What you get
A report you can read in one sitting.
-
The questions your buyers ask
A fixed set of buying questions for your category, the ones people type before they choose.
-
Five tries each
Every question is asked five times, each in a new conversation, because one answer can mislead.
-
The questions where you’re never named
The gaps: questions where your name does not come up in any try.
-
Who is named instead
The other brands that appear in the answers when yours does not.
A real example
A CRM brand, 18 questions, five tries each.
Real answers from gemini-3.5-flash-lite on October 2, 2026, five tries per question. The brand is not named here, and neither are its competitors.
7
questions where the brand was never named
5
questions where it appeared only sometimes
6
questions where it was named every time
A CRM brand · gemini-3.5-flash-lite · October 2, 2026 · five tries per question
Named Not named. Each row is one question, each mark one try. Across all 90 answers the brand was named in 44.
All 18 questions, five tries each
Never named7 questions
These are the gaps. In every try, the answer did not include the brand.
- What's the best CRM for a 40-person outbound sales team?
- What's a good free CRM for a freelancer or solo consultant?
- Which CRM works best for a B2B SaaS company with a product-led sales motion?
- Which CRMs have the best built-in email sequencing and call tracking?
- What are the best enterprise-grade CRMs that cost less than the market leader?
- Which CRM is best for managing a long, multi-stakeholder enterprise sales cycle?
- What's the best CRM for a company that wants sales and marketing in one tool?
Named only sometimes5 questions
The same question gave a different answer from one try to the next.
- Which CRM is best for a small business that sells mostly over email and phone?
- What CRM do mid-market companies with 200 to 500 employees usually choose?
- What are the best CRM options for a real estate team?
- What are the most affordable CRMs that still have workflow automation?
- Which CRM has the best reporting and sales forecasting for a VP of Sales?
Named every time6 questions
In every try, the answer included the brand.
- Which CRM should a 10-person startup use that won't need a consultant to set up?
- What are the best CRMs that integrate natively with Slack?
- Which CRM is easiest for a non-technical sales team to adopt?
- What CRM should an agency use to manage clients and deals together?
- Which CRMs integrate well with Google Workspace and Gmail?
- Which CRMs are best for sales teams that work mostly from mobile?
How a run works
Four steps, and you only do the last one.
-
The questions
We use questions that do not name any brand in the category. A few mention an integration platform such as Slack or Google Workspace, because that is how buyers ask.
-
Each asked five times
Every question goes to one model five times, each time with no memory of the others.
-
We record two things
Whether you are named, and whether you appear in a heading, list label, table or bold text.
-
You receive the report
The questions, the marks for every try, and who is named when you are not.
Read this before you act on it
What a run can’t tell you.
- One model, one date. A result describes one model on the day it was run.
- Answers shift. A reworded question, or the same question next month, can give a different result.
- It is not a ranking. It records whether a name appears, not where your brand sits against others.
- Nothing about other assistants. A result for one says nothing about the rest.
- No traffic forecast. Being named is not a prediction of visits or sales.
- Named is not an endorsement. A name inside a dismissive aside counts as named.
Questions
Citation tracking, briefly.
What is citation tracking?
It is a report on whether an AI model names your brand when people ask the questions your buyers ask. We run a fixed set of questions for your category, ask each one five times, and show you when your name comes up and when it does not.
Why ask every question five times?
Because the same question can get different answers. In the sample, 5 of 18 questions named the brand in some tries and not in others. One try can mislead you. Five show the pattern.
What does named mean?
Your brand name appears in the answer. We also record whether it appears in a heading, a list label, a table or bold text, rather than only inside a sentence. A name inside a dismissive aside still counts as named. It is not an endorsement.
Can I run it myself?
A self-serve version is still to come. Today we run it for you. The question sets are also published on the Citehound MCP server, with a short protocol for running them yourself in any assistant.
Which model does it use?
The sample used gemini-3.5-flash-lite on October 2, 2026. Today we run Gemini models only. A result describes one model on one date and says nothing about any other assistant.
Is it a ranking?
No. It records whether a name appears, not where your brand sits against others. Answers shift with the wording of a question and over time, so a result is a picture of one day.
How do I get early access?
Email us your brand, domain, category and three competitors. We reply by email.
Early access
Find out when AI names you, and when it doesn’t.
Tell us your brand, domain, category and three competitors. We reply by email.