Heads In Beds Show - Can We Build A Marketing Strategy With Just AI?

Episode Date: January 15, 2025

In this episode Conrad and Paul run a fun test: can we build a marketing strategy that makes sense with ChatGPT and / or Gemini from Google? Enjoy!⭐️ Links & Show NotesPaul Manzey Conrad O'ConnellConrad's Book: Mastering Vacation Rental MarketingConrad's Course: Mastering Vacation Rental Marketing 101ChatGPTGemini🔗 Connect With BuildUp BookingsWebsiteFacebook PageInstagram🚀 About BuildUp BookingsBuildUp Bookings is a team of creative, problem solvers made to drive you more traffic, direct bookings and results for your accommodations brand. Reach out to us for help on search, social and email marketing for your vacation rental brand.

Welcome to the Head to Med Show presented by Buildup Bookings. We teach you how to get more vacation properties, earn more revenue per property, master marketing, and increase your occupancy. Take your vacation rental marketing game to the next level by listening in. I'm your co-host, Conrad. I'm your co-host, Paul. All right, Paul, how's it going? Happy New Year to you. What's going on? level by listening in. I'm your co-host Conrad. I'm your co-host Paul. All right, Paul, how's it going? Happy New Year to you. What's going on?

Happy New Year indeed. It's, you know, 2025. I feel a lot of good stuff coming here in 2025. This is, we're still following sports closely. I'm still following sports a little close more closely than you are on the on the fall side of things with football. But we've got basketball to enjoy here as the I'm still following sports a little close more closely than you are on the on the fall side of things with football but we've got basketball to enjoy here as the Year continues to progress but it's this is kind of that weird time of year where it feels like everybody's Just gotten off of the they've been they've had a 45 day sprint since about Thanksgiving. So

Maybe a little harder to get a hold of people these weeks here, but it is it's a good launching point offer into 2025 here. So how are you doing, sir? Yeah, pretty good. Not in the sports realm necessarily, as you know, it's been a to say that the Patriots are limping in as a bit of an understatement. I guess I realized they're trying to lose at this point now for the whole you know, but it feels like now this is the second year row of doing that. And I guess the logic, the logic is that if you get a high pick that automatically that high pick will high picks, I should say, because you're at the top of the future rounds, right? Will turn out

to make you win a lot of games. And in theory, the Patriots have done that correctly with Drake may, but it's not really turning into better results on the field. So that's where there's some concern, you know, seeping into my body. I feel like that still was the right choice. And I still wouldn't go back and rerun it or do it differently given the options available at the time. But that's the funny thing about sports, you just never know, you know, and sometimes it takes multiple years to kind of see the fruition of things. And you've got to get sports is interesting. It's kind of like

business in a way where each piece of the funnel or each piece of the process has to be set up correctly. And if you have one spot that's really weak, you will get exposed, right? If you've got good players or mediocre players, even but bad coaching, like you're gonna be really bad. If you have good players and good coaching, you're gonna be at least okay, you know, and if you have great players and great coaching, then of course you can, you know, win and take it all the way to the to the finish line as my old Patriots used to do so often, but they're for discussion. Well, speaking of change, I guess one thing that's definitely changed, you know, in the past few years is AI.

And I guess, I guess this is probably, you know, every few podcast episodes, we end up touching on AI, whether directly or indirectly. But this one's very directly because we came up with an idea on generating marketing strategies with AI and comparing it or giving it our kind of our grade on it. If that was given to us, and we were the one actually having to execute the marketing strategy. So here's kind of the premise of today's episode for the listener to kind of get tuned into what we've been thinking about. There's all these AI tools out there. Like Paul and I are different in many ways, similar in some ways, many ways, but also different in many ways. We've kind of chosen, it seems like, two a little bit different AI platforms for our day-to-day use. Paul's been

a little bit more in the Gemini train. I've been a little bit more on the chat GPT train, which is okay, because I think there's good stuff coming out of both both those tools. That's kind of where the leans have been so far. And not only that, but we have different slightly different backgrounds, at least over the last year or two, where Paul's done a little bit more in the homeowner marketing side. I've done obviously more on the guest marketing side of things. So we've got some different perspectives. So the thesis today was we're going to have these AI tools make a strategy based on prompts, and we're going to compare the outputs or compare the strategies that the AI tools gave us and see what we think of the different outputs. If they're good, if they're bad, if they're giving you the right direction. If we think that this is something that's eventually going to replace us,

then we're going to be working bread lines or chilling somewhere unemployed. That's a joke. Well. Yeah, maybe. Did I set it up properly, Paul, did I, yeah, maybe, did I say, did I set it up properly? Paul, from your perspective and. Perfect setup. I thought I truly that this really is. I mean, this is, this was a fun thought experiment, kind of a scary thought experiment.

I will say that because yeah, as I lightly put out there, maybe we're going to get replaced. There's some stuff here that it's not going to replace one-to-one. It's not going to replace an individual person, but there's some great information and there's some great research and there's some great insights that these systems are generating. So this is fun and I'm excited to share just what we found in this because there's some pretty cool stuff here. So let's go through my premise first, just because I'm first in the document. No reason for that.

We'll go through your second and we'll see what did. So I actually ended up using three tools. And one, I kind of cheated on, I use the Google, Google tool as well. So I use chat, TBT, a four dot or like four. Oh, I also used, um, chat TBT four, Oh, mini, which is quote unquote, the free version. So obviously some of these tools, um, maybe you can walk me through this too in a second. Because there's Gemini, which is like 1.5 Flash, which you use the free version of Gemini.

And then I think you use the more premium upgraded version of Gemini. One of the discussion points today is what's the difference between these two things? Is the free one good enough for this type of task, or do you need the more premium version of the tool to get better output? We'll come back to that, dear listener. But I also use a tool that I've been testing a little bit called notebook LM. And it's interesting. Everyone shared notebook LM a few months ago, I feel like, and it was all about the audio output where you could like upload a document and then you get this like podcast style

thing and people are sharing it. That was fun. That was kind of interesting. I don't find the podcast to be particularly engaging, although it's unbelievably convincing on how they sound. I think training on voice is very easy, obviously for these LM LLM tools. It sounds like a person. It's honestly kind of crazy. When I listened to them, I was like, okay, to me, this is slower than just going and reading the equivalent in text. So unless you're a big audio learner, that wasn't the function of notebook LLM that appealed to me. What appealed to me a lot about notebook LLM and using it for this task was that if you give it the right context, if you give it the right documents, it references and will give you citations within your document

to what you're asking it. For example, if you upload a book to notebook L.M. and say tell me all the times the author talks about XYZ topics, tell me all the times the author talks about paid advertising. I uploaded my book to notebook L.M. and did the same thing. It will say page 17, click here, goes to page 17, pulls it out. It is, in my opinion, notebook LLM in this test was without a doubt the most truth-based LLM system that I've used because it's not inferring, it's not guessing. It's taking exactly what you give it and giving it back to you in the format you've requested. So if you said give me bullet point list of all the ways the author talks about paid advertising, it goes through the document, spits it back out to you, and then does it in that way. It doesn't infer or make estimates or guess like ChatGBT,

and I feel like Gemini does the same thing. So that can be a pro and a con, depending on your perspective. Depending on what you're trying to do, that could be a really good thing. Like if you uploaded your paid ads results from last year, gave it a spreadsheet, and then said, tell me what you expect all my results to be next year based on these growth metrics, it's going to pull exactly from there and not do much guessing or inference, which I think can be a positive thing. The downside is that what I learned in doing this sort of experiment, this project, is that it's not going to write anything new as new as you would like. So if you don't put it in there, if you don't give it enough context, if you don't upload your last year's results, if you don't upload some

instructions, it just kind of goes like, yeah, I can't help you, basically. So that was kind of my feeling on Notebook LL. It's kind of like a little bonus one. So here is kind of the premise that we went through. I'll go through mine first, like I said. So Chat GPT, Chat GPT Premium, so it's like 4, 4.0, and then 4.0 Mini is kind of like the different platforms there. My goal was to take a property manager in a fictional market. We picked like a fictional, I think we chose San Diego

as a fictional market, although you have a client there. 100 unit company. And I want to spend $30,000 on paid advertising, and I want to drive at least a six to one ROAS. So these were like the inputs that I gave to chat GBT or notebook alarm. So here's my unit account, here's my company user or Matt, here's how much you want to spend. And here's how much I want to get back out of it. And then I waited back for the results. So here was kind of my grade of the three different outputs based on what I saw. Notebook.alarm, like I mentioned, took the information that I gave

it. So when I put the prompt in there, I didn't just put the prompt in there. I gave in additional instructions from our strategy documents and other things that we've done before. And I asked it to recommend what I should do, like what should the campaigns set up that I should have. Here's what it suggested. You should set up dynamic search ads for property detail page campaigns. You could set up a branded search campaign. And you can set up long tail keywords targeting

key vacational terms. And then you can set up primary or area named vacational terms. So those are the four campaigns that I suggested. And then it gave examples of each of the different things that it wanted to do there. That's true. I don't disagree with what Notebook alum gave back to me.

The downside is that it gave me no additional information. So it just said, here's four things you could do. It looked at my information, and then it spit back out, here's four things you can do. And I'm like, OK. They gave a little bit more information there of like, for example, I won't read all these, because we'll be here all day if I read the whole output.

But with brand search campaigns, it's crucial to own your brand name and search results, preventing competitors from bidding on it and ensuring that guests searching for your brand find you directly. This strategy helps capture high intent traffic and maintain brand visibility. That's completely true from my perspective. I'm with AI on that one. Excellent. But it didn't tell me how much to spend. It didn't tell me what bid strategy to use. It didn't tell me how broad should my brand keywords be. Should it be just the brand? Should I put some geo modifiers in there? It didn't give me a lot should my brand keywords be. It didn't give me a lot of the extra context that would be necessary to make a strategy. If you want to use notebook LM to take what you're doing and ask questions about it, get feedback from it, ask it to research

long documents, I think notebook LM is fantastic. It doesn't give you back anything that's like, hey, based on other knowledge I have, let, let me take your information and then combine it and put it back in. So I'll pause before I go to chat tbt in the next one. But when you read through my notebook, LM output, what was your thoughts on it? It is. And I think this is something that I've seen that with my use of notebook LM is it needs source material, you give it source material, boom, it can create some crazy stuff. but you hit the nail on the head with it doesn't think that well, it can think it can regurgitate what you've got and help you find I mean, that's I did the poll of

I pulled video clips from like September forward on LinkedIn, and I had it do the same thing. So it's pulling from 80 different video files and kind of picking up the same types of things where I like how it always sites back. I think that's great. That's very helpful. And again, understanding what you've said, or in the case of this, where you're pulling that information from, I like that it always draws back to some type of source information. Because when you're doing that research, I think that is going to be important at a certain point. But as strategically, there's no strategy that's being outputted from this model. So yeah, I mean, the overall insights, again, if you're teaching

it with a high budget with some of these very specific modifiers in place, it's going to give you some type of form outline, but you have to start with that form. I think you have to have created that form initially in order for it to be effective. So yeah, I think that of the three that you have, of the three outputs that you have, this is probably the weakest because it is, because it's stuff that you already know. That's what I mean. That's that's that's my my assessment is this is this is the table stakes. This isn't going to move your overall strategy to a level of I'm going to improve I'm going

to hit these results based on the information I'm receiving from notebook killer. That was my thesis or that was kind of my conclusion as well, which is that this could be a helpful little tool. Like you said, it's almost like if you know what you're doing, this can be a helpful little, you know, research assistant for you. It's kind of like one of my AI theories right now, which is like, this is the best junior employee you've ever seen. Never takes a sick day, doesn't have any time off,

can do any research you want, like it's awesome in that way. But would you trust a junior employee to run $100,000 of Google Ads for you, who's never run Google Ads before? Probably not. You'd want to have someone with a little bit more skill and intellect, one layer up, that would actually be able to run that campaign.

Now, would you want to have that junior level employee, entry level employee do a bunch of research for you? Hey, go take all the clicks from last year, put them on a plot, and tell me. It's amazing in that respect. I also think these AI tools have gotten better and better at what I call first draft thinking. So it's amazing to like put a bunch of inputs in there

and then say, take all this stuff and give me back, you know, a strategy. I then take that copy and paste that, put it in Google doc and then I start. So it's like, oh my God, that's such a more pleasurable feeling than starting from a blank cursor, which is how so many documents that I've written

in my career started a blog post, et cetera, written from a blank cursor syndrome to then have like some kind of scaffolding. And when I look at my final, final thing that I'm written in my career started a blog post, et cetera, written from a blank cursor syndrome to then have some kind of scaffolding. And when I look at my final thing that I'm publishing or finishing or even outlining through this podcast and stuff like that, it was shaped by the AI. It was helped by the AI, but it certainly wasn't like, take the AI and copy and paste it and put it in there.

Otherwise you have no unique insights. So let me go into my next thing here and then I can beat this up a little bit more. So my next version was the ChatGBT Free. I'll go a little out of order here. So the me go into my next thing here, and then I can kind of beat this up a little bit more. So my next version was the ChatTBD Free. I'll go a little out of order here. So ChatTBD Free explained nothing. So I gave it the input and it said, run these two things. Hey, your budget's okay.

And then it gave just the most like kind of useless, non-specific suggestions. So it's like with an annual advertising budget and these raw expectations, your vacational business is well positioned to implement strategic advertising plans. It just all does explain each campaign type, you know, brand campaigns help you cover your brand, long tail keywords help you find people looking for specific vacation rentals. So if you just need an explanation for what these things are,

ChatGBT free is helpful. Like if you didn't understand like the terminology in a report that you got, or if you didn't understand specific language that someone was telling you. Like I'm thinking, if this makes any sense, I'm thinking like, I don't know if you ever got like a blood test report back and you're like, I don't know what any of these numbers mean. I don't know if a red count of this is good or bad or whatever.

Like I think that's what these tools are awesome at is like upload it, give it to it and go, I don't understand any of this, explain it to me like I'm five. And that's where chat GPT, even the free version excels. Like it explains everything very clearly, but it doesn't give you any insight on if what you're doing here is good or bad or a different It doesn't actually give you any specific Strategy if you will remember back to a strategy conversation for a few weeks ago strategy is putting what resources you have against an outcome and saying

What's the best use how do I get closest to my outcome of 6 to 1 ROAS? based on the inputs that I'm getting in of a budget of whatever, 50,000, 100,000, whatever the case may be. And the answer is max out brand and then go to PDPs. And then like, I know what it is because I've done it a lot, but this thing doesn't know. It doesn't have any context into what's actually gonna work well.

So it can explain things at a high level. Let's go to chat, GPT premium, explain things with more detail. So I gave it that same input and then I got back much more specific instructions on what exactly to do. Here's how to set up the campaigns. Here's where to go, on what exactly to do. Here's how to set up the campaigns. Here's where to go.

Here's what to do. Here's some of the information based on the input that you provided me of how to go and set up these campaigns. And it gives some, you know, it did some more math. So for example, it was like based on the number of properties you have, you're spending $1,600 per property per year for advertising based on, you know, my understanding of advertising, meaning Chai GPT's understanding of advertising. That makes sense.

You know, that's the logical budget. What I should have done, maybe I can doT's understanding of advertising, that makes sense, that's not the logical budget. What I should have done, maybe I can do that here while you're going through your piece, is like, let's just say I wanna spend $10,000 per year and just see what it says, is that okay? And like, see if it just agrees or if it like pushes back on me, like, oh, that's pretty heavy.

I don't know what data it's trained on, so that's kind of an interesting kind of point there. But although the premium version or the paid version of Chai GPT explains things a lot better, goes into more detail, gives me more instruction, it still didn't make any calls, it didn't make any decisions on how to best deploy that budget. It basically said, like, yeah, try all this stuff and keep an eye on it, and if it's working well, do more of it. If it's not working well, do less of it. I don't know how helpful that is.

I don't know how much strategic decision-making is happening when, you know, I'm giving it an input, and then it's mostly just giving it back to me a little more context-level information and going from there. So my grade, if you will, to kind of round up my prompts here, is that I liked notebook LM the least in terms of the output it gave me. Chat GPT premium or the free version, excuse me, I'll put a click above, but barely. And then I'll say the premium version was in theory the best. But my concern that I had going in was, you know, oh, it's going to put me out of a job. I don't feel that way today. Like if this is the most advanced model that they've ever done, well, again, I do feel like this is great for like researching or breaking out ideas or like hashing out ideas, stuff like that. It didn't actually make any decisions. It didn't

actually tell me how to break up the budget on a monthly basis, on a daily basis, et cetera. It kind of said like, yeah, you can do these things. Here's how to do them. Here's the context. And then go from there. So my experience here, I think, was if I got this from a client or if I got this from an employee and said, hey, here's what we're going to do with the, this is how we're going to spend this money for Google Ads in 2025 for this vacational manager, I would give it a C plus, maybe, maybe a C, somewhere in that range. It's not useless. It gives me some direction, but it didn't give me anything that I don't think wouldn't have, you know, it didn't give me a new insight, didn't give me any new information.

So when I was finishing my test, I felt secure in my job, I felt secure in my ability to provide value for clients, because I can confidently make strategic decisions based on the inputs that I get. And today, chat GPT cannot. So Paul was turning the page. What was your learnings as you built your marketing strategy with

Gemini? Mine was definitely different. And I think part of it is that what I'm doing on the owner side, certainly that's going to be a little different. And my prompt is I want to do direct mail, because that is still a pretty consistent way of doing homeowner acquisition. That is a campaign. I think it's because I've referenced it many times, the digital side is just, it's tough, it's tricky, you're limited by a whole lot of factors there. So I did. I made it much

more general, much more broad, because that was kind of one of the things I wanted to test as well. If I give it a very light prompt, not going down into not giving a whole lot of budget specifics right off the bat, what's it going to give me? So my prompt was I'm a vacation rental management company in San Diego, California. I currently have 100 vacation rentals in my portfolio, would like to grow to 150 rentals in my portfolio over the next two years. I would like to do this by running a series of direct mail campaigns. Where should I start? So we're at the at a very high level. This is a pretty similar process to what we do on the Venturi site or what owner point does or what a lot of the existing owner acquisition solutions

are doing. You know, it's it's kind of taking and define your ideal audience, build a targeted mailing list, craft compelling direct mail pieces, all pretty straightforward things that you know, Venturi can do that, all these people can do for you. Plan a series of mailing. Don't rely on a single mailing. Don't just do it once. Consistency is key. Things we've talked about over and over again. Track the results. Use a unique code, phone number, integrate with other marketing efforts. I thought that was interesting that even though I was very explicit on the direct mail side, make sure you're putting website online advertising and other networking opportunities. So the example they give is attend local real estate events, connect with property owners,

everything like that. The one thing I thought was interesting about the free version is that they actually gave some ideas for the mailers, Maximizer rental income mailer, stress-free property management in mailer, local expertise global reach mailer. So very, very high level, not really talking about the how you're going to do it USPS and stuff like that. But, but if someone handed that as a strip strategies sheet to me, or as a property manager, I could probably read through that and put together something, you know, feel like I had a pretty good outline or plan for how I was going to do that. And then, but again, didn't really address the budget there.

When I moved up to pro, I did, I kind of probed it a little more. We got a pretty similar information, refine, optimize, track measure results, do all that stuff. But I think the key difference there was then going down to factors that are gonna affect your budget, so how much budget will I need on a monthly basis to reach my growth goals? Factors affecting your budget, mail list quality,

mailing piece design and printing, postage, campaign frequency, geographic area. So then I did, I got an actual breakdown of a sample budget, a monthly estimate, you know, for 50 cents a record for to get the data, 250 bucks, 75 cents per postcard for design and printing, 375. Postage is another 55 cents per postcard, another 275. Total monthly cost, they're saying about 900 bucks a month. Now, you know, that's a nice budget with which to work and is that gonna help someone reach a goal of adding 50 over two years?

I would tend to think that's a little short, but they're giving us something. They're giving us something to work with and they're breaking down the costs based on what they're finding that's out there. Their important notes I think is also important. They're telling you what to consider. This is a rough estimate.

Your costs may vary. Consider your conversion rate. If you have a higher conversion rate, 2% of your recipients sign up, you may need to mail the fewer people. So some of the stuff here that that really, I think it is it's it's painting an opportunity for how you could do this. I think you could probably go to a next level here on the pro side and feel pretty comfortable. You know, you could you could start to run through this process and really feel like,

if you found a printer, and if you found a designer, and if you found the data, I think you could execute that strategy. You could have someone else do it as well. But I think you would feel comfortable with the detail that's given here in executing the strategy. This next one is where I really, the deep research that I think Gem and I are has rolled out in the last week and a half, maybe two weeks. It's still pretty new. I've only done it a couple of times, but this is probably the most fun I've had with it. Again, same prompt, vacation rental company, 100 properties, would like to grow to 150 rentals.

But how this, the deep research breaks down is they're giving you how they're going to do the research. So find areas in San Diego that have the highest concentration of vacation rentals. Demographics of the vacation rental owners in San Diego, age, income, property, ownership. Find the typical marketing channels used by vacation rental management companies in San Diego. Don't just focus on direct mail campaigns. Look at all marketing channels. Find the cost of direct mail campaigns in San Diego, including printing, postage, and list acquisition. Find the typical response rates for direct mail campaigns in the vacation rental industry. Find examples of successful direct mail campaigns used by vacation rental management companies, and find a list broker who can provide

targeted mail lists for vacation rental owners in San Diego. That one prompt gave me all of that. Google just said, okay, this is the research I'm going to do. And I mean, you can see the results here are in-depth. I mean, neighborhood, medium price of a three-bedroom annual gross rental revenue, the rental yield, key features, it's nuts. This is where, I mean, if I were going to go down to San Diego, it's giving me Middletown, Mission Hills, North Park, Pacific Beach, Point Loma, places that I know,

I've been on Keywords previously, that I know are good markets. And now I'm getting the data to back it up. And I did, I kind of took it down even further and said, so how would I identify even more markets? Because maybe five isn't enough markets for me within the San Diego area. That's a huge area. And it kept going.

I think it's cited 35 sources in the research. One of them was buildup bookings, which I was very happy to see him knock on. That was a big win for you. That's right. So that was when I saw that logo pop up. I'm like, yes, vacation rental email marketing campaign ideas is where you were cited there. So it's an older blog post, sir, but it checks out. It should tell you. I mean, that's there's there's value in that older content there. That's that's that's classic. That's that's that's evergreen content. But this is the one that it does. And I did I did Gemini 2.0 Flash experimental there. There. Google's got I think,

six different models right now. So that anything with the deep research really feels like it's gonna find an answer for you. It still might not be the right answer for what you're looking for. So if I'm gonna take the 10,000 foot overview, I'm gonna say that the free versions of Gemini still aren't worth sending alarm over. But the lighter versions of Gemini are okay. The 1.5, the flash, that's all right. 1.5 Pro is solid. I think the key with Gemini is you really do have to, you have to converse back and forth.

You have to kind of, that initial prompt isn't going to do it for you in most cases. Now, so I think, you know, we can we go about C minus for the free side of things. I think we give it a solid B for for the pro. And I think with this deep research, depending on the uses you're using it for, I give it an A minus, maybe even a solid A. I think there's some definite value strategically on what you can, I think it is, you still have to have some type of strategy. What's your thinking? It's the the LLM is not still not

designed to do all the thinking for you. I think that's a good thing. We should have to, we should have to add that nuance. But that deep research is something that I will continue to do more and more, especially on the owner side of things, because it's not just that they're finding all this information. They included how to find appending information, and we're some top vendors for appending information,

for compiling your mailing lists. That stuff that, I mean, it went back and forth for, for years, just trying to figure out who has the best data and how can we get that for people? So I, uh, I was, I was much more impressed by Gemini as, as I. Ben from time to time, then I think we probably both were on the chat GPT side of things. Yeah, you know, based on based on your output, I'm looking at both, not just because buildup was mentioned, but because I'm looking at the I feel like if this was a battle, I feel like I got to give it to Gemini, you know, maybe because this is such a specific topic, you know, Gemini has access to Google crawling information,

and they're just consuming more data, they're consuming more information than JackTBD is right now or OpenAI is right now. I think that's a fair assessment. And it's so funny, this was a topic on the All In podcast a while ago when JackTBD was blowing up, is the first one that comes out often always the best one? And it's like, not always. In theory, Gemini or Google with Bard and all that was a little bit later to the party here. But in the grand scheme of things, we're going to be using one of these AI LLMs for the next like 10 or 20 years. Like, does six months or eight months or 12 months, you know, mean much

in the grand scheme of things? Not really. So like, maybe Gemini is quote unquote catching up, if you will. You know, compared to, you know, where it was before, when it first launched. But based on these outputs I'm seeing here, I think it's giving the citations and research is super helpful, because then it feels a little bit more like, you know, chat tbt often gives you these like sort of final answers where it's like, here's the input, try to give you back the final answer, where sometimes that's perfect. Sometimes it's exactly what you want or what you need. But with some of these tools, I don't think that I think Gemini is a little bit stronger in that respect based on what you're showing me, because especially with that research mode that you were

flipping on, and putting there it's like giving you the exact mail mail pen services to use. It's giving you a link to Venturi. It's giving you a link to some of these other services. I think that's a good thing. I think it gives you some opportunities to like research, here's some different ideas, here's some things that you want to measure. So I feel like, yeah, based on these two prompts and this testing, the free version of Gemini didn't overly impress me, but I think the free version of Gemini is better than the free version of Chat GPT for this particular prompt. I think the research and the links that you put in there definitely put it a click ahead in my opinion, the paid version of Gemini versus the paid version of chat GPT. And I think if you're doing research and you're trying to gather information, I

think you're in a better spot, you know, do using Gemini at this moment in time based on the output here versus my prompt over to chat GPT. So I got to, you know, if there's a, there's a ding, I don't know if you need to like put a ding in post or something like a ding, ding, ding, ding. I think the winner here is Gemini from my perspective. So what, do you think there's something about your prompt or like previous history, or do you think it's just like

this research mode just gives you a lot more context as you're giving these prompts in? What's kind of your read on this, given that you've been using this tool a little bit longer? Yeah, I think the, so admittedly, the first research one that I did was go back and evaluate the 2002, 2003 NBA draft or something like that,

and find me the top people who were drafted and give me the backups behind that. And that was, I think that was a research article that had 79 sources. They went back through and did all that. So I don't, I mean, it's not, I don't think it's because of my search history. I don't think it's because of the learning it's done

because of me specifically. I was purposefully a little broad, a little vague in my prompt, just because I didn't want, I wanna see what the LLM can do without the perfect problem. And that was kind of how we how we positioned it, that we wanted to kind of see what it would take. I mean, is it really a matter of, if you give it a really good prompt, can it replace us? And I think you gave it a really, really good prompt. And it didn't come

close on the chat GPT side of things. I mean, and maybe if there's a little more nuance there, if I put in a specific budget, monthly budget goals and stuff like that, I might get a little different output. But what I got for quality wise, it is, it's, I think the ability to see those sources important, but to monitor, no, like we're pulling from BNB, Calc and AirDNA and sources of information that I find to be of a high high value. If we're gonna do domain authority, these are the high domain authorities that I should that we should be looking for as far as market research in some of these specific case rental markets. So I think that's the also the nice part about that deep research system is that it's showing

you like before you even see before you see a single output, you're taking a look at the sources that are being sourced. That's kind of the first thing that they're pulling through and they're reading I'm sure and doing the research, the deep research there. So it is if we do 10 more of these or do 10 more exercise and figure it out. I would, I'd be interested to see like, if guest side is easier than owner side or different marketing channels are easier than this or, you know, whatever that would happen to be.

I think the one thing I did notice from a lot of the sources that they were pulling from in the deep research side is their blog posts. So, you know, blog posts from some agencies in the space too and then we have to say, okay, is there, are they the experts? Or is there more, someone who have more expertise? There may not be anybody of more expertise.

I think that's the reality is that blogs are just, they don't, maybe they don't feel like the traditional user generated content of a social media. It's still user generated content. Someone has to be that expert. Someone, you know, however we're defining that expertise, any blog posts, you know, that's been referenced from here to the beginning of the internet, came from someone individually. So, you know, I think we're dealing with the same questions of citations as we have been forever. So if you're gonna

question the citations, that shouldn't have stopped from like when you're doing offline research. So I think that's the ability to go back and check and to see. I think that's something that we give AI overviews a lot of guff for lack of a better word there. But what they do is every time you click to a citation even from an AI overview It's gonna bring you to the exact spot on that website where that content has been placed where they pulled that citation from They're not making you dig for it They're not making you know so you can see you can see the relative content that it's around and then what else is there? And you get to have an idea of how qual high quality that source is there so

Okay, Gemini is doing it on that side, now they're doing it even better on the research side of things. So, I think that, like I said, I'm going to be using it a lot more frequently, I'm gonna be recommending a lot of other people, use that, I mean, it's that alone. If you're doing some type of planning for a year, I would invest in premium for the

first quarter of Q1 of 2025, just to take advantage of some of the deep research opportunities that were there. Even if you only tested it for three months, and you used it for nine months of content after that, I think it's well worth the investment. So. Yeah. Yeah, I think my parting thought on Gemini versus TrachiBT is why battle them against each other? Why do we feel like we need this?

I've seen this happen a few times, you know, to only put, you know, the output into one place. That doesn't make a lot of sense to me because in the grand scheme of things, they're so cheap. Right? Like imagine the cost of TrachiBT because in the grand scheme of things, they're so cheap, right? Like, imagine the cost of GTPT is 18, 19 bucks a month. The cost of Gemini, it looks like it's about 16, 17 bucks a month. Collectively, for less than $50 a month, you could just run the prompts in both and see which one gives you better output because there's certainly been things I've tested before

Gemini that have not been very good that I tried to have a GTPT and is better. So, you know, test both these things. But yeah, like I said, I think we got to give Gemini the winner right now. And maybe it'll be fun to kind of revisit this on an ongoing basis and see how we can get there. I want to turn the page a little bit, because this is just more fun for me at the end of the end of this particular episode to go through this idea as we kind of theorize from today's episode.

I'm trying to figure out if there's a way for us to build like an AI strategy for clients or at least an AI assisted strategy from clients. And I think there is, I think the question is, the quality of the inputs matter so much in order to get the right output, whether it is from Gemini, whether it is from Jack GPT. And I guess what I'm questioning at this moment in time is that what is the best layer to get the information from the property manager into one of these AI tools, which we can then get the output from. So my kind of current thesis is that it's like a really long intake form. Like you fill out a bunch of questions, like, you know, what are your goals? What's your revenue? What's, you know, what's your

commission model with you and all these things. And then you click a button, it goes through all these long list of prompts, then it sits back a bunch of ideas out of you. I think that that can work if the Property Manager has a pretty detailed amount of patience, or pretty large amount of patience, I should say, and a pretty detailed list of what they actually are looking to do. I still think the gap that exists though, is that people are not excellent at honestly grading where they are. And they're often not excellent at putting in this level of detailed information into the prompt to get the right stuff back from the AI. So when I, this was a long time ago, but at one point in high school, I thought I was good enough to be a programmer So when I this was a long time ago, but at one point in high school, I thought I was

good enough to be a programmer. And I took like this computer programming class. And this is like a very common phrase in program programming, which is garbage in garbage out. So if garbage goes in, then garbage goes out. And that was that was what we were told, you know, when we were trying to figure out how to write code and C sharp and stuff like that for programming. I think AI is pretty much the same thing, which is that if you give it the wrong information, or you don't give

enough information, you're going to get pretty much the same thing, which is that if you give it the wrong information, or you don't give enough information, you're going to get pretty much garbage back out of it, it's going to give you very bland, you know, blame, plain recommendations, like my notebook alum complained from, you know, a few minutes ago at the top of the show. So it really is, I'm wondering if then the the onus is on the people who are still involved in the process for now, to figure out how do I get all the right information out of the person that's giving me this, you know, the owner of the

company, the leader of the company, I need to get all the right information out of them so that I can feed it to the AI in exactly the right way. And then it's going to give me back, you know, all the research that I could do, but it would take me a lot of time to do. And it sort of cements in my mind that at least today where we are is that if you don't prompting is a skill that you can develop and learn and continue to improve over time. And if you're not working on this skill, you are going to get left behind by someone who could produce probably, even if they're not producing necessarily better reports, my concern is that because it's AI, like there is people who are thinking it's better because it's AI when, you know, we just gave examples today of like, track GPT not really giving very good

recommendations, people might think that they're decent recommendations because they came from some robot. So the final version of this thought process is I have to take the 50 links Gemini spit out of you and distill it down into the four things we're going to do. It's still going to take that final layer of human judgment to say, all right, there's a bunch of good ideas from Gemini. Thank you so much. Gemini, we appreciate your research. Now how do I decide what five things to do first? And I don't think that the AI is quite smart enough yet or has quite gotten there yet, where that's gonna be the thing that actually gets us

to a live strategy that we have to go to play in the real world. But I think that it's a lot closer now than it was a year or two ago. And I think that's where we have to keep our eye and tuned into this space to see how much better is it gonna get. And this whole agent concept that's coming into play,

how far away are we from going, okay, Gemini sounds good, go ahead and design the mailer for me and send it back to me. Go ahead and send the first 10 postcards and tell me how they're doing. Go ahead and build the landing page and connect it to my CRM. Because I think we're not that far away from that. At least, again, almost like the first draft version of that,

where it's like, hey, it gave you the first version of the postcard. It gives you the first version of landing page. Then you have to go into Modifier, change the text, and stuff like that. And I think what honestly is just going to happen is that it's going to be expected or it's going to be the ideal outcome for one person or maybe two people to do the work today of what is currently like six or seven people. So if there's currently like eight people in your

marketing department all doing different things, I think the way we're headed is like, you're going to have one or two marketing people and it's going to be assumed that they're going to be using a lot of these AI tools. So like we still need them, but they're going to do a lot more. you're going to have one or two marketing people, and it's going to be assumed that they're going to be using a lot of these AI tools to like, we still need them, you know, but they're going to do a lot more, they're going to get a lot more output out there into the world. As they go through this, you know, these AI tools and all the all the things they have, it's going to be expected that you're going to have this AI tool sitting here using, you know, giving you a bunch of stuff back, you're prompting, you're giving stuff back. And that's going to make you in theory, a lot more productive, you're going to be able to put a lot more information or content or strategy out into the world.

So that's kind of my final thought here. I think that people who are anticipating that their work load is going to go down like 90%. That's not what AI is meant to do. That's not what like you, the people who think they're going to write a paragraph and that's going to turn into five pages of content. That's great and powerful and everything. That's not how it is. Now you can, you can write incredibly good prompts, but to write a really good prompt,

it takes work. It's in, it takes a lot of words. And I think that that is something that it might take just as much work to do AI driven tasks initially. And I think after the system is learned and after you've learned and after you've figured out the process there, that's when the efficiencies come in. But I think it is if you're not even thinking you hit the nail on the head is If you're not working on those prompts if you're not working on the initial efficiencies Then you are going to be behind because the those the efficiencies that people have been using ai for the last year They've already built prompts on top of prompts on top of prompts. They can probably create

Amazing images made probably some videos and and they can probably create some pretty cool content just by putting in some extensive prompts. I mean, the prompt, the prompt, and I use those air quotes very heavily here that you have just for your overall kind of breakdown. This decision tree as it is, is intense. I mean, it's like a four page document in and of itself. So I think that's the key here is I think we're limiting ourselves if we think of prompts as something quick, like a conversational chat, like something you're going to see in a text message back and forth between Google. That's not what this is. The more good data you put in, the more powerful data you get out. Now, hopefully it's good. I mean we get some stinkers in there too, but You are going to play

You are going to get out what you put in so learn how to put those prompts in in a way that is going to generate better results because That is that's the only thing especially the free versions That's the only thing that's that's gonna limit you ultimately is your ability to create a prompt that is detailed enough for the LLM to return an answer that will be effective for you. Yeah, good agree more. So, all right, well, that concludes, I believe our first AI battle, the chat tbt versus Gemini versus

human battle. And we feel like Gemini has kind of taken a pretty interesting lead, although you need a human to sit there and refine it. So thank you Paul for allowing us to use your Gemini license here. I think that I will take mountain medicine and I'll upgrade and start to test these a little bit more. Maybe we'll do an update episode here in a little bit. Maybe it'll convince me, maybe we'll get on the same team

for once instead of being on opposite teams like we are so often in some of these tech decisions, but it's all good because it allows us to put good fodder out and good information here on the podcast. So if made all the way to the end, again, happy new year. Hope you're off to a awesome start to 2025. And I think it's going to be an awesome year, but it's going to be work. It's going to be work. You're going to focus on what you need to accomplish. You're going to have to make sure you understand what's around the corner. And hopefully we are at the, we here at

the heads and beds show can help you do that with some of our episodes. So thank you for listening. We appreciate it. I got a few nice messages about the Merry Christmas podcast, which is fun, because it puts some Amazon gift cards in there and stuff like that. If you go to look, they're all taken. So sorry, you missed the button on that one, but we appreciate the feedback on that one. One thing we do need, though, to kick off 2025 for us is a review. So go to your podcast half of choice. iTunes, Spotify is where we get the most downloads, therefore we appreciate the review the most.

Click follow or click subscribe that we get future episodes that we're going to publish every week going forward. And also, please do click five star, leave appreciate the review the most. Click follow or click subscribe. That way you get future episodes that we're gonna publish every week going forward. And also please do click five star, leave us a review so more people can listen to the show. We appreciate that as well. Thank you and have an awesome rest of your day.