The AI tool that only answers from what you give it

The tool I don't use nearly enough

Of everything I've written about in this newsletter, this is the one I underuse the most, and the more I dig into it the more I think that's a mistake. It's a free Google tool that most coaches have either never heard of or tried once and forgot about. It used to be called NotebookLM, and this summer Google renamed it Gemini Notebook. Same tool, new name, so if you've heard of the old one, this is it.

What makes it different from everything else we've covered is simple. It only answers from the sources you give it.

Why that matters in our field

Back in Issue #8 we talked about the biggest problem with using AI for research, which is that a general chatbot will sometimes hand you a study that doesn't exist, formatted perfectly, from a real journal. That happens because a general tool is drawing on everything it learned in training, and when it doesn't have a real answer it produces what a real answer would look like.

Gemini Notebook works the other way around. You load in the papers, position stands, articles, and transcripts you want it to work from, and it answers inside that stack. Every answer comes with a citation that links to the exact passage it pulled from, so you can click and read the actual sentence. And when you ask about something that isn't in your sources, it will usually tell you it doesn't know instead of filling the gap.

The research on AI in sports medicine backs this up. A review published this summer looked at 32 studies on AI tools in our field and found that the risk of made-up information was high for general chatbots and substantially lower in tools built to pull answers from a defined set of sources. That is exactly what this tool is.

It's also different from the Projects we set up back in Issue #2. You can upload files to a Claude or ChatGPT project too, but the tool will still blend in its own general knowledge when it answers. That's useful when you want its broader perspective. It's the wrong setup when you need to know exactly where a claim came from.

Inside the Lab: build one notebook on one question

Pick one question you actually care about. Not a whole topic like speed, but something specific you're trying to figure out. How to manage hamstring risk in-season. What the evidence says about deceleration training for your sport. How other programs structure return to play after an ACL. One notebook per question keeps the answers focused.

Gather ten to twenty good sources. Research papers, position stands from the NSCA or other governing bodies, long-form articles, clinic notes, and YouTube links to talks worth hearing, since it can read those too. Quality matters more than volume here, because the tool can only be as good as what you put in it.

Then ask questions across all of it. This is where it separates itself from reading the papers one by one. A few worth copying:

• "Where do these sources agree, and where do they disagree?"
• "What populations were these studies done on? Which ones used athletes similar to high school or college athletes, and which used untrained adults?"
• "Which claims are supported by multiple sources, and which come from a single study?"
• "What questions do these sources leave unanswered?"

Use the Studio tools on the side. It can turn your sources into a briefing document, a mind map, study guides, or an audio overview where two AI voices talk through your sources like a podcast. That last one is worth trying on the drive home, which is the same idea from Issue #7 about learning away from the desk.

What it looks like in the real world

Say you've got a folder of eleven papers on in-season hamstring work that you saved over the last year and never sat down with. You load all of them into one notebook and ask where they disagree. Within a minute you know that most of them agree on the general approach but differ on volume, and that three of the studies were done on professional soccer players, which may or may not describe your athletes.

That second part is the one that matters. You would eventually have figured that out reading all eleven papers, but it would have taken you a weekend you do not have. Now you know which papers to actually read closely, and you know it because you clicked through to the passages yourself.

Where the coach stays in charge

It only knows what you give it. If you load it with ten articles that all agree with each other, it will tell you there is strong agreement, because in that stack there is. The quality of the answers is capped by the quality and balance of your sources, and that part is still your job.

The audio overviews are fun and genuinely useful, but they are built to be conversational, and conversation smooths over nuance. Treat them as a way into the material, not a replacement for the parts that matter.

And the privacy rules from Issue #10 still apply. On a personal Google account, Google says your uploads are not used to train its models in normal use. The exception is the thumbs up and thumbs down buttons, since submitting feedback can send that content to human reviewers. If your school gives you a Google Workspace account, use that one instead, because the protections are stronger. And keep athlete information out of it entirely. This is a research tool, not a place for your athletes' records.

The Coach's Scouting Report

• Signal. A general chatbot answers from everything it knows. This answers only from what you gave it, and shows you exactly where every answer came from.
• Noise. Saving papers you never read. Trusting a clean answer without clicking through to the source.
• Next Rep. Pick one question you've been meaning to dig into. Pull ten sources you already have saved, load them into one notebook, and ask where they disagree.

From the Lab

If you want to take this further, I built a free workflow called Research to Podcast. It starts with one coaching question, builds a source-grounded research packet, and turns it into a Gemini Notebook podcast you can listen to on the drive, plus a slide deck or infographic for your staff.

And if you or someone on your staff is just getting started with AI, the AI Starter Guide for Strength Coaches is a $20 ebook that covers where AI fits in your week, what to hand off first, and how to build habits that stick.

See you next Sunday.