Direct answer: there is a free Google tool, NotebookLM, that answers questions using only the documents you upload, and it shows you the exact spot in the source where each answer came from. Almost everyone uses it to study faster. Here is a far more useful job for it if you run a practice, a clinic or a firm: use it as a mirror. Feed it your own material (your website copy, your services, your prices, your FAQs) and ask it the questions a client would ask. If NotebookLM can't give a clear answer using your own information, that is your signal that ChatGPT and Gemini, the AI already deciding whether to recommend you in front of a client, can't either. The good news: this is fixable, and fixing it is the front door to getting recommended by AI.
First, what NotebookLM actually is (one minute, no hype)
NotebookLM is a free tool from Google. The logic is simple: you upload documents (a PDF, a Google Doc, your website text, your FAQs, an old proposal) and then you ask it anything. The difference from asking ChatGPT the same thing cold is the whole point here: NotebookLM answers only from what you uploaded, and every answer comes with a citation to the exact passage it pulled from (Google, NotebookLM). It does not invent out of thin air, because it only works with your material.
It can also build useful things from those documents: a podcast-style audio overview, a set of FAQs, a study guide. And one honest limit worth saying up front: NotebookLM only looks at what you give it. It does not connect to your CRM or your email or pull live data from anywhere. That sounds like a weakness. It is exactly what makes it such an honest mirror.
The turn: NotebookLM is not a productivity trick, it is a way to see yourself the way AI sees you
Here is the idea almost nobody hands you. Every tutorial sells NotebookLM as a productivity trick: summarize, study, save time. Fine, it does that. But for a professional who lives on being chosen, it has a much more valuable use.
Think about it this way. When your client asks ChatGPT "who's the best orthodontist in Coral Gables" or "which immigration attorney should I use in Miami," the AI answers with the businesses it can understand from whatever is written about them. It does not recommend you because you are nice or because you have been around for years. It recommends you if it can read you and understand you.
NotebookLM lets you rehearse that exam in private, with your own documents in hand. You upload what you publish, you ask what a client would ask, and you watch, with your own eyes, whether your information is enough to produce a clear answer. It is the lab version of the test the public AI is already giving you out on the street, every day, without telling you.
And here is the line worth writing down: if NotebookLM can't answer your client clearly using your own paperwork, the AI that recommends you can't either. Nobody checks the mirror before they leave the house and then wonders why the meeting went sideways.
Why this is pure GEO (and why it pays off for you)
GEO is getting AI to understand you, trust you, and cite you when someone asks. The first of those three, understanding you, is the foundation. If the AI can't understand you, the other two never even start.
The problem is that "being understandable to AI" sounds abstract. Nobody knows whether they are or not. That is where the mirror comes in: NotebookLM turns something invisible ("does AI understand me?") into something you can see on screen in ten minutes ("I asked it my price and it couldn't find it"). It stops being a theory and becomes a concrete test you can run yourself today.
And the timing is not an accident. 37% of searches now start on an AI platform (Eight Oh Two, 2026), and traffic arriving from language models had been growing 527% year over year (Previsible, 2025). More people ask AI who to hire every week. Being understood stopped being a technical luxury and became the difference between sitting inside the answer or watching it name your competitor.
The mirror test, step by step (10 minutes, buy nothing)
You do not need to be technical. Anyone with a Google account can do this.
- Go to notebooklm.google.com and create a new notebook.
- Upload three to five documents you already have: your website text, your list of services, your prices or an old proposal, your FAQs, your warranty or first-visit policy.
- Ask it, one at a time, the real questions a client asks. For example: "What does the first consultation include and what does it cost?", "How is this clinic different from another?", "Do you see patients in Spanish?", "How fast do you respond?"
- Do not just read the answer. Watch how it answers: does it reply clearly and show you the citation, or does it ramble, fall short, or flat out say it can't find it in your sources?
- Write down every question where it stumbled. That list of stumbles is gold: it is exactly what your client and the AI can't find either.
If you want to squeeze more out of it, ask this so it leaves you something built: "Using these documents, write the 8 questions my ideal client would most likely ask, with an answer to each ready to publish, and tell me what information you are missing to answer better." In two minutes you have a draft FAQ and, more important, the list of your gaps.
How to read the result: the three diagnoses
When you finish, your result lands in one of these three. All three are good news, because all three tell you exactly what to do.
One: it answers clearly, with a citation. Your information is understandable. Congratulations, you passed the rehearsal. Your job now is not to write more, it is to distribute that where the public AI reads it: your website, your Google profile, the places AI goes to look. You have the raw material; it just needs to be in the right window.
Two: it answers halfway, or too generically. Your information exists, but it is buried in long paragraphs, written in pure adjective ("excellence," "commitment," "quality"), or scattered. The AI has it right in front of it and still can't pull a clean answer. The fix is structure: direct answers, your specialty and city stated plainly, concrete facts instead of filler.
Three: it says it can't find it. The fact simply is not written down anywhere. And this is the most valuable finding, even if it stings: if it is not written, the AI can't cite it, and that is why it names someone else. It is not that the AI dislikes you. It is that it has nothing to say about you.
The honest objection (before you think it)
I will say it for you, because in this shop honesty runs the place: "Todd, so if I pass the NotebookLM test, I show up in ChatGPT?"
No. And anyone who tells you yes is selling you smoke. NotebookLM is a lab, not the street. It only looks at the documents you upload, so passing the rehearsal does not automatically put you in ChatGPT's or Gemini's answers. Showing up out there needs the other two legs of GEO: the AI trusting you (reviews, mentions, signals that you are real and good) and your information being published where the AI reads it, not sitting on your laptop.
But flip the coin and you will see why the exercise is worth so much: failing the mirror test does guarantee you fail outside. If the AI can't build a clear answer about you from your own documents handed to it on a plate, there is zero chance it does so in the real world, where it competes with everyone else's noise. It is necessary, not sufficient. Passing the mirror does not guarantee the exam, but failing it guarantees you flunk.
The Todd and Naty corner
Todd: What I like about this is that it kills the mystery. People hear "GEO" and assume they need an expert and a budget. They don't. In ten minutes, with a free tool, you see for yourself whether the AI understands you. That is a diagnosis you normally pay to get.
Naty: And on my side: it is empowering. There are excellent professionals who are invisible to AI for one reason, nobody ever wrote their information where a machine could read it. That is not a talent problem. It is a paperwork problem, and paperwork you can fix.
What to do this week
- Run the mirror test with your real documents, not the ideal ones. Upload what you actually have published today.
- Ask the five questions clients ask you most. Note which ones NotebookLM stumbled on.
- Sort each stumble into one of the three diagnoses: understandable, buried, or missing.
- Start with the "can't find it" ones. Those are the facts neither your client nor the AI can see. Write them down.
If you want the exact playbook we use with clients to go from the mirror to the real world (the full checklist for getting AI to understand you, trust you, and cite you, in English and Spanish), text us the word UNDERSTAND on WhatsApp and we will send it straight over: https://wa.me/17866869497?text=UNDERSTAND
FAQ
Is NotebookLM really free? Yes, Google offers a free version. You upload your documents and it answers only from them, citing the exact spot each fact came from (Google, NotebookLM). There are paid tiers with more capacity, but the free version is plenty for this test.
Why use NotebookLM and not ChatGPT for this test? Because they do different jobs. ChatGPT answers from what it knows about the whole world, so it can paper over gaps with generic information and give you a false sense that everything is fine. NotebookLM answers only from what you upload. If the fact is not in your documents, it tells you. That is why it works as an honest mirror: it does not cover your gaps, it shows them to you.
If NotebookLM understands me, do I already show up in ChatGPT? Not automatically. Passing the test means your information is understandable, which is the foundation. Showing up in ChatGPT's or Gemini's answers also needs that information published where the AI reads it, plus trust signals. Passing the mirror is necessary, not sufficient. Failing it, on the other hand, does guarantee you do not show up.
Does this work for a practice or clinic, or only for big companies? It works best for the professional. A dentist, a lawyer, a doctor or a coach lives on being chosen, and more people choose by asking AI. The mirror test is just as valid for a one-person practice as for a large firm. In fact, the smaller the team, the faster you can fix it.
Is GEO the same as SEO? No. SEO is about ranking you in Google's list. GEO is about getting AI to understand you, trust you, and cite you inside the answer. Different systems, different signals. We break it down in full in our map of the 5 disciplines of search.
