You have certificates, you have experience — and when you work with clients, you change lives. But then your website says something like “I help you step into your power” or “We lift your business to the next level”. Your prospects scroll on. Not because they do not need your coaching — but because they do not understand what you are concretely offering them.
Note (as of July 2026): Google has renamed NotebookLM to Gemini Notebook. Product, features and your existing notebooks remain unchanged — only the name is new. In this article I keep using “NotebookLM”, because the term is (still) the more widespread one.
The problem is your own tunnel vision. I call it astronaut language: you know what you mean by “transformation”. Your customer sees only empty word shells. So today we feed NotebookLM your coaching texts and see whether the AI understands what you are actually selling. Because: if the AI does not get it, your customer does not get it either.
Why this matters doubly in 2026: SEO was yesterday
There is a second aspect: if the AI does not understand your offer, it also cannot recommend you to the right customers. It is 2026 — your page is no longer read only by people, but above all by AI systems deciding whom to present your offers to. That is called context engineering: the AI assembles the information about you and considers which customer this could be interesting for.
- Back then: SEO. It was enough to be found in Google with the right keywords.
- Today: GEO. The AI has to understand what you offer.
- And: AEO. Your offer has to solve a concrete question your potential customers have.
Setting up NotebookLM — and a word on data protection
NotebookLM comes from Google and you can in principle test it for free — all you need is a Google email address. If you are at the very beginning, first read What is NotebookLM? — there I explain the basics step by step. But since we are in a business context, I recommend the paid variant via a Google Workspace: only then do you get the extended data protection, can decide yourself where the server sits and where your things are stored — and Google does not use your content to train the AI.
Still, please be careful: Do not upload sensitive or confidential data. Only what makes you say: this is completely fine. One hundred percent data protection will never exist on the internet.
Step 1: upload your texts as a source
Open a new notebook in NotebookLM. For this analysis you would ideally upload your own website as a source. For the video I prepared a typical coaching text, the kind I have read again and again on various pages over the past years — “3 months of mentoring, we lift you to the next level” — and paste it in via “Copied text”.
As soon as the source is in, NotebookLM immediately runs a first analysis and suggests questions. If you want it even more specific — NotebookLM should be a strategy partner with particular traits — go to the three bars at the top, choose Custom and enter your exact instruction there.
Step 2: the €5,000 test
Now comes the first prompt — and it packs a punch:
“If I transfer you €5,000 today, what exactly will I hold in my hands in 3 months? Name concrete, measurable results.”
Why this question? The AI has to understand what you do so that the customer understands what you do. You have to solve a concrete problem. And I know — we coaches, I am a coach too, we tend to revel in flowery, literary texts. But that solves no problem for the customer.
With my example text, the result was: concrete services are named, and also how they are delivered — but no concrete problem is being solved. And with that we have problem number one: the AI cannot tell whom to suggest this to. And the customer too is left with only a vague feeling: okay, what exactly am I getting here?
Step 3: identify the target group — hello, blind spots
Next, have NotebookLM read your target group out of the text: “Identify my target group for me.” Maybe your website already yields a much better answer — but maybe you too still have blind spots.
In my example: “affluent high performers”, plus a few buzzwords like energy work, human design, purely intuitive. All very soft phrasings. I do not want to attack that — it can be the optimal solution for many. But people do not understand it. Times have changed: today we need concrete address. And think of GEO and AEO — the AI too has to understand what you do.
Step 4: reverse engineering — turning mistakes into recommendations
Recognising mistakes is one thing. Drawing a consequence from them is another. So now we approach it from the back and make your communication better:
“How can I define my target group better in my text so it becomes clearer? What recommendation do you give me so I can better address the right people?”
“What tip, what recommendation do you give me to work out the concrete added value of my program? What can I concretely say, what concretely promise — what are the concrete to-dos for me?”
In case you are wondering how I enter such long prompts: I work with Wispr Flow (affiliate link) — an AI tool on my computer that I speak my prompts into. I think out loud. Via the link you can test the Pro version free for a month.
And you can see how I work with AI: in dialogue. A ping-pong game. The AI answers, and I treat it like an employee I tell: “Please explain that to me again.” A big problem many people have is thinking it all has to work at the snap of a finger. But the AI has to learn from you what you really want — which means you too have to learn to become clearer in your communication.
NotebookLM then names very concrete points: so the AI understands whom it can suggest your offer to — and so the people on your website understand what they are getting and can build a relationship with it. Many move far too much in the emotional realm alone and never pick up the logic. But it is precisely the logic we need to make a buying decision.
Step 5: secure the results — note and source
What do you do with the analysis now? Two things, right in NotebookLM:
- Save as a note — so the chat is preserved.
- Set as a source — this way you can use every discussion again and again as the basis for future conversations on this point.
The counter-test: does the AI even know you exist?
The analysis above checks whether the AI understands your text. But there is a second, more uncomfortable question: Does the AI even name you when someone asks about your service?
The test takes two minutes and you need no tool for it:
- Formulate the question your dream client would ask. Not your name — her problem. So not “Who is [your name]?”, but for example: “Who helps the self-employed sharpen their offer? Name concrete names.”
- Ask it in an AI chat with active web search. And see who gets named.
- Then ask the control question with your name: “Who is … and what does she offer?” Check the answer for two things — is the description still accurate, and are the offers named current?
Both results are revealing. If you are not named on the first question, that is not a technology problem — it is a clarity problem. And if an outdated offer surfaces on the second question, you know the AI is working with old information about you.
Do this test regularly, not once. The answers change, and only over time do you see whether anything is moving.
My own test — and the uncomfortable result
So you know I do not just recommend this: I ran this test on myself in July 2026.
I asked the first question from my dream client’s perspective — not a word about me, only her problem: “I am a self-employed consultant, 45. I have tried AI, but I just collect tutorials and never get into implementation. Who concretely helps solopreneurs integrate AI into their existing business? Name concrete names in the DACH region.”
The AI named ten providers. I was not among them.
That was unpleasant, and I am writing it here anyway. I have been doing nothing but this topic for years, run a YouTube channel about it, work with it every day — and was not named on precisely my core question.
The control question was even more revealing. Asked about my name, the AI did know me — but described a version of me that no longer exists: old phrasings and offers I have long stopped selling. The AI was working with a state from years ago, because those were the pieces of information it had found about me.
What I did with that: tidied up. One single, everywhere-identical description of who I am and whom I work for. Current offers stored so machines can read them. Outdated pages cleaned up. And the test set as a fixed monthly routine, instead of doing it once and hoping.
The lesson from it matters more than my result: Not being named has nothing to do with how good you are. It has to do with how clear you are. And clarity is something you can create — today.
SEO, AEO, GEO — what the terms really mean
I used these three abbreviations above, so let me explain them briefly, because they increasingly get mixed up:
- SEO (search engine optimisation) is the familiar one: your page should be found on Google when someone enters a keyword.
- AEO (answer engine optimisation) goes one step further: your content should answer a concrete question so clearly that it works as an answer — not just as a hit in a list.
- GEO (generative engine optimisation) is the new one: it is about generative AI understanding and recommending you. No longer just “be found”, but “be named”.
All three build on the same foundation, and it is called clarity. An AI can only recommend you to someone if it can say in one sentence what you do for whom. That is exactly why the analysis in this article is no marketing-cosmetics topic — it co-decides whether you appear in AI answers at all.
Clarity is the key to high prices
As a coach you are not paid for your time — even if many customers keep thinking that — but for the certainty of the result. NotebookLM helps you put that certainty into words before you go into the sales call.
Yes, it is hard to see your own “baby text” dissected like this. It hurts. But better an AI tells you than your bank account telling you in six months.
Your next step: 5 prompts for the AI check
If you are now noticing that your positioning still wobbles, I have something for you: a worksheet — from platitude coach to authority. In it you will find five killer prompts for having the AI examine your offer. Get it for free, run NotebookLM over it — and write in the comments under the video what the AI said about your business. I genuinely want to read it.
Frequently asked questions
What do SEO, AEO and GEO mean?
SEO means being found on Google for a keyword. AEO (answer engine optimisation) means answering a concrete question so clearly that your content works as an answer. GEO (generative engine optimisation) means generative AI understanding and recommending you — being named, not just found. The foundation for all three is clarity about what you do for whom.
How do I test whether AI knows my offer?
Ask an AI chat with web search the question your dream client would ask — not your name, but her problem, plus a request for concrete names. If you are not named, that is a clarity problem, not a technology problem. Then ask the control question with your name and check whether the description and the offers named are still current. Repeat the test regularly, not once.
What does NotebookLM cost — and what should I watch for in data protection?
You can test NotebookLM for free; all you need is a Google email address. For business I recommend the paid Google Workspace: extended data protection, selectable server location, and your content is not used for AI training. Still, the rule always applies: do not upload sensitive or confidential data.
Why should an AI of all things examine my offer texts?
Because it is your toughest and most honest critic: if the AI does not understand what you are selling, neither does your customer. And because in 2026, AI systems co-decide whom your offers are presented to — see GEO and AEO.
What do I do with the results of the analysis in NotebookLM?
In NotebookLM, save the chat as a note and set it as a source — that way you can use every discussion as a basis again later. And then: keep at it, ask follow-ups, keep working in dialogue. The AI is like an employee who has to learn from you.
Conclusion: better the AI now than the bank account in six months
Vague communication costs you customers — with the people who scroll on, and with the AI systems that cannot recommend you. With NotebookLM you make your blind spots visible: first the €5,000 test, then the target-group analysis, then concrete recommendations via reverse engineering. It costs you one afternoon and a bit of courage.
If you would rather not be alone in this: in my free community we exchange ideas about exactly these kinds of workflows. And if you notice that this is not quite enough and you want to work more deeply on your business, have a look at the AI Business Community — or let us simply talk.
