Which LLMO Book Teaches You to Get Named by AI Assistants?


Key takeaways
|
Somewhere right now an AI assistant is telling one of your buyers which companies to shortlist. It might name you, it might name three competitors, or it might describe your company using a positioning you retired two years ago. Most marketing teams have no idea which of those is happening, and that bothers me more than any ranking drop ever did. LLMO is the LLMO book I wrote to fix that blind spot, and it's Book IV of The Demand Doctrine™.
This post walks through what the book covers and the research behind it. I'll be candid about the timelines, because LLMO is the slowest engine in the system and the one most likely to tempt you into shortcuts. I'll also show where it sits next to the other three books, since getting named usually depends on work you did for the first three. Buying details are at the end.
What is the LLMO book, and who is it for?
The LLMO book is a short, practical guide that teaches B2B marketers how to shape what AI assistants know about their company, so assistants name it accurately when buyers ask who to call. It's Book IV of The Demand Doctrine™. It ends with a 30-day plan. Learn what LLM visibility means. |
It's written for the marketer who owns the brand story and has started to suspect AI assistants are telling a different one. Product marketing leads, heads of content and founders get the most out of it, along with anyone who has been asked in a board meeting why ChatGPT recommends a competitor. You don't need to understand how models are trained to use it. You do need patience, because this is the engine where results take months rather than weeks.

The book has five chapters, and each ends with key takeaways and an exercise that fits in under an hour. Chapter 1 explains what a model knows about you, Chapter 2 builds a distinctive point of view and Chapter 3 builds the public record. Chapter 4 covers author and entity links, and Chapter 5 sets up the prompt panel that measures all of it. The 30-day plan runs one theme a week, baseline, point of view, public record and finally identity and reporting.
There's one thing the book refuses to do, and I want to be upfront about it. It won't teach you to game a model with hidden text, fake reviews or a flood of self-published listicles. Those tricks are fragile, they're easy to spot and they can poison the very record you're trying to build. The book is about earning the mention, which is slower and lasts far longer.
What does an AI assistant actually know about your company?
An AI assistant answers from two sources, what the model learned in training and what it finds by searching right now. OpenAI's publisher FAQ says any public site can appear in ChatGPT search. The training crawler is separate. Learn how search grounding changes AI answers.
Chapter 1 separates those two sources because they move at completely different speeds. Live search can reflect a change you make within weeks, sometimes days. Training-based answers change only when a new model is released, and every model has a cutoff after which it learned nothing. If an assistant describes your company the way your old website did, you're usually looking at training data, not a bug.
The crawler documentation makes the split concrete. OpenAI separates GPTBot, used for training, from OAI-SearchBot, used for ChatGPT search, and lets sites control each independently. Anthropic documents a similar split, with ClaudeBot for training, Claude-SearchBot for search and Claude-User for user requests. Block the search crawler and you can drop out of that assistant's live answers entirely, which I cover in more detail in which AI crawlers to let in.
What shapes training data, the book argues, comes down to three things. Frequency is how often credible sources mention you in connection with your topic. Consistency is whether those sources describe you the same way. Timing is whether the mentions existed before the model's cutoff, and that last one explains why patience is part of the job.
Why does a distinctive point of view get you named?
A distinctive point of view gets you named because models repeat specific, consistent ideas and can credit them to a source. Generic category language gives them nothing to attach to you. A named idea does. Learn how entity authority makes a brand AI's default answer.
Chapter 2 defines a strong point of view as specific, defensible, consistent and relevant to a problem buyers care about. Most B2B positioning fails the first test before it reaches the second. If your homepage says you help organizations unlock value with an innovative platform, so does every competitor, and a model can't tell you apart. The book makes you write your point of view in two sentences that a competitor couldn't sign.
Then it asks you to name the idea and define it in one sentence, and use that name and definition the same way everywhere. I did this with the Engine Optimization Matrix™, which has one definition, four engines in a fixed order and five levers with citation as the payoff. It appears the same way on my site, in my books and in what I publish elsewhere. That repetition isn't vanity, it's what gives a model something stable to learn and credit.
The chapter's warning is about restlessness. Teams that rewrite their story every quarter scatter their signal across contradictory descriptions, and the model averages them into mush. Pick the idea, write the citable sentences and hold them for longer than feels comfortable. If you want the content side of that discipline, I wrote about building authority through thought leadership content.
How does a prompt panel measure whether assistants name you?
A prompt panel measures LLMO by asking AI assistants the same set of questions on a regular schedule and recording whether you're named and how. Because the prompts stay fixed, changes in the answers mean something. One run means little. Learn what share of model measures.
Chapter 5 is where LLMO stops being a vibe and starts being a program. Assistants don't publish reports on how often they mention you, so the only way to know is to ask, consistently, and write it down. The book has you build a panel of about 20 prompts and run each one three times on different days. Same wording, same settings, same assistants, every time.
The prompts come in four kinds, and each one tests something different. Write them the way your buyers would actually ask, not the way your messaging doc phrases things. Keep the competitor names real and the problems specific. A vague prompt gets a vague answer, and a vague answer tells you nothing.
Category prompts. Which companies help with a problem, or which tools are best for a use case.
Problem prompts. Why a problem keeps happening and how a team should approach it.
Comparison prompts. How you stack up against a named competitor or an alternative approach.
Identity prompts. What your company does, who it serves and what it's known for.
From those runs you track four numbers. Mention rate is how often you're named, accuracy rate is how often the description is right, idea attribution is whether your named ideas get credited to you and share of voice is how you compare with competitors. Every pattern becomes a specific fix, like a stale description to correct at its source or a comparison where a rival's page is doing your talking. And the book is explicit that you measure LLMO over 6 months, not 90 days, so nobody panics after the first run.
How do you build the public record that models learn from?
You build the public record by earning mentions in credible, independent sources that connect your name to your topic. Ahrefs found branded web mentions correlated at 0.664 with AI Overview visibility, versus 0.218 for backlinks. Mentions carry more weight. Learn how PR works for AI visibility.

Chapter 3 starts from a simple premise. Models learn from the whole public record, so what matters is mention density across sources the model has reason to trust. Your own site is part of that record, but it's the part that says nice things about you, and models seem to know it. Independent sources describing you accurately do far more work.
The book lists the ways B2B companies earn that coverage without a big agency retainer. Contributed articles in publications your buyers read, expert commentary for journalists, original research others will cite, podcast appearances and event talks all count. Each one connects your name with your subject in a source other than your own website. The trick is consistency, a steady cadence over a year rather than a burst around a launch.
The stakes keep rising. G2's March 2026 survey found 71% of B2B software buyers rely on AI chatbots somewhere in their research, and 51% start there more often than with Google. That means the public record isn't just shaping awareness anymore, it's shaping shortlists. And when you find an outdated or wrong description, the book's rule is to fix what you control first, then correct it at its source, which I walk through in fixing AI hallucinations about your brand.
Where does the LLMO book fit in the Engine Optimization Matrix™?
The LLMO book covers the fourth engine in the Engine Optimization Matrix™, large language models, which answer from training and live search. It comes last because it builds on the other three. Citation stays the payoff. Learn how AEO, GEO and LLMO differ.
Everything you did in the earlier books feeds this one. Pages that rank through AI SEO, answers lifted through AEO and citations earned through GEO all become part of the public record assistants learn from. That's why I tell teams not to start here, however tempting it is when the CEO screenshots a ChatGPT answer. The foundation work makes the LLMO work stick.
Chapter 4 of the book handles identity, which is where LLMO and structured data meet. Use one official name and one description everywhere, give your experts real profile pages and connect it all with Organization and Person markup that lists your official profiles in sameAs. Google's own Organization structured data guidance says this helps it disambiguate your organization. If you'd rather own all four engines in one volume, Mastering AI Visibility is the complete edition.
Where can you get the LLMO book?
You can get the LLMO book as a PDF or audiobook on The Demand Doctrine™ site, on Audible or as a paperback and Kindle ebook on Amazon. As of November 2026 the PDF is $5. The audiobook is $6.99 on the series site. Learn how to track brand mentions in ChatGPT.
The fastest route is the PDF from the LLMO page on the series site. Checkout runs on the site, the download is immediate and the page links to free posts on this blog that go deeper on each chapter. Run the Chapter 1 exercise the same day, asking three assistants the same five questions about your category and your company. Whatever you find will tell you exactly which chapter to read twice.
The audiobook is on the series site and on Audible, written and narrated by me. I recorded it myself after testing AI voice cloning and deciding it wasn't good enough to put my name on, which felt like the right call for a book about how AI describes people. The figures are described in the narration. It's a good listen for a commute, followed by an hour with the 30-day plan.
For print, the paperback and Kindle editions are on Amazon. As of November 2026 the paperback is $5.99 and the Kindle edition is $4.99, with the same 30-day plan and one-page cheat card inside. It's published under my own imprint, MQL Magnet, so I can update it as the models change. Given how fast they change, that may be the best feature the book has.
What else do people ask about the LLMO book?
What is the LLMO book?
LLMO is Book IV of The Demand Doctrine™ series by Harold Bell. It teaches B2B marketers how to shape what AI assistants know about their company so assistants name it accurately.
What does LLMO stand for?
LLMO stands for large language model optimization. It covers the content, identity and public record work that shapes how AI assistants describe and recommend your company.
How is LLMO different from GEO?
GEO focuses on being cited as a source when a generative engine writes an answer. LLMO focuses on what the assistant knows and says about your company, from training as well as live search.
How long does LLMO take to work?
Live search answers can reflect changes within weeks, but training-based answers change only when new models are released. The book recommends measuring LLMO over 6 months rather than 90 days.
What is a prompt panel?
A prompt panel is a fixed set of about 20 prompts you run in AI assistants on a schedule, three times each on different days. It tracks mention rate, accuracy rate, idea attribution and share of voice.
Does blocking AI crawlers affect LLMO?
It can. Blocking an assistant's search crawler can keep you out of its live answers, while blocking a training crawler only limits future training data, so check each provider's documentation first.
Should I read the other books before LLMO?
Ideally, yes. The work from AI SEO, AEO and GEO becomes the public record assistants learn from, so LLMO builds on it.
Is there an LLMO audiobook?
Yes. The audiobook is written and narrated by Harold Bell and is available on The Demand Doctrine™ site and on Audible.
Related guides
About the author. Harold Bell is the founder of MQL Magnet and has spent about 16 years in enterprise technology marketing, working with brands like Cisco, Rubrik and Nutanix in different capacities. He's a member of the Forbes Communications Council, the creator of the Engine Optimization Matrix™ and the author of Partner Over Product. |





