There has been a lot of buzz about Gemini 3, Google’s LLM. In this episode, I dig into Google’s big announcement and try to get past the hype to what it really means for healthtech marketers.
Google is positioning Gemini 3 as a highly multimodal, context-aware AI system that can handle text, images, data, and reasoning in one place. I will do my best to explain what that means in English and why you should care about that. I will also share how I benchmarked Gemini 3 vs the other guys to see if it lives up to the promise.
I also cover what this all means for search. This is kind of a big deal – potentially. This may presage the likely evolution of search from text-only answers to rich, AI-generated “micro-sites” with visuals, maybe even video, built on the fly.
I will wrap up with five key takeaways on when to use which tool (Gemini, ChatGPT, Claude, Perplexity), where Gemini really shines, and why Google’s evolving ad model should be on every healthtech marketer’s radar right now.
Key Topics Covered
- “(00:00)” – Introduction & setup
- “(01:10)” – What Gemini 3 actually is
- “(03:40)” – Nano Banana Pro for visuals
- “(05:30)” – Multimodal workflows & creative speed
- “(07:30)” – Deep integration with Google apps
- “(09:30)” – AI Overviews & the future of search
- “(12:30)” – Visual, interactive AI results & declining SEO value
- “(15:10)” – Rethinking Google Ads in an AI-first world
- “(17:00)” – Introducing the HLP BrAIn & benchmarking approach
- “(18:30)” – Benchmark results: BrAIn vs ChatGPT vs Gemini
- “(21:00)” – Script-writing test across four LLMs
- “(24:00)” – Strengths and weaknesses of each LLM
- “(26:00)” – Five key takeaways & closing
If you are interested in discussing this or any other topic, let’s have a chat. Reach out to me directly to schedule a no-obligation discussion. This isn’t a sales call, but rather an opportunity to talk through your questions and challenges.
Subscribe to The Healthtech Marketing Show on Spotify or watch us on YouTube for more insights into marketing, AI, ABM, buyer journeys, and beyond!
Thank you to our presenting sponsors, HIMSS, a leader in advancing health equity, digital innovation, and data-driven care through technology, policy, and community collaboration. And also HealthcareNOW, 24/7 expert shows, interviews, and podcasts, powering healthcare leaders with innovation, policy, and strategy insights.
Full Episode Transcript
Adam Turinas: [00:00:00] Hello and welcome to the Health Tech Marketing Show. I’m your host, Adam Turinas. welcome to the show. if this is your first time listening, I want to thank you for joining us and you can subscribe to this show on any of your favorite podcast apps. And we’re also on YouTube as well. I wanna give a big thank you to our sponsors, the people at HIMSS who’ve been a long time supporter of what we do, and also the wonderful people at Healthcare Now Radio.
Thank you for all your support and all that you do for us. Really appreciate it. So I want to talk about something which you’ve probably seen all over LinkedIn. Certainly been in the media. It’s kind of a big deal. And that’s Google’s announcement in the last few weeks of Gemini three and nano banana.
is kind of interesting.
I wanna get beyond the hype cycle, so I’m going to talk a little bit [00:01:00] about.
Way that Gemini, by the way, that Google is talking about Gemini three. So break that down for you so you can understand what the strategic themes are and what to expect out of Gemini. And then I’m gonna share, what I’ve been doing myself with it, which is I’ve been doing some experimentation with it and particular.
Benchmarking it against chat, GPT and some of the other applications out there. And I’m gonna kind of conclude for you what I think the pros and cons are of Gemini and how it fits in compared to the other LLMs. Okay. So let’s talk about the way that, what, Google is saying about it. So, Gemini three is Google’s newest.
AI system and multimodal basically means it handles text, images, data. reasoning. So in [00:02:00] that sense that’s a really cool new functionality. It’s very, very important. And compared to earlier versions, it’s much better at reasoning. It has much stronger world knowledge and it has huge context awareness.
And so a lot of people are talking about this as being really a, big leap forward in. LLMs and how, you know, in many regards, it’s actually better than chat, GPT or any of the other LLMs. But I’m gonna share some of my own experiences with it where I can see some things, which I, I was, surprised, I was surprised.
A little disappointed in some regards actually. so. you know, what you should expect is that, you get a better quality of output.
So rather than say, you know, a, a high school essay, you get a college level essay, maybe that’s, one way of thinking about it. and so it really means, You know what you should expect as a health tech marketer is that it [00:03:00] understands context better, so it understands the difference between. Clinical workflow and business workflow, for example, between what it understands what prior auth means. It understands what CMS is. and you know, as we know, right, context is everything in healthcare.
And so I, I guess it’s better. I hadn’t really benchmarked what compared to what it is was before, but I think as you’ll see later, I don’t think it’s. In terms of expertise on healthcare, it’s everything that it, they say it’s cracked up to be. one of the really cool things that they’ve announced is nano banana.
So while it’s got a pretty silly name, it is a pretty cool tool. and so what, Nana Banana Pro is, is their image generator and, paired with Gemini three. It’s really very powerful and that is actually pretty impressive. I’ve certainly found in the past that. Using the image generators for [00:04:00] things like chat GPT, they really don’t work very well.
you have to kind of use things like midjourney. maybe it’s just the way I’ve been writing the prompts, but I’ve just been pretty, disappointed with it. Uh, Nana Banana Pro, does it really well. And, that’s pretty, pretty impressive. So what you can expect to be able to do with Nana Banana Pro is create diagrams, infographics, user flow charts, mockups, visual explainers, and it’s also creates legible texts, which.
As I understand, it’s actually something that a lot of the LMS have struggled with, and it can do it in multiple languages. So you actually can do multilingual visuals, which is, you know, may, I dunno if it’s that important in healthcare, but you know, maybe if you have a Hispanic audience as well as an English speaking audience, that’s important.
But, you know, that’s pretty cool, I guess. according to Google, it’s shockingly accurate. And, when you pair. Nana Banana with Gemini Threes reasoning engine. You get a creative [00:05:00] system that can produce a full explanation and the visuals to go with it, and they, and it seems to actually.
Push that quite actively when you, you know, when you use it, just in a simple sort of text way that you might use chat, GPT. I think that’s pretty cool thing. The ability to create visuals easily is, compelling. So, let me kind of break down.
Some of the, three key things that they talk about, and those three things are multimodal workflows, how Google integrates it with all the other applications. And the third thing is, what it means to you as a creative. Producer, so of somebody who’s actually creating creative, what it actually means.
And then there’s a kind of a I’ll talk a little bit about afterwards about what the impact, what that means in terms of search and potentially what it means in terms of page search and Google as an advertising platform. So let’s just talk a bit first all about, I’m gonna go into a little bit more [00:06:00] detail on this one is multimodal workflows, because that’s kind of the biggest deal here in my view.
So you can generate your copy, your visual diagrams, your campaign structure, all in one shot. And it’s not just, you know, you don’t actually have to change LLMs to do it ostensibly. You can do all of this with Gemini. You’re gonna have to kind of play it out to see whether it lives up to the promise.
I’m still a little bit on the fence, frankly, as to whether or not it can do it, but the way they’re certainly designing it, that way. So you know, that means that you can get faster creative briefs, faster drafts, smarter edits, and you also can create a more consistent voice. So it actually makes it easier.
Gemini Met three apparently makes it easier to load up your brand voice and have that reflected. So if that’s something which, you know, I found is a bit hit or miss with the other LA Lambs. And so, you know, if you are building, you know, an infographic like say for Care Pathways Nano Can Drive, can create your first draft of [00:07:00] that in seconds.
If you give it a really good text prompt your designers can spend time refining what the output of Nano Banana Pro rather than starting from scratch. So that’s pretty cool. I thought, you know, ostensibly it also is good for thought leadership. it’s good for synthesizing complex material and, structure so you know, you should be able to build better white papers faster, and better LinkedIn content.
as I’ll come on to later, I’m not so sure it is that great at it actually. and I’ll explain, I’ll, I’ll tell you a bit about the bench, where benchmarks here. So the second big thing about it is Google is integrating this with its own applications. So that means it’s much easier to integrate it with Gmail, with Google Docs, with Google Sheets.
with Chrome, with Android and you know, I haven’t quite thought through all [00:08:00] the possibilities there, but I’ve certainly find it frustrating with the other LLMs that you have to kind of copy and upload other documents. unless you’ve got it well integrated with, some of your applications.
But if Google is actually doing this in a much more seamless way than the other LLMs, it does give them a, an advantage. So if you are, what an organization that uses. A lot of the Google applications, Gemini three could be actually a much better LLM for you to use because, you know, you can pull data out of your Google sheets, you can pull, communication out of your Gmail.
You can pull, reports out of your Google Docs, out of your, Google Drive rather. And, it can process all of that quickly and intelligently in any way. That’s, the promise. so the thing, it’s, it, it says it’s better at, and some of the benchmarks seem to suggest is, is that it’s faster, which means that you can produce.
Work faster. I don’t really know, [00:09:00] you know, if it is, I, found it to be, you know, when I was doing my benchmark tests, didn’t find it to be sort of like so dramatically faster that I would stop using any of the other lms. But maybe if you do really complex coding or things like that, you get an advantage there.
Certainly they’re sort of. the benchmarks that they’re reporting seem to indicate that. I think the most interesting thing is actually what’s behind this? And that’s Google’s strategy here. So, you know, Google’s cash cow is Google Ads, right? That is, you know, Google makes money in lots of different ways, but still its biggest revenue generator is what it makes from advertising and as.
Consumers and B2B customers shift their searching behavior away from searching on Google to using LLMs. And we’ve talked about this, you know, ad nauseum, right? So you just need to check out several of the past episodes. We’ve gone [00:10:00] into that in depth. So as people shift towards using the LLMs for doing.
For doing research, it’s a big problem for Google because it means they’ve got fewer eyeballs to monetize and it means that you are less, you know, it’s you as an advertiser, you are less likely to see the value. Of Google Ads and what they seem to have been doing. I, I don’t think they’ll admit to it, but it’s only looks like it to me, is that they’ve just been increasing the, number of sponsored links that you get in search results as a way to counter that.
But that’s not sustainable. So they do have a, you know, they’ve got a serious strategic issue and they’ve fully embraced it. They’re not in denial about it. And so I think that. If you, you know, you, I’m sure you’ve seen AI overviews, right? Whenever you, you know, for most searches these days on Google, you get an AI overviews that actually is powered, as I understand, it’s powered by Gemini.
And so with Gemini three powering it, it means that the [00:11:00] output of it is gonna be much richer and there’s some really interesting stuff. Sort of between the lines here. So I was recently, listening to one of my favorite podcasts, which is this old marketing show. I’m sure many of you’ll know. It, it’s, you know, it’s been around for a long time.
they’re so smart, those guys, and they went quite deep into something that Google is hinting at. And some stuff, in my view is really interesting. And I think the notion here is that. In AI overviews, Google is gonna move from just having text-based results to having stuff which is much results, which are much more interactive, are much more visually engaging.
I mean they, written a publisher white paper on this that surprise, surprise that AI generated visuals are much more satisfying to search people searching than text. I mean, that’s, you know, that’s not surprising. The old adage picture’s worth a [00:12:00] thousand words. And so, you know, the idea here is that rather than just getting, you know, an essay that responds to your prompt, you might get like a little micro size, which has got visuals.
It maybe eventually even has little, little videos. and some of that may be pulling from other sources, but. what these guys on the, this old, marketing show we’re hinting at is, that the likelihood is, is essentially it creates original content on the fly.
And that’s certainly because it’s multimodal and because it’s able to process data a lot, lot faster. That actually technically should be possible. And that gets really interesting because, you know, as I’ve talked about this a lot there’s a decline in SEO driven traffic, right? People are not, clicking through from Google to go to your websites or your competitors websites.
They’re staying on Google because they’re getting the answers on AI overviews, and if they [00:13:00] make the experience much more engaging, that is just gonna get worse. So for us, you know, it means less traffic, it means less opportunity to get people to come to our website. And I think you know, that we’ve gotta think about that as marketers because, what it means is that.
We start to think of Google rather than a place where you want to get, you know, to get the top links, either through paying for it or through SEO. Through thinking about Google the way that you think about traditional media companies in the, you know, you don’t, people don’t tend to click through from TV commercials.
I know. You know, they’re much more interactive with Direct TV and QR codes and things like that. And so, that’s all great, but it’s not a huge part of advertising revenue. And so effectively, Google then becomes kind of a notion of, you know, it’s an old, a term I, I heard very early on in my digital days back in the nineties called the [00:14:00] notion of a contextual envelope.
So, you know, you go to Google, let’s say, you go to Google. to research, patient engagement or patient behavior, something like that. If you are, you know, if you’re a, one of the people you know, uh, you’re a CMO of a, healthcare system, so as a marketer. Right now you’d be buying AdWords related to patient engagement, patient behavior, maybe adherence to a, a medical program or a, you know, to a, a regimen of some kind.
and so that when the buyer is using Google to get the answer to the questions, they are gonna click on, hopefully click on a ad, which will link ’em back to your website saying the not too distant future AI overviews will evolve. To presenting a much more engaging article. Maybe it’s a, a buyer’s guide created on the fly.
Maybe it’s a video, maybe it’s an infographic. maybe it’s, processes a lot of different research from different [00:15:00] sources to give the buyer the answers, to all of their questions right there within Google. And so you as a marketer. I’m gonna need to place ads there. And just a simple text link, you know, with whatever it is, like 20 words or however many loads you, you allowed these days with Google ads isn’t gonna cut it.
You are gonna have to produce commercials, you’re gonna have to do ads with offers. And so I think it, it’s quite interesting in terms of what it means to us as, as advertisers and, in terms of the timing of when this is gonna happen, it, it’s not clear, but it could be, we start to see this within the next year.
I don’t think that’s outta the question and sort of certainly the way that Google’s hinting about it would suggest that’s possible. and so as marketers, I’d suggest you really start to think about this. So you gotta think about. A-E-O-G-E-O-A-I-O, whatever [00:16:00] initials you use for this in terms of getting your content cied in the answer engines.
But you also are gonna have to think about what Google is doing with a IO overviews. And the type of advertising that you are going to need to develop to get people at the right time. And so how do you, you know, how do you create ads that are tailored to different prompts? You know, it’s not just about sticking a single banner ad up, I mean, maybe it is, or a single video add up.
You’re gonna have to create stuff which is tailored to the prompt to do it, you know, to really take advantage of what the medium is. Okay, let, let’s move on to a different topic. And this is, my own personal benchmarking of Gemini three. And so, you might have heard, if you listened to one of our pastor episodes, a hundredth episode that we’ve developed an AI system, an application if you like, called the [00:17:00] HLP Brain.
And, this is not publicly available. We’re just using it internally, but, you know, we we’re looking, we’re toying with the idea of of making it available to clients, in the new year. once we’ve got it sort of posh up a little bit, but the whole idea behind it is to AI enable sort of deep healthcare marketing expertise and particularly.
Expertise and knowledge. of the buyers. And so basically, the way we’ve done it is we’ve created, this way of storing data, for AI called vector storages, where you’re essentially embedding chunks of data that makes it easier for the GPTs, the LLMs, to process that data.
What we’ve been doing consistently over the last couple of months, and we do this on a weekly basis, is to benchmark how the brain is performing versus chat. GPT. [00:18:00] we’ve just done it against chat, GPT to start with. We’ll probably expand it. And to do that we’ve created a set of three different tasks.
So one is creating a persona narrative. about a chief marketing officer, sort of a day in the life of a chief marketing officer. Second one is creating a blog post for a CIO. I think it’s about interoperability. And then the third one is a set of personal personalized emails are for a patient engagement solution targeting a chief nursing officer.
And these are really detailed written prompts. So there are sort types of prompts that you would use if you’re really serious about using the LLMs to create sort of thoughtful content, that leverages the depth of knowledge of the LLMs. And so what we do is, you know, we create the content using.
M So using chat, GPT, and then we use it using the brain. And then I did it this [00:19:00] time using Gemini. And the interesting thing was, that chat, GPT, is a good script writer. it’s a good, you know, writes, well, you know, if you didn’t have a lot of knowledge of the industry, you’d look at it and say, well, this is really smart.
Yemini didn’t, you know, wrote elegantly. It’s. It’s a really kind of really well written what, was out the output, but it was really not very expert. The brain produces significantly more expert content and I’ll give you sort of the numbers on it. And so we, when we have about 10 different criteria that we look at, so the brain was getting sort of eight to 10 out of 10.
On a variety of experts attributes, particularly on the sort of level of expertise and reflecting nuance about the persona, sort of use of, sort of, of language that’s specific to healthcare. chat, GPT, was [00:20:00] getting, sort of sevens, sixes, sevens and eights. So about 30 to 40% less expert. Gemini is getting between four and six.
And so, in our view, the brain, as we’re currently benchmarking, it is already 30 to 40% more experts in producing content for our market. Then chat, GPT and it’s at least two x more expert than Gemini. And you can read and you, when you read the, the output, you can see it, see it specifically.
the marketing is saying that, in terms of context. Gemini’s gonna outperform everything in our view. From a healthcare perspective, it’s not that great. and certainly against our benchmarks. Then the other thing I did, I did a second test, which is I thought, Hey, you know what?
be great to get, all of, I’m going to use the LLMs to help me create the podcast script. I actually got it right here. I’m, [00:21:00] I’m refer You can, if you’re watching on YouTube. And you can see my eyes go down and I’m, referring to the script quite frequently. So, I wrote a detailed prompt, which is I am writing a podcast about Gemini three.
I want you to start by doing the research and what Gemini three is about, what the pros and cons are about it, what it means for health tech marketers, et cetera. Sort of, you know, you can get a sense of what that was. It was, I don’t know, about a 200 word long prompt. and then I said, and then you, the second part of the prompt was create an outline for the podcast.
And then the third part of the prompt was then write a detailed script. And then there was a fourth part component, which is the script turned out to be an hour long script, which is, I think, don’t think anybody wants to listen to me blather on about this for an hour. So I said, reduce it down to 15 to, 20 minutes.
and I, I did it with Gemini three. I did it with chat, GPT, I did it with Claude and I did it with perplexity. it was sort of interesting doing this [00:22:00] because actually my go-to on these types of things, actually, Claude, and I was surprised what I found out. So the first one was chat.
GPT actually came out as the best. It was the best script writer and the most balanced, all round performer. And so I was genuinely surprised about that. I was fully expecting that Claude would do the best of that, and I felt the output, and certainly the benchmark. I then basically, I then also ran it through, back through all of the LLMs.
I got ’em to, basically all four of them to benchmark the outputs of all four of them to sort of see what happened. And, I actually got fairly consistent results. So they all kind of benchmarked. What their pros and cons are fairly objectively. And so they, so it actually created a fairly consistent benchmark, across all, four of the LLMs.
so again, chat, GPT, sounds like a real person in many regards. And, you know, with flows and transitions and [00:23:00] humor. I think it’s still, think it’s a bit corny. I don’t, I don’t feel comfortable reading it verbatim, but it’s pretty good and the flow’s pretty good. and as I said, I would surprise me, ’cause Claude’s always been my go-to for it.
I’ll come on to Claude in a minute. Gemini, great structure. it actually seems to be, its strength is actually in structuring information. It was very weak on healthcare depth as I mentioned before. So it shines on structure. Writes really great outlines. this is sort of clean segmentation of the information.
The tone is professional. Maybe a little corporate, but that’s also probably because I haven’t used Gemini three or Gemini before much, so it doesn’t know my voice, whereas, you know, chat GPT and Claude. You know, I’ve, I’ve been using it for a while so it does know my voice, so, You know, actually Gemini produced the best, outline quality.
interestingly, again, going back to the, versus the brain benchmark. This surprised me is, is [00:24:00] that, there’s a marketing effectiveness, criteria that, we benchmark, chat, GPT against the brain. And that’s the one where the brain does well, a little bit better, but not dramatically better.
Gemini three actually got a three out of 10. And then, the brain and chat GPT do significantly higher. and I think the reason is, is because it writes, Gemini writes like, it’s talking to a general audience, not an enterprise buyer. And you make it really clear who, and I made it really clear in the prompts, this is why I’m write who I want you to write it for.
And it still didn’t get it. So I don’t think for expert stuff that it’s, it’s that great. Claude, best professional tone and strategic thinking. So Claude writes like a McKinsey consultant, and, the output is polished, it’s articulate. It’s the most kind of executive briefing style.
So again, a little bit corporate, not quite, you know, not quite something that makes a particularly engaging pod podcast, [00:25:00] which surprised me. perplexity. If you’ve used perplexity, you probably won’t be surprised about it. It’s fantastic for doing research and, You know, really it just outperforms all the other ones.
But it writes crappy writer. I mean, it really, the script that comes out of it is just awful. and when it benchmarks itself, it says it’s just, that’s not, its, it’s not its thing. so, you know, phenomenal for research. Terrible, for storytelling. So, I thought this was interesting and again, I’ve got the, notes in front of me actually, in terms of, just going back to the first benchmark, the brain got a sort of score of 9.3 out of 10.
for the three tasks that I told you about earlier, the sort of writing content chat, GPT got 6.1, so not bad, but not great. Gemini got 4.8 and so. creating an expert source of content, sort of designed for the LLMs and really carefully curating it, not you gotta fill it with quality, not, just quantity.
the [00:26:00] brain is actually outperforming chat GPT by 53% and Gemini by 94%, which I just think is kind of staggering. Anyway. So, let me put it all together for you.
Chat, GPT seems to be the best storyteller, Geminis best for structure. Perplexity is, the best researcher. Claude has the best strategic tone and you know, based on our benchmarks, the brain seems to be so far the best at healthcare specific executive or sort of B2B content.
I wanna leave you with five key takeaways. So, number one. Gemini seems to be really good for structure an an organization. The second thing is the combination of multimodal power plus nano banana for visuals.
And VO three too, by the way, for video, means that you can have the potential to create really great. Dynamic content or sort of interactive, [00:27:00] engaging content much faster by combining the Google tools. And so we’re certainly gonna be playing around with that. ’cause that is, you know, that’s really cool for creative businesses like ours.
I think, you know, based on what I, what I, my own research, my own benchmarking, again, it was, you know, a limited test, but still, you know, using some, a pretty robust testing methodology. it seems to be weak on healthcare depth and understanding the B2B healthcare persona. and I don’t think it’s, I wouldn’t use it for executive level thought leadership.
I think the output of it is just too shallow. I’d still use Claude or chat GPT, but with a ton of different information, like, you know, accessing what we have in the brain to make sure that what comes out of it is expat. The last thing is the strategic shift, and I think this is the most interesting thing, which is this is SA signaling a major change in the way that Google is [00:28:00] gonna work as an advertising platform.
So I think that’s something that we all need to be thinking about. So, okay. I’m gonna bring it to a close now. I hope you found this useful. and if you did, please share it with your friends and colleagues it would mean a lot to me if you would do that. It would also mean a lot to me also if you give it a thumbs up on YouTube or positive racing on whatever your favorite podcast application.
So with that, I want to bid you farewell and say thank you for listening.
