit this exact wall at first. They feel reassured that they're "showing up in AI" — but in reality, they only show up to people who already know their name. In the conversations of people who don't know them yet, they're not even a candidate. This is GEO's blind spot.
Why This Mistake Happens
The cause is simple. People design GEO queries with the same mindset as SEO keywords.
In SEO, you're searched via short keywords like "X tool" or "X comparison." Including your company name to check your ranking is a natural instinct. But users phrase things differently when talking to generative AI.
"How can I check whether my company shows up in AI search?"
"What should I look for when choosing a GEO tool?"
"What AI search tools would you recommend as of 2026?"
These aren't keywords — they're conversational context. And most users, when asking these questions, don't yet know your service's name. In other words, if you only check branded queries, you're completely blind to the most important scenario: a user who doesn't know you yet, consulting AI.
So the problem isn't the number of queries. It's which customer, and which moment in their research, you're actually looking at.
There Are Three Axes to Query Design
To solve this, I started thinking about query design across three axes.
Axis 1
Funnel Design
Split queries by consideration stage
×
Axis 2
RAG-Triggering Design
Mix in queries that trigger web search
×
Axis 3
Named / Unnamed Design
Are you a candidate without your name?
Axis 1: Funnel Design — Splitting Queries by Consideration Stage
How a user phrases a question to AI changes completely depending on what stage of consideration they're in. Organizing this into TOFU, MOFU, and BOFU stages makes it easier to understand.
TOFU: The Awareness Stage
At this stage, users aren't yet looking for a specific tool. Questions like "what is AI search strategy" or "why doesn't my company show up in AI" emerge — questions about becoming aware of the problem itself. Your brand name doesn't need to appear here. This confirms whether AI can correctly explain that problem space.
What is AI search strategy?
What is GEO?
Why doesn't my company show up in ChatGPT?
What's the difference between SEO and GEO?
MOFU: The Solution-Seeking Stage
Having recognized the problem, users are now looking for specific methods or tools. What matters is whether the "GEO tool" category appears in AI's answer at all.
How can I check whether my company shows up in AI search?
What is a GEO tool?
What should I look for when choosing a GEO tool?
What should I check first when addressing AI search strategy?
BOFU: The Comparison and Adoption Stage
At this stage, users are comparing specific tools. What matters isn't just whether your name appears, but whether you're presented with the right context and the right strengths.
What GEO tools would you recommend?
What GEO tools are usable as of 2026?
What is Genview?
How is Genview different from other tools?
The client I mentioned at the start had only registered BOFU branded queries. Without TOFU and MOFU, you become blind to users at the awareness stage and the solution-seeking stage.
Axis 2: RAG-Triggering Design — Mixing In Queries That Trigger Web Search
Generative AI doesn't perform external search (grounding) for every query. Queries about general concepts tend to be answered from trained internal knowledge alone, while queries requiring current information or specific fact-checking are more likely to trigger a web search.
Genview, for example, launched in April 2026. It's barely represented in AI's training data yet. For queries answered purely from internal knowledge, we might not appear at all. That's why it's important to deliberately mix in RAG-triggering queries containing words like "latest," "as of 2026," "official information," or "comparison."
Internal-knowledge type: "What is GEO?" "What's the difference between SEO and GEO?" → Confirms AI's general understanding
RAG-triggering type: "What GEO tools are usable as of 2026?" "What's the latest GEO tool?" → Confirms whether information on the web is being picked up
Axis 3: Named / Unnamed Design — Are You a Candidate Without Your Name?
This is where it connects directly back to that opening story.
An unnamed query is one that doesn't include your company name — things like "what GEO tool would you recommend?" or "what AI search tool would you recommend for an e-commerce company?", phrased by category or use case. Most users don't know a specific service name from the start; they consult AI about their problem or goal. Whether you appear here is, in my view, the fundamental measure of how well your GEO strategy is actually working.
A named query includes your company name — something like "What is Genview?" What matters here isn't whether you appear, but whether you're described accurately. If a named query returns an inaccurate description, you need to shore up the information on your official site, FAQ, and feature pages.
Named queries alone only tell you the state of people who already know you. What matters in GEO is how you show up in the conversations of people who don't know you yet.
How to Think About the URL You Attach to a Query
There's one more important thing when registering a query: whether the URL you attach actually contains "the answer to that query."
For example, if you're attaching a features page to "how can I check whether my company shows up in AI search," that page needs to clearly contain a sentence like this:
"With Genview, you can check how you're described across ChatGPT and Gemini (standard), plus Claude, Perplexity, and Grok as optional add-ons."
Simply attaching a URL isn't enough. The design order is: query → the answer AI is looking for → the URL that supports that answer. Thinking in this order makes a page more likely to be cited when it's picked up by RAG.
① Query
The question a user asks AI
→
② The Answer AI Wants
Write a sentence that directly answers it
→
③ Attach the URL
Point to the page containing that answer
A Real Query Design Example
Here, using Genview itself as an example, I'll walk through 15 queries and their related URLs. This combines all three axes, covering everything from the awareness stage through to the comparison stage.
Rather than setting up only queries that include your own brand name from the start, including queries that reflect a user who doesn't know you yet consulting AI lets you check whether you're a candidate without being named.
* The Base plan includes 15 queries as standard, all monitored daily. If you want to add more than 15, you can add queries at ¥2,000/month each. See the pricing page for details.
Common Failure Patterns
Registering Only Named Queries
This is the case from the client I mentioned at the start. You only see the state of people who already know you. Always include unnamed queries.
Skipping TOFU and MOFU
With only "recommendation" or "comparison" style BOFU queries, you can't isolate why you're not appearing as a candidate. Is your problem category not even recognized, or are you just not in the solution category? You only find out by including TOFU and MOFU.
Skipping RAG-Triggering Queries
The newer your service, the less likely you are to appear in queries answered purely from internal knowledge. Including at least one or two queries with words like "latest," "as of 2026," or "based on official information" lets you check whether AI is referencing information on the web.
Attaching a URL Without Building Out the Content
Registering a query is pointless if the target URL doesn't actually contain an answer to that query. Designing in the order of query → answer → URL matters.
Summary
Most cases of "we show up in AI" are actually only confirmed via branded queries — you don't show up in the conversations of people who don't know you yet
The problem isn't the number of queries — it's which customer, and which moment in their research, you're looking at
Funnel design: split queries across the TOFU, MOFU, and BOFU stages
RAG-triggering design: the newer your service, the less internal-knowledge-only queries will surface you. Mix in RAG-triggering queries
Named/unnamed design: whether you appear in unnamed queries is the fundamental measure of your GEO strategy's real strength
Make sure the URL attached to each query clearly contains the answer to that query
While writing this, something kept nagging at me: could Genview measure "whether RAG gets triggered" as its own metric? Right now, the same query can trigger a web search on one day and not on another. It varies by platform too. If we could surface "how likely this query is to trigger RAG" as a score, I think it would meaningfully improve the precision of query design. We haven't built this yet, but it's one of the challenges I'd like to tackle someday.