Ask
Again
A simple guide to what AI says
about your business

Free illustrated book 88 pages
A free book by Dmitrij Żatuchin. Each chapter tells one study on AI answers as a short story, with one step to try on Monday. Each chapter page ends with its reference: a link to the paper or its data.
Every chapter is free to read here, with no sign-up. For the PDF, give your email address and the download link arrives by email.
Why I wrote this
My children are 9 and 12. They wanted to know what research I do and what my book is about.
So I tried to explain it in a way they could understand. This book is the result.
If you run a business, it is for you too. Your customers can now ask AI which company to choose. It could be a pizzeria, an accountant, a lawyer, an online shop, a media agency, a publisher, a car dealer or a company that builds and sells homes. You will want to know what the answer says.
In this book I call AI assistants robots. ChatGPT, Gemini and Perplexity are all robots here.
Each chapter starts with a short story, explains a finding from research and gives you one thing to try on Monday. The references at the back link to the studies. A short glossary explains a few useful words.
Please open those links. I founded Rankfor.AI and I run it. We sell businesses tools that measure what the robots say, content written for them to read, and publishing where they read it. We also help publishers sell sponsored content. I sell a book on this subject too. So I earn money when businesses care about these answers. Keep that in mind as you read.
We do not promise a place in their answers.
Dmitrij Żatuchin
Tallinn, November 2026
The people in this book
You will meet eight business owners here. I made them up, so their stories are examples. The numbers come from the studies, and each chapter says what was measured.
The pictures make the numbers easier to see: pizzas, cars, houses, parcels, books and more.
You will also meet these two in the pictures.
A few useful words
- AI
- Software that can do some tasks people do, like writing or spotting patterns.
- Average
- Add all the numbers, then divide by how many you have.
- Brand
- The name and look that help people know a firm or product.
- Data set
- A group of records kept for a study or test.
- DOI
- A fixed code that helps you find a paper or data set.
- GEO (Generative Engine Optimization)
- Trying to help a firm get named in AI answers.
- Large language model (LLM)
- A kind of AI trained on lots of text to write more text.
- Peer-reviewed
- Other experts checked the study before it came out in a journal.
- Persona
- A made-up shopper with needs and questions you can think about.
- Preprint
- A paper shared online before it appears in a journal. It may never appear in one.
- Prompt
- What you ask an AI to do or answer.
- Tone score
- A program’s number for text it reads as positive, neutral or negative.
- Version
- One form of a paper, data set or robot’s software.
- Web search
- Looking on the web for pages to help answer a question.
- Wobble
- How much a count changes when you ask again.
Part One
Ask again: how answers repeat and change
- Chapter 1Peer-reviewed
Robots make a list, and you can check it
The same names keep coming back, so you can count them. Check more than one robot, because they do not always agree.
- Chapter 2Preprint
Ask robots for lottery numbers, and tickets repeat
A robot’s answers can change from one try to the next and still keep landing on a few choices.
- Chapter 3Preprint
One answer misses names, so ask again
One answer shows part of the list. Ten tries found most of what 15 tries found, yet most questions still held a name seen in one try alone.
- Chapter 4Preprint
No single number of tries fits every question
Ask one question about 10 times as a small trial. See how much the answers wobble. Set that beside the difference you care about.
- Chapter 5Preprint
One robot’s named companies barely changed in 20 days
Over about three weeks, this robot kept naming nearly the same known companies on the same questions, and kept leaving out nearly the same ones.
Part Two
What ends up in the answer: links, people and names
- Chapter 6Preprint
A small share of websites got most robot links
Most links went to a small share of websites, and thousands of others got a few each.
- Chapter 7Preprint
About one answer in four named a person
In three trades, about one answer in four named a person. The share changed a lot by trade and by robot.
- Chapter 8Data set
Made-up firms with gloomy names got gloomier words
On made-up firms, a gloomy word in the name came with gloomier words from the robots. After one letter changed, most of the gap was gone.
Part Three
One word, one language: how the question changes the answer
- Chapter 9Preprint
A tone score changed more with language than company
Most answers got a tone score of zero. Where scores differed, they went with the language of the question far more than with the company.
- Chapter 10Preprint
When language and place both changed, the names changed
Local names came up far more when the question used the language of the company’s head office and named the area around it.
- Chapter 11Preprint
Same question, other language, other names
Check each language your customers use, because the names can differ.
- Chapter 12Data set
Gift lists differed for husband and wife
The husband and wife answers named many different brands. Robots also differ a lot in how many brands they name, and some answers name almost none.
Part Four
Before you believe a number
- Chapter 13Data set
One count can fool you. Mine did.
A number about AI answers comes from a counter, and a counter can be wrong. Two of mine were, and their numbers looked fine.
- Chapter 14Preprint
A robot’s report card can flip with the question
On gift questions, Gemini named more brands per answer than GPT. On business questions, GPT named more.
- Chapter 15Preprint
Robots share a taste for robot writing
In the authors’ test, robot judges picked robot-written product text far more often than people did.
- Chapter 16Preprint
Pretend shoppers give you questions to try
A pretend shopper is an invented person on a card, written by a robot. Its question can help you look for a need your list missed.
Questions
Is the book free?
Yes. Every chapter is free to read on this site. The PDF is free too: give your email address and the download link arrives by email.
Who is it for?
Dmitrij wrote it for his children, who are 9 and 12. It is also for anyone who runs a business and wants to know what AI assistants say about companies.
Why does the book call AI assistants robots?
To keep it simple. In this book ChatGPT, Gemini and Perplexity are all called robots.
Where do the numbers come from?
From the studies the chapters describe. Each chapter page ends with its reference from the back of the book. It links to the paper or its data and says what kind of source it is, such as a journal article, a preprint or a data set.
Does Rankfor.AI sell anything linked to this?
Yes. Rankfor.AI sells businesses tools that measure what the robots say, content written for them to read, and publishing where they read it. It also helps publishers sell sponsored content. Dmitrij also sells a book on this subject. Rankfor.AI does not promise a place in AI answers.
Keep asking
I started this book because my children wanted to understand what I was working on. You have met robots choosing toys, naming companies and filling boxes with pretend shoppers. Behind those pictures is a question a business owner can ask: “What would a customer hear about us?”
You can begin with a question a customer has already asked you. Put it to a robot. Read the answer. Ask again in a fresh chat, then try another robot. Keep the answers, so you can compare them.
Look at the names. Who keeps coming up? Who did you leave out of your own list? Read the reasons too. If a report gives you a score, open the answers behind it and see what was counted.
Then choose something useful to work on. Perhaps your website makes it hard to tell what you sell. Perhaps an answer repeats an old fact. Make the information clear and correct, and come back to the same questions to see what changed.
My children started this with their questions. I hope the book gives you a few of your own.
Pick a question that matters to your customers. Keep asking, and keep reading the answers.
About me
The real me
I am Dmitrij Żatuchin. For more than twenty years I have built technology companies. I also do research, and I teach.
I founded Rankfor.AI and I run it. There I study how AI assistants describe and name companies, and I work with marketing teams on what to do with those findings.
Before that I founded and led DO OK, a software services company with a team of around sixty people. It was listed on Poland’s NewConnect market and recognised in Deloitte’s Technology Fast 50 Central Europe. I also co-founded SoDA, Poland’s association of IT services employers.
I hold a PhD in computer science from Wrocław University of Technology. I was a lecturer and researcher at the Estonian Entrepreneurship University of Applied Sciences, and I led the Data-Based Decision Making and Leadership microdegree at Estonian Business School. I have been teaching business owners about AI visibility across the Baltics, Finland and Poland.
My book, Become the Answer, tells this story for marketing teams.
I live in Tallinn with my wife and two children, whose questions started this book.
Did you try the Monday steps? Tell me what you found.
References
Chapter 1. Robots make a list, and you can check it
Peer-reviewed, journal version (paid access). Dmitrij Żatuchin (2026). The synthesised consideration set: brand inclusion and concentration in large language model recommendations across five industries and three systems. Journal of Marketing Analytics, published online 31 October 2026. 10.1057/s41270-026-00562-3.
Free preprint. arXiv:2606.23057. Data, version 3.0. 10.5281/zenodo.22927143.
Chapter 2. Ask robots for lottery numbers, and tickets repeat
Preprint. Dmitrij Żatuchin (2026). Ask a Language Model for Lottery Numbers: Concentration in Repeated Six-of-49 Outputs. arXiv:2610.00052, version 1. arXiv:2610.00052v1. Data and code. github.com/Rankfor/rankfor-open/tree/main/research/lotto-models.
Chapter 3. One answer misses names, so ask again
Preprint. Dmitrij Żatuchin (2026). Measuring Brand and Source Discovery under Repeated LLM Queries: A Finite-Sample Audit. arXiv preprint, version 2, September 2026. arXiv:2609.05059v2.
Chapter 4. No single number of tries fits every question
Preprint. Dmitrij Żatuchin (2026). The Dice Roll Method: A Standardized Protocol for Repeated-Query Auditing of Large Language Model Brand Recommendations. arXiv:2609.04047, version 2, 14 October 2026. arXiv:2609.04047v2. Data. 10.5281/zenodo.22927143. Gift answers are listed in chapter 12.
Chapter 5. One robot’s named companies barely changed in 20 days
Preprint. Dmitrij Żatuchin (2026). Twenty Days Later: A Pre-Registered Test of History-Based Forecasts of Brand Mentions in Gemini Answers. arXiv preprint, version 1, 14 October 2026. arXiv:2610.19608. Data. 10.5281/zenodo.23351313. Register entry. osf.io/x8h3p.
Chapter 7. About one answer in four named a person
Preprint. Dmitrij Żatuchin (2026). Who Gets Named: Citation Type Predicts Individual Naming by Grounded Language Models, and a Roster Instrument Captures 0.5% of It. arXiv preprint, version 2, 1 August 2026. arXiv:2607.23893.
Chapter 8. Made-up firms with gloomy names got gloomier words
Data set. Dmitrij Żatuchin (2026). Does an AI describe your company, or your company’s name? 9,600 model answers about invented and real European firms. Version 1.0, 12 August 2026. 10.5281/zenodo.21904654.
Chapter 9. A tone score changed more with language than company
Revised manuscript. Dmitrij Żatuchin (2026). Query Language Sensitivity in Automated Sentiment Scores of LLM Responses. Original preprint. arXiv:2607.13304v1. Related aggregate data. 10.5281/zenodo.20794390.
Chapter 10. When language and place both changed, the names changed
Preprint. Dmitrij Żatuchin (2026). The Language Blind Spot: How Query Language and Brand Recognition Tier Shape AI-Constructed Brand Reputation Across Twelve European Languages. arXiv:2606.23165. Data. 10.5281/zenodo.20794390.
Chapter 11. Same question, other language, other names
Revised manuscript. Dmitrij Żatuchin (2026). Query Language and Network Exit Context in Generated Supplier Mentions: An Exploratory Audit. Original preprint. arXiv:2608.30052v1. Raw records. 10.5281/zenodo.22181306.
Chapter 12. Gift lists differed for husband and wife
Data for the reanalysis. Dmitrij Żatuchin (2026). Supplementary Materials: Volume and Distribution as Separable Dimensions of Gender Bias in Large Language Model Brand Recommendation. Zenodo, version 1.0. 10.5281/zenodo.21904788.
Chapter 13. One count can fool you. Mine did.
Data for the reanalysis. Dmitrij Żatuchin (2026). Supplementary Materials: Volume and Distribution as Separable Dimensions of Gender Bias in Large Language Model Brand Recommendation. Zenodo, version 1.0. 10.5281/zenodo.21904788.
Chapter 14. A robot’s report card can flip with the question
Preprint. Dmitrij Żatuchin (2026). System Attribution in LLM Brand Recommendations: Single Responses Identify the System, Aggregated Brand Profiles Do Not Transfer. arXiv, version 1. arXiv:2610.00253. Data. 574 of the gift answers are at 10.5281/zenodo.21904788.
Chapter 16. Pretend shoppers give you questions to try
Preprint. Dmitrij Żatuchin and Daniil Dzemesjuk (2026). Demand-Side Measurement for Generative Engine Optimization: Constructing and Validating a Million-Persona, Intent-Annotated Buyer Corpus. arXiv:2608.30023, version 1, 30 August 2026. arXiv:2608.30023.
Public sample. PersonaGen-15K, 14,955 pretend shoppers. huggingface.co/datasets/rankfor/PersonaGen-15K. Research use only, CC BY-NC-SA 4.0.
What does AI say about your business?
Put the book into practice. Tell us your company and website to request an AI visibility audit. We’ll reply by email to agree what to check.
What the audit shows
- Where your company appears in the answers, and next to whom
- The pages AI cites in your market
- What to publish first on your own site
