Slides | Download Audio | Transcript
Transcript
My name's Rich Brooks. Welcome to the AI Marketing in Action, a Practical Guide for Your Business. Today, I'm going to be sharing all the ways that I'm using ai, and my team is using AI here at flyte new media.
I just want to start by saying I drive a 2013 Toyota 4Runner. I only bring this up to share with you that I am not a car guy. I can't tell you how many miles to the gallon my car gets. I don't even know how the engine works or how to fix it. I can't even tell you if it's a V6 or a V8. But I don't care.
I only care that it's reliable, that it can fit all my gear in it when I head to the mountains, and that I can connect my phone to the car speakers so I can listen to the latest Jack Reacher novel. At some ways, that's how I approach AI, too. I don't care which model is faster, or if some Chinese company just released a model that scores 12% higher on some scale than the tool I'm already using.
I assume that at some point my tool will catch up, and if it doesn't, I'll switch. I do care about whether AI can help me do my job. To be better and faster than the day before. To create better outcomes for flyte and for our clients. And that's the approach I'm going to share with you today. I want to share with you some of the ways of how I'm using artificial intelligence and how flyte is using it to run better, smarter campaigns for ourselves and for our clients.
So, who am I? If you don't know who I am, my name is Rich Brooks, and I am the president of flyte new media. We're a digital agency located here in Portland, Maine, offering branding, web design, and digital marketing like search engine optimization, social media, digital ads, and more. I'm also the founder of the Agents of Change, a weekly podcast, and an annual conference that's all about digital marketing.
I'm also the tech guru on 207, the evening news program on the NBC affiliates here in Maine, where I do hard hitting news stories like “What are the best recipe sites online?” And “how can you take better pictures with your smartphone?” And then I'm also the author of The Lead Machine, the small business guide to digital marketing.
In other words, this is my playground and I spend all my time in digital marketing. And because of the changes in the world today, I spend a lot of my time thinking about artificial intelligence. So, we started with a car metaphor. Let's shift to a housing metaphor. Congratulations! You just purchased some land on which to build a house.
You're talking to two builders. One has been doing this for a long time and only uses hand tools to create one-of-a-kind homes for families just like yours. The other uses power tools. She prefers them for their speed, consistency, and precision. She'll be able to build your home in a fraction of the time it would take the other builder.
And it's going to be more energy efficient due to precise cuts. On top of that, she has cutting edge technology that lets her personalize her home designs for your needs. Would you like a bigger mudroom because your family loves the outdoor? Push of a button. Need a bigger bedroom because your twins prefer to sleep in the same room, but they'll be teenagers soon and will need more space?
Want a four-season porch with a suspended fireplace? Updated. Did I mention the quote for the second house is significantly less? Which builder would you hire? Now, there's no wrong answer here. And maybe you prefer the more imperfect, organic vibes from the first builder. But chances are, most people are going to choose the second builder.
Faster, less expensive, and arguably better. And that's the point. AI is a competitive advantage, and if you don't start adopting artificial intelligence into your business, into the way that you work, you will be at a competitive disadvantage to other companies that are willing to take that leap. Now, I focus mostly on marketing, so today I want to talk about how AI is transforming marketing.
In just a short list, speed and efficiency. AI can automate time consuming tasks, the drudgery of your job. Ideation. AI can help you get started, even if you're starting from scratch and you have no idea what you're doing. It's a great way to brainstorm and come up with new ideas. It can help you with research and forecast.
It can sift through insane amounts of data to find answers, to find trends and make predictions. This can also include competitive research. It can help with content and SEO. It can help us create content at scale, but more importantly, it can help us create better, more valuable content. And AI tools can help us optimize that content for search visibility.
Personalization. AI can customize messages to specific audiences or people. Just a couple weeks ago, I was having a conversation with somebody who had created video courses. And I'm embarrassed to say, I don't remember what the video course was actually about. It might have been about leadership, but it could have been about something else.
Basically, it was a series of videos that led people through this online course. All the videos were identical. But when you signed up for the course, you provided the course creator a lot of information, including your job description and your job responsibilities. After you watch each video, there is an automation that went through AI that basically created a summary of the video. Even though everybody saw the same video, each summary of the video included examples that were relevant to your job and to your job responsibilities. And that's just a great example of how AI can really help with personalization.
Now, it's obvious that AI is making all these big changes, and there's a lot of opportunity here, but we'd be foolish to think that it's all good news. There definitely are concerns that you should have when it comes to using AI. One of which is bias. The bottom line is all of these AI models we're using today have been trained on content that has been created by humans.
And the problem is, humans are biased creatures. We're just a little bit prejudiced. It's an unfortunate fact of life. So as these tools are being trained on the content that we've created over the years, it absorbs some of that bias into it. So, if you're going to be using AI to generate content for yourself, for your clients, whatever the case may be, you need to be aware of this bias so that you can review everything and fix it before it goes out.
Otherwise, that might lead to some big problems for you and your company. Also, hallucinations. AI is still making stuff up. AI has created the AI tools we have today to be helpful. And sometimes being helpful seems like a bigger priority than telling the truth. And there's a lot of examples of people who didn't pay attention to this.
And just recently I was seeing that One of the biggest law firms in the nation is in a lot of hot water because a number of their lawyers has been have been using chat GPT to write documents for the courts and to write briefs and some of those documents that they submitted to the court included some gobbledygook and others had these amazing case studies or cases that they were bringing forth that sounded really important like Rich Brooks versus the state of New York and sounded official, but it turned out they were completely fabricated.
And the lawyers never checked to see if these things were real, but they did find out when the judges checked to see if they were real. And many of them got into hot water because of this, again, another reason why humans have to be part of the loop. Data security is another one. When you upload documents to AI, a lot of the AI tools we're using today, you're training the AI.
I'll show you some tips on how to avoid that in the future, but if you're uploading sensitive or sharing sensitive data with the AI, that data then goes into the AI's knowledge base and might get output somewhere else. Maybe somebody is doing some competitive analysis and suddenly information that should not be publicly available is being created and shared with that person. So, anonymizing the data before it goes live and taking some other actions is critical.
Job loss. Now I can tell you with absolute certainty that every single time there's been a technological advancement in the history of humankind there has been job loss. But on the flip side of that, there’s been job creation, and those new jobs usually pay more and are a lot safer.
But that's not going to make you feel any better if your boss just told you, you're fired because I can get Chat GPT to do your job for $20 a month. There is also environmental concerns. For whatever reason, AI uses a lot more of Earth's resources than typical computing. The good news is there has been some advancement in how these things are being processed. I think Microsoft, just the other day, had an announcement about how they're using less resources than they had been. But this is something you need to be aware of. And lastly, AI lacks human experience. It can make up a story. It can pretend to have a personal experience. But it doesn't really. The best it can do is just regurgitate things that it's already been trained on.
So, it's really important again, to put a human being in the loop when it comes to AI. Now, as we go through some of the things today that we're talking about, I want to talk a little bit about some of the major categories of AI that we'll be looking at. Large language models or LLMs. These are tools like ChatGPT or Claude or Gemini.
Then you've got image, video, and audio generators, as well as image, video, and audio editors. So, we've got tools that can both create or at least generate images and video and sound effects and music even. And then we've got tools that use AI that can also edit it that have been already created. We've got AI chatbots, we've got AI note takers, we've got AI powered search.
We've got AI agents that can autonomously perform tasks and make decisions on their own. Those are just starting to come to market. Plus, you have AI co generators, productivity tools, AI assistants and more. We're not going to talk about all of these. In fact, we're only going to really go over a few of them today.
And mostly we're going to be focused on actual things that you can do in your job. So, here's what I have planned for today. There's a hundred different ways that I'm using AI but I chose five of them that I thought would be most effective for today's webinar, for today's session. And here's what I want to cover. First, I'm going to show you how to create effective buyer personas for your business. Then I'm going to teach you how to use AI to create brand voice so that AI will sound more like you or more like your company brand. I'm going to show you how to do a competitive analysis. I'm going to show you how to do SEO, search engine optimization, and web optimization, also known as conversion rate optimization. And lastly, I'm going to show you how to create custom AI assistants in a couple of different tools.
So, let's start by creating a buyer persona. Like everything in artificial intelligence, this is an iterative process, meaning that you're never going to get it right the first time, and it's a little bit of back and forth that really makes things work a lot better. By the way, I'm not going to be stopping for any questions right now, but I've left plenty of time at the end for any questions.
Ask them as they come up. I might answer them on my own, or we might just do them in Q& A, but I would love to have, I want to make sure that everybody walks away with an understanding and gets their questions answered. So please post them into the Q& A section or the chat. I'll check both places.
All right, so most of what I'm going to share with you today can be done in ChatGPT, even though I often use Claude. Many people believe that Claude is a superior product in many ways, however, ChatGPT can do two things that Claude doesn't do right now. One of which you can generate images through DALI, although I would argue that the images in DALI are not nearly as good as some of the other image generation tools like Midjourney, which is my favorite.
But it can also surf the web. So, you can have it go out and check things on the web. Claude currently does not offer that. So that's an important thing to keep in mind. So, let's start by creating a prompt. And a prompt is just the way that we talk to AI. We'll create a prompt to generate a basic persona for our business, which in this case is an online spice retailer.
So, I go into ChatGPT, and I create a prompt. Create a buyer persona for a customer who buys high quality spice blends online. Include details such as demographics, cooking habits, pain points, goals, preferred communication style, and where they get their cooking inspiration. And boom, ChatGPT generates this answer.
And now we have demographics, cooking habits, pain points, goals, preferred communication style, and where Avery, the name of this avatar, gets her cooking inspiration, plus a nice little summary, all in four seconds. Which just amazed me, quite honestly. Even though I've done this activity at least two dozen times.
But this time, as I was putting together these slides, I got curious, and so I asked ChatGPT, how did you generate this persona? What sources did you use to create the persona of Avery? And chat GPT thought for a few seconds and then basically told me that I created Avery by drawing on well-known buyer persona frameworks and combining them with common demographic and psychographic traits often observed in consumers who value premium food products, especially spice blends.
Here's the thought process broken down. Marketing persona best practices, general consumer insights. Common pain points and goals for home cooks. Preferred communication and content consumption patterns. And then combining all of the above. In other words, this was a guess. An educated guess based on publicly accessible information.
It might take a human being 5, 10, even 20 hours or more to gather all of that. and synthesize it into a report. And again, it took ChatGPT about four seconds. And if there's something in this buyer persona that you don't agree with, since it's your online store, you can tell AI that, and it will update the buyer persona.
Maybe you know that most of your customers are actually men, or that they tend to live near coastal cities. Or they tend to order sample sizes first. So, you can feed that information back into the AI, and it refines the buyer persona until you're happy with it. For many of us, this might be good enough.
But you might be curious, how can we make this even better and more accurate and more valuable? I asked ChatGPT, what information could I provide you, such as customer surveys or e-commerce data, that could create a more accurate persona? And ChatGPT reasoned for one second and basically told me that to create a more accurate persona, the key is having specific real world data points that reflect actual customer behaviors and preferences.
Here are some of the types of data that you could share that would significantly enrich and validate a buyer persona, demographics, purchase behavior, and e-commerce metrics, product level insights, customer surveys, analytics, social media engagements, customer support interactions, user generated content, and reviews.
Now you may not have all of these things, and if you're just starting out, chances are you don't have any of these things. But the more accurate data that you can feed the AI, the more that you can move away from a well-reasoned guessing game to data driven marketing. Once you have this buyer persona, you can put it into a document and attach it to future prompts to ensure that you're always creating content with this buyer persona in mind.
Based on the attached, so basically in this case, I've already taken that buyer persona, I've saved it, say, as a Word document or PDF and then I'm basically going to attach it to this prompt by clicking on that little paper clip icon. And then I tell ChatGPT, based on the attached buyer persona, create an email promoting our new line of African based spices with a strong call to action to visit our e-commerce store.
I've also attached the new product names, spice profiles, and introductory pricing. And by doing that, I'm going to, or ChatGPT is going to generate that email for me, which I would still want to review and make my own, but it definitely is helping. And now we're not just creating content, we're creating content that should resonate with our buyer persona or our ideal customer persona. So that's Buyer Personas. And hopefully that gave you a sense of what you can do.
Let's talk now about Brand Voice. The first time I used ChatGPT, I'll tell you, I was just blown away. It was like magic. But if you're like me and you start using it quite a bit, you realize that by default, ChatGPT and similar tools, the conversation is pretty middle of the road, banal, vanilla. Choose your adjective and that's probably not your brand voice. It's not how you want to be seen, how you want to talk. So, we want to train these AI tools to speak in our voice or to speak in our brand voice. And this is how I've accomplished this myself. I'm attaching several examples of brands writing style.
Please summarize that writing style. In this case, it was flyte new media. And I don't remember the exact tools I uploaded or the exact examples I uploaded, but it was like examples from our website, blog posts, email newsletters, social media excerpts, and so on. ChatGPT went through all of those documents and then basically started coming up with the writing style, basically breaking it down, almost a document that describes that.
And some of the highlights included the writing features lively promotional tone that heavily incorporates flyte and aviation metaphors. The language is energetic, confident, and uses frequent calls to action. While it remains professional, the copy leans conversational and approachable rather than overly formal.
And I would say that pretty much describes our brand voice. I will say, although we do use flyte puns, I think the examples I gave it were a little over the top on those flyte puns. So maybe that's not what I really want. So again, iterative process.
I might say, hey, chat GPT on a scale of one to 10, how would you rate, or how would you rank the incorporation of flyte and aviation metaphors where 10 is too much and 0 is not at all? And maybe it comes back, and it says 7 out of 10. I say great. As you create this brand voice, I want you to make it 4 out of 10. I don't want to have as many flyte metaphors in there as you've identified.
And you can do that on a lot of different levels. You could make it more helpful, or you could make it funnier, or which is tough because chat GPT is not very funny. But you can make it more sarcastic, you could make it more formal, whatever the voice is, because you can ask it to give you samples, and if you don't feel that it's on point, you can continue to reiterate this.
And for some people, they like to actually have, here are the things that describe our writing, and then give it a ranking from one to ten, where one would be very little, and ten would be a whole lot. So, if you're going for formal language. You might say, I want the formal to be like an eight or nine out of ten.
And that will basically, it's almost like an equalizer. It'll basically raise the formality of the speech. Or you could say things like, always include an Oxford comma, because the Lord wanted it that way. No, because you like to write with Oxford commas. And so, from then on, that's going to be part of the brand voice.
So, once you've gotten it to the place where you want it, what I would recommend doing is. Using this prompt or similar prompt. Please rewrite as a document that I can use with AI prompts to ensure any copy created is in our brand voice. So just like before with the buyer persona, I then get a document and it basically gives me this brand voice prompt. And here are the instructions to AI, tone, style, personality, key traits, so on and so forth. It even gave me example usage on how exactly I should use it with this prompt, which I thought was very helpful.
Next, I might create a custom GPT, which is what we're going to get to at the end of today's webinar. That includes both sets of instructions. One for buyer persona and one for brand voice. And from there, I could ask AI to write a blog post, email, social post, landing page, and so on that would be in our brand voice. As well as focusing on the needs and pain points of our ideal customers or clients.
Or if you're like me and you don't want AI to actually do the writing for you, you could use these instructions to either create a framework for those content pieces that I just listed, or use it to provide feedback on content you write yourself. So maybe you write a blog post and you run it through this custom GPT that you've created that has the buyer persona, that has the brand voice in it, and then it could provide you feedback on whether you've optimized that content for your buyer persona, and if it actually does reflect your brand voice.
Next, I want to talk about using AI for research. Obviously, it can do research for you. If you've asked it any questions in the past, AI has its own knowledge base, which gets updated fairly frequently more frequently these days than it used to be.
And that's how Claude works. It's also how all the tools work. They have their own knowledge base, but you can also ask say chat, GPT or Gemini, to look at a webpage, something like that. So, it can do some minimal amount of research, but let's say we want to take it a step further there we want it to do deep research.
There are tools out there that can do that deep research. So instead of in 1, 5, 10 seconds, giving you the answer, you're looking for. It might take. 10, 15, 20 minutes or longer and look at hundreds of websites and dive deep into those websites to bring you back the information to synthesize it and to write out a report for you.
Often those cost extra even if you're doing the paid version of chat GPT that doesn't include doing that extra level of research. 3 of the more popular tools right now that do deep research one is by and it was the first one out, one is by Google as part of their Gemini platform and they call it Deep Research, not to be outdone, ChatGPT came out with one which was called Deep Research.
And then Perplexity, the AI powered search engine, also came out with one and they named it, you're not going to believe this, Deep Research. I think we have a branding problem here, people, but let's not go into that.
Let's take a look at the first one, Deep Research by Google, which Gemini is the name of their AI tool. So, here's a screen grab. I go down and I'm choosing the tool, the version that has deep research in it. And I asked Gemini deep research, I'd like you to perform competitive analysis for redacted, a residential propane company in Southern Maine. The website is redacted. Please identify their toughest competition and then run a SWOT analysis with the goal of finding opportunities for new marketing ideas. Let me know if you need any extra info.
I often like to ask that question of AI because sometimes they'll ask you questions that you wouldn't have thought of that provide the right insight to get the best possible answer. And then it gives me a summary of what it wants to do, and it includes things like visiting the redacted website, identify other residential propane delivery companies in Southern Maine, and boom, all the way down the line.
And again, this is a great opportunity for me to weigh in and provide additional information. To get it to do the right type of research for me. And basically, I get this output, which is the SWOT analysis, strengths, weaknesses, opportunities, and threats. And then it also writes up the marketing opportunities.
I'm going to be honest. I was a little underwhelmed by this particular example. I've done one since then that I got better answers with, but if you look at this, you'll see there's some good stuff in here, the residential only focus, and build relationships with local businesses, which is something I might not have thought of, and embrace technology, such as tank monitoring systems, so on and so forth.
But there was also stuff that didn't make sense, develop a strong online presence, which they already had, offer competitive pricing and promotions like. Duh, so a little bit of a mixed bag and if I was doing this for real, I might push back and I might say, hey, listen, no offense, but that's some freshman level marketing stuff you got there, Gemini.
I want you to dig deeper and I want you to find ideas that aren't so common. AI responds to those kinds of commands, so it will go out and it will try and find a little bit more obscure stuff that might be much more valuable. Again, if you're not happy with the output, keep on iterating.
Alright, let's shift to search engine optimization, always one of my favorite topics. And, for this, I'm actually going to move away from ChatGPT and some of these bigger and less expensive LLMs, and I'm going to move to a very specific tool that I like called Market Muse. And there's a couple of ways that you can use, or at least that I use Market Muse. The first one is for doing research into a new topic.
So here I have I'm doing some research into the topic compounding pharmacy, and I'm here in the topic navigator. And if you've ever used a tool like Ahrefs or Moz, you've probably seen things like this, where it's going to go out and do research into the search volume and competition. And the way that MarketMuse works is that it takes a look at the top 20 results for the different keywords.
So not surprisingly, one of the main related keywords to compounding pharmacy is compounding pharmacy, but you can see further down there's other topics as well. But there's the word and then it tells me the monthly search volume. If I'm interested in doing some paid search, it tells me the typical cost per click of that particular phrase.
And then if I'm planning on writing copy about this, it tells me the number of times I should plan on using it on this page if I want to rank well, given the current competition. So that's one way in which you can use it, and then actually so this is, we were in the topic navigator, now we're shifting over to the SERP, the search engine results page, x ray, and I can start to see things like again, the top 20 pages, and this top box is the average, so I can see that, these pages tend to have images, videos, 7 internal links, less than one external link per page, which I found surprising.
Most of these pages serve the intention of knowledge or information rather than transaction. And I can see the average content score, the average word count, and the number of H2 tags per page, which is all important stuff if you care about search engine optimization. And then 1 through 20, I can look at every single individual page and see what the score is for that particular page.
The other way you can use it. is if you already have an existing page, you can see how it should be ranking or how you can improve it to get into the top 20, hopefully the top 10, hopefully number one by comparing it to what is currently ranking for that keyword phrase. So, in this case, I actually narrowed it to compounding pharmacy, Maine.
I put in a redacted URL, specific URL. What happens is once I put it in there and then click that run fetch button, it will go grab the page and pull it right into here and a text editing document. So, you can see it here and It gives me my content score, which in this case was 16.
The average of the top 20 pages was a 15 and it's telling me my target should really be 22. My word count, the average word count and the target word count, which is actually very strange in this situation because it's so low where mine is a lot longer. So, this might be a situation where I don't necessarily agree with the algorithm of the AI, and I would probably ignore it.
So, we come down here and you can see that this is, these are the words that were literally pulled from this page and you can see compounding pharmacy has been highlighted because that's one of the phrases that tends to appear on the top 20 results. And you can see other phrases too pharmacist and dosage forms are all on this page too. They've been highlighted because they appear on these top 20 results page. Over here in the left column, I can see all of the phrases that they suggest we use on this page, and how many times I'm actually using them, and then how many times I should be using them based on those top 20. And then what I can do is, so for example, we're actually, or they're not, using the phrase compounding pharmacy Maine, at least those three words, right in a row.
I could go in and say independent Compounding Pharmacy Maine and Vitamins and Supplements Store, which I wouldn't because that doesn't read well. But if I added the phrase Compounding Pharmacy Maine, suddenly my mentions in real time would go up to one to two, and then I could keep on adding it to get it if I wanted it even higher.
So, I can keep on editing this page and my scores, both in the columns here and then up at the top of the page, will keep on changing, getting me to a point where it doesn't have to be perfect, but that I can improve my rating on this page. It doesn't publish it back to the page, so I have to copy and paste this back into my WordPress or whatever website CMSM might be using.
But I've done this in the past and I've seen improvements on pages ranking. So, this is a great tool to use. Especially if you've already got content, but maybe it's not ranking as high as you want it to. This AI power tool is fantastic. Like I said that one's a little bit more expensive. That one starts off at $1,500 a year and goes up to $5,000 or more a year depending on the level that you're at.
But here are some ways that you can use ChatGPT to improve your search engine optimization as well. So, I go into ChatGPT and I say what are subtopics missing from top ranking pages for the phrase compounding pharmacy. ChatGPT goes out and pulls me back some information on here are some of the topics that it feels should be covered by top ranking pages for that phrase, but Aren't or underrepresented in other words; these are what we call content gaps So it's in my best interest to create content around these topics because people are likely searching for them, but they're not getting great results. So, this is a great opportunity for me to elbow my way in to the top results for the phrase compounding pharmacy by talking about these topics and they include things like regulatory compliance and standards patient education, quality assurance, and so on.
Another way that we can use chat GPT to improve our SEO is by asking it about schema markup.
Now if you're not familiar with schema, schema is a way to organize information on your website in a way that search engines understand it. And as it turns out, in a way that AI tools understand it as well. So, it's a great tool if you want to improve your search engine optimization. And there are tools that will help you build this.
And this is something that we often do for our clients as well. But let's say, you're not sure how to do it. So, I just basically ask ChatGPT, generate FAQ schema markup for an article about brown tail moth rash. So, in other words, I know from personal experience that one of the types of schema cause there's different types.
There's recipe schema. There's event schema. There's like hundreds of different types of schema out there. You can Google and find a long list, but one of them is for FAQ sections of web pages. So, I say generate FAQ schema markup for an article about brown tail moth rash. So that's what I'm planning on writing about.
And seconds later I get, Hey, here's a FAQ schema markup in JSON-LD format for an article about brown tail moth rash. And I look at it and I'm like I'm sure that's great, but. I don't really know JSON-LD, I just want to put this in HTML code on the back of my page. So I write, I'm looking for that for my web page, can you do it not in JSON?
So even though that is probably the most poorly formatted prompt ever, chat GPT response. Sure, here's an FAQ schema markup using microdata instead of JSON. You can embed this directly into your web page's HTML, and it gives me all the copy I need. And I'm like, that's great. But now it's got me thinking, because if it didn't understand what I was looking for the first time, maybe it still doesn't understand it.
So, I opened up another window, another chat GPT window, and I posted this question, this prompt. Is the following good schema for an FAQ page? And then I copied and pasted what I got from the other chat GPT right next door. What was the result? Your schema is well structured, but has a few issues. And then it pointed out the issues, and then it wrote the better FAQ schema markup.
I'm like, okay, but how do I know it's better now? So, I asked the same question by opening up a third window and I got back basically the same thing could be better. I did it again a fourth time, and a fifth, and a sixth, and a seventh, and finally an eighth, which is when I gave up. Every time it said it was good, but could be better, and gave me improved copy.
But sometimes it seemed to be reversing what it had already said was better copy. And this is another important thing to keep in mind. I'm sure any one of these seven or eight versions would have been good enough. But if you're really looking for the optimal schema, then either you need to B, have some expertise in schema to know which one's best or be willing to go look it up to see is there one that's actually preferred by the search engines and by AI tools out there and then you would know which is the right answer.
So again, we can't get away from keeping humans in the loop. So, let's shift from search engine optimization to conversion rate optimization, also known as CRO. This is also basically just trying to get more people who visit your website to take a desired action.
Here's a prompt. If I've got, let's say I have a landing page and one of my goals is to get people to download something on the landing page I can say to ChatGPT, you are a conversion rate optimization expert. You've been provided with this URL and asked to provide recommended improvements to increase the chances that site visitors will download the three-part spec. Please provide a list of recommended changes. See how I like to make ChatGPT feel good about itself? Now, as a side note, giving it a role, like I've done here, is a common approach to better prompting.
Generally, if you tell it that it's an expert in whatever you're looking for, it will actually rethink things in a way that persona would do it, and you'll get better results because of that. That being said, there are hundreds if not thousands of videos out there about creating the perfect prompt.
Many of the prompts that I'm using today tend to lead more towards the simpler side, basically for the sake of this webinar. But well structured, detailed prompts are definitely the way to go, but getting to that point shouldn't hold you back from a lot of the things that we talked about today. You could always improve, but you might as well get started.
So, I pose that prompt, I gave it a URL, and basically ChatGPT takes a look at that page and comes back. And gives me some specific advice for that page. Enhance the call to action. Talks about where it should be placed on the page and the design of it. The compelling copy. It tells me I should simplify the form.
I should have less fields on it. I should make some of those fields optional. Tells me I should be incorporating social proof with like testimonials and certifications. So on and so forth, you can take this document and now hand it off to your marketing team or your development team and have them get to work.
Now, there's no guarantee that this will actually make your page better, but it's definitely worth experimenting with for sure. It's using best practices. In fact, as you can see, it actually names checked Unbounce and Hotjar as part of its work, and those are two CRO companies that provide CRO software.
Now, maybe you've got your own contrarian approach to CRO You could train your custom GPT or using these prompts, you could train ChatGPT to behave that way. Maybe you believe that long form sales letters are the only or the best way to convert people. So, you could tell it that, so it would always be prompting you to create longer form content.
Or you believe that only video testimonials will move the needle. So, you tell it to look for video or promote the idea of getting video testimonials rather than written testimonials. Maybe you believe that every call to action should use the word me rather than you. You train it up in what you think is the best practices if you have the expertise. And from then on, your chat GPT or your custom GPT. Will respond in that way,
So, I want to wrap up today. The last thing I want to talk about is creating custom AI assistance. These are basically you're taking a large language model like chat GPT or Claude and you're focusing it. You're getting it to become an expert in one area at the expense of many of the other areas. So, if you're on chat, GPT, these are called custom GPTs. If you're on Claude, they're called projects. If you're on Gemini, they're called gems, but they're basically the same thing. I'm going to show you how to do it create a custom GPT right now. So, I'm pretty sure this only works with the paid version of chat GPT and 99 percent of the AI tools out there, the subscription is $20 a month. For $20 a month to have this tool at your disposal, to be able to have something that can help you do SEO and CRO and run automations for you. It's a no brainer like just get it, just pay for the 20 bucks a month.
It's such a small business expense in the scheme of things. So you can get to your custom GPTs by logging into your account, you can see up here in the corner there's my account and then in the pull down you can choose my GPTs or you can click on the explore GPTs, either way you're going to end up on a page that looks like this that lists all of your GPTs if you've created any.
And if you want to create a new one, you just click on that, create a GPT. So, I click on that and I'm taken to a page that looks something like this. And if you're just getting started, the default is create. And with create, you're basically having a conversation with chat GPT to create your custom GPT.
It'll ask you questions, you respond. And if you're, if this is your first time doing it, that's a great way to get started. That's the first few that I did as well. But once you feel more comfortable and after you've created a few you're probably going to want to switch over to configure where it's going to be less conversational but you're going to be able to put content in right away and get started. So, here's an example of one that I created which helps me Create the promotional assets for my podcast.
Once the podcast interview is done. I created a custom GPT to help me with everything that comes after that. I gave it a name podcast producer I gave it a description. It's a producer who creates and manages the agents of change podcast assets. The next section is instructions. And the two really important things in these custom AI assistants are instructions and the knowledge base.
So, the instructions are the detailed instructions of what you expect this chat GPT or this custom GPT to do, to perform. And then the knowledge base, which we'll get to in a second, is basically any supporting documents, the instructions can only be so long, but the knowledge base allows you to create.
I wouldn't say unlimited amount of information for it to pull from, but you can put a lot more information into those documents and then have the instructions reference those documents as part of the process.
So, this is what my, my instructions look like you don't see the whole thing here just because you'd have to scroll down, but you can see I identify certain things as all caps and then give it a little bit of a description like role. You are an experienced producer and promoter of podcasts in charge of creating the assets and promotional pieces for the Agents of Change podcast or AOCP.
And then I identify other things like, SEO keywords. That's one of the things that's important to me as part of this process. SEO title, meta description, intro paragraph, so on and so forth. And I give it a little bit of information about each one and then I tell it the order that I want it to create everything in.
This is something only you'll find in ChatGPT conversation starters. And these are, if you're going to be using this ChatGPT for a number of different type of things regularly, then this makes sense. They become buttons on this side, but to be honest, if you're just going to use it like I do for one thing, you don't need to do this.
Also, a lot of people like to publish these. Some of these you can even buy from other people if you wish through their OpenAI store. So that might be a case where you'd want to do those kind of buttons and things. But again, not something that I find super important. And then there's the knowledge base.
And so again, I find that some of the best ways to get AI to do what I want it to do is to give it examples, to give it a, in this case, what we have here is we've got a uploaded document for the agents of change buyer persona, my voice document that we talked about earlier, although that was for flyte new media, I grabbed the about page off the agents of change website and I uploaded something called social media.txt.
Once you start uploading documents in chat, GPT, suddenly the additional settings setting pops up and it's pre checked. Use conversational data in your GPT to improve our models. I always uncheck that. I don't want to be training their models on my data. Now just because I've unchecked that, I'm still not 100 percent sure that they're listening to me when they say that, so best practices would still be to anonymize that data, but, you do what feels right for you.
There's nothing in here that was too private, so I didn't really worry about it, but I still don't want to have them trained up on my data. Now, I did this in Custom GPT as part of a Keystone project that I was doing when I was taking an AI course. And I found that despite me saying, I don't want you to stop until you're finished, ChatGPT kept on stopping at each phase and being like, would you like me to continue?
So, I did the entire thing again in Claude, because I have an account over at Claude as well. You can see it's basically the same thing. I copied and pasted the instructions into a field here in Claude. I uploaded three of the documents. And then as soon as I get the transcript from one of my interviews, I upload that, hit the go button.
And Claude gets to work. And it gives me things like, it gives me this document that includes the SEO title for the show notes page. The meta description, the introductory paragraph in my voice, five key takeaways from the episode, an email promoting the episode, which I almost never use, Facebook, Instagram, and LinkedIn post, a prioritized list of SEO keywords based on the transcript and general knowledge about SEO, and then in that little block below, an entire blog post written in my voice that we can copy and paste into our show notes page.
But of course, I want to read it first to make sure it makes sense. And then it also gives me a bunch of social media quotes, which we can use basically as text clips. But I can also take those, and I even had Claude figure out the timestamps for those. And then I can bring it over to a program called Kapwing, which is a video editor.
And that allows me, I never take good screen captions, by the way. That's just my fate in life. But it can do things like dynamic captions, and it provides a transcript, and you can edit your audio and video in this tool just by editing the transcript. So, if you have a great paragraph, but there's a sentence in there you don't like, you just delete it just like you would in Microsoft Word, and the audio and video just disappear.
It's really a slick tool. Now before I wrap up today, I just want to give you a list of my AI toolset these days. These are all subscriptions that I'm currently paying for. And when I made this list, I realized that probably I should start to use fewer tools, but ChatGPT, Claude, Gemini, asterisk, because I'm using a free trial right now, MidJourney, which is my favorite AI generation tool. Magi, which we didn't talk about today, but is a really interesting program because it allows you to use ChatGPT and Claude and other text editors as well as image generation tools and even a couple of video and audio generation tools all for $20 a month. So basically you're able to access all those tools through Magi, and if you're not sure where to get started, that's a great tool.
It's like a smorgasbord of AI. Market Muse, which we looked at. Fathom, which we didn't get to, but I use that tool as much as any other tool in the world. It's an AI notetaker and it connects with Zoom. And descript, which is a great audio and video editing tool I use for my podcast all the time.
And then Kapwing as well. And that doesn't even include Photoshop, Canva, MS Office, Microsoft Office, Google Drive, or any of the tools that I was using previous to AI that now have AI built into the software.
Also, I just want to give a little shout out to my friend and coworker John Paglio. He has another great webinar coming up on April 29th.
SEO is being changed every day and especially now that AI is impacting all of our results and a lot of our traffic coming from the search engines. You'll definitely want to sign up for this one. It is also free, SEO betting odds, where to place your money in 2025. So, check that out. When I send the email recap for this webinar, we'll have a link so you can sign up for that.
So again, as you think about today, hopefully you have a better understanding of how to use AI to create buyer personas, to create your brand voice, to run a competitive analysis, to do SEO and CRO, and to create custom AI assistants. That was a lot. And there's still a lot of things that we could cover. I did see that I think we had some questions going on.
So, I'm going to stop my share right now. John, maybe you can help me, but I'll see if I can't just pull up the Q& A myself. And I'll, I guess I'll just go through it.
So, Jeremy asks, wonder if you think clients care about how the AI they use is built, where it's hosted, etc. Great question. We've been having a lot of conversations internally about AI and how transparent we want to be with our clients.
And our decision is we want to be absolutely transparent. Like we're using these tools. We're different people in my team are using them at different times. Levels. I use them quite a bit, but I do less client work than a lot of the very talented people on my team. But I do think that if you are using AI, somewhere in your documentation, you should let people know if you're not like just telling them up front hey, we did this with the help of AI.If you're creating content for people that's AI generated and you're barely overseeing it, I think you need to be transparent about that. Hope that Hope that helps.
Cybele asks, what say you about confidentiality when it comes to sharing your private buyer persona with AI? So hopefully I trust that a little bit, but I think you need to be very careful with any sensitive data, whether it's your own or customers because you don't know where it's going to go. And I've, Google told us that anything that we did while private browsing was on that no one, not even Google would know about it. And it turned out they were lying to us the whole time. I hope a open AI can be trusted, but I've seen some of the people that run that company. I'm not so sure.
So, I would say, depending on the level of sensitivity, consider anonymizing it. So, for example, let's say that you have a bunch of Google analytics reports or P and L reports for clients. You should probably anonymize their names before uploading it and asking AI to provide some reports or recommendations for you.
That would be the approach that I would take. Some other stuff that's publicly available I wouldn't care as much, but these days I'm trying to err on the side of caution. The next question is, and by the way, if you have questions, please go ahead and fire them off.
AI detection software is now readily available as AI itself is there any reputational, risk to using undiluted AI content, even if it appears human and completely congruent with our brand. So first of all, there is a lot of debate. Whether AI detection software can actually detect AI and how many false positives it gets. I was just talking to a buddy the other day and his daughter, who I think is about 12 or 13 years old got like a failing grade because the teacher said when she ran it through AI software, it said it was fraudulent.
This young woman was savvy enough to show her Google Doc and step by step how she created this, the research, the iterations, and it was evident that it wasn't AI and the teacher I wouldn't say she apologized, but said, okay AI makes mistakes and gave her 100 on the paper. I'm not 100 percent sure that we can trust AI detection.
I think it's a weak tool at this point. Maybe it'll get better, but then I think AI will get better to not be detected by other AI. It's going to be an ongoing battle. But is there a reputational risk to using undiluted AI content? Absolutely, I think there is, because we talked about bias, and we talked about hallucinations.
And I can't imagine trying to run a company that wouldn't have human oversight of the AI output. I think this is an assistant level tool at this point in history, and I would not be trusting it to go off on its own.
Elizabeth has a question. Once we put our brand voice info into AI, is there a way to protect, or keep private our work inputs and outputs from AI as it's learning our voice and work, we want to benefit, but we don't want our copywriting and content shared with the world.
So again, if you use the way that I showed you it, when I was creating the custom GPT, I think that's going to be helpful to you, but anybody could take your examples of your writing style from content that you've published publicly and create their own version of your brand voice. So, it's not. that difficult for somebody to reverse engineer what you've already done. It’s more about just making sure that you've got a tool out there that makes sure that, especially if you've got multiple people working on your copy, that it all has a consistent brand voice. I think that's the goal, rather than trying to protect or copyright or hide that content.
Let's see, I think I may have lost some of the order, maybe they're being taken away as I answer them.
When we talk about creating custom AI, say in chatGPT, what do we actually mean? You mentioned it in relation to personas. Is it an individual account that we fed loads of information to around, say, our brand, writing style, etc.? One second, please.
All this talking got me thirsty. All right. So I think I went through that. You are taking the LLM and you're fine tuning it to be really good at one job, whatever you lay that job out as. Hopefully hopefully I answered your question when I showed that example, but feel free to ping me again if that's not the right answer.
Here's when I discover the hard way. Cybele asks, does ChatGPT remember your prior conversations when you've signed in? Can it recall things you've asked, and it's answered? Yes. It updates its memory. And you can go into ChatGPT's memory, I'm not sure if all the platforms are the same, and ask it. There's a section in that pull down menu where you log in up in the top right corner, and you can see its memory.
And I went through it because I was, I discovered because I was having some conversation, I wanted to start making hot sauces. And I said, Hey, can talk me through the process? What do I need to buy? What kind, what are some simple hot, fermented hot sauces I could create? And it said, and this is a brand new conversation, and it said since you're already growing habanero peppers, ghost peppers, and Carolina reapers in your garden…
And I'm like, I didn't tell you that. I know I told you that, but I didn't tell this instance of it. And that's when I discovered that some of the information that you share with it. Does go into its permanent memory so it can better help you can shut that feature off. You can go into that memory and edit things out and this is important because if you are in a business where you have a lot of different clients If you're talking about one client that might get into the memory and color everything else that you've output Which is why custom GPTs are so important. And now they've got this new thing in chat GPT called projects so you can give, I haven't played around with it too much ,but so you can give custom instructions and that may be one way to organize different customer or clients so that you are keeping that thing separate. But right now, I’m using chat GPTs to really keep them separated as the song goes.
Paula asks, which are the apps that the flyte team uses most? And does each team member have their own or is there one shared across the company? Again, we're still in the early stages of our AI adoption here and we're figuring out as we go along. I have my own copy of everything because I'm the guy who's really most excited about this, but my team, we do have a team.
We have a chat GPT for everybody else in the company and they have organized it in their own way. It's not the team version. That's $200 a month. We're getting by right now on the 20 version a month, but we are developing out a prompt library. And we're hoping that we will create, that everybody will have their own area to play with, but we'll also have areas where it's these are the prompts that we use at flyte new media.
So hopefully that is helpful.
And hopefully I showed you at least my favorite tools. Other people may have their own favorites.
Bill asks, Do you know of any AI that can do phone calls with minimal errors? When you say phone calls, do you mean that they can actually speak like robocalls? If so, I've definitely seen that they exist, because we've all gotten those calls, but please don't be that guy.
If you mean Ones that are great note takers and then you can get the whole thing transcribed like a zoom call, then yes, I use Fathom constantly. I always ask for permission before I get on a call with somebody that yeah, I can use my AI note taker. I've never had anybody say no, but those are great recordings, and it really helps me because I'm a terrible note taker to make sense and make sure that I didn't forget anything that happened during the sales call. Feel free to Ask a follow up question, Bill, if that wasn't your, the answer you're looking for.
Marydale says, what's the most surprising and valuable thing you've learned with when experimenting with these tools? What a great question, Marydale. I've had a lot of surprises, but I think one of the most important things that I've learned Is, you used it, you used the word in your question, experimentation.
Like I really think it's critical to just start using these tools and see what you can do. Like that schema thing was brand new to me. Another thing that I've discovered is if you get stuck or you're not sure how to get started, Ask ChatGPT. I've often asked it like what is the prompt that I should be using here if I want to accomplish X, Y, and Z?
And it will start to shape that and then I can use that and continue to iterate it. So, I think there's a reason why the, it starts with the word chat. And I really think that you should think about it as having a conversation with somebody with a great amount of expertise in a wide variety of areas that you can use to your best abilities. So that's probably my answer.
Somebody asked. Related to the reputational risk to using AI. If people realize you're getting AI rather than humans to create content. Yeah, I know. I agree. If anybody asks me, hey, do you use AI to create content? I use it right up until the point of, and I've tried everything. I want to say I've experimented. I had chat GPT, an entire article in my voice that I knew nothing about. It's probably the worst article we've ever created. It's I think still up on our website. I don't want to tell you which one it is. It was absolutely horrible. And I spent more hours editing it than actually if I just done the research myself and again, if somebody is paying you to create content for them, and you're literally, after you set up the prompts, just pressing buttons, like that's not valuable.
That's not anything. It's not worth anything. So, I would not risk my reputation or my company's reputation by just basically having a simple prompt and asking somebody to pay for it as if human beings were involved with it. That's my take on it. Obviously, different people may have different opinions on that.
Some people have no problem just saying to chatGPT, I need an article on the history of cornhole. And then they put it up to their website with absolutely no corrections.
Bill asked like robocalls, but the customer signed up for it because they were interested. So, I think what, it's not an area I have an expertise in, but I'm sure there are automations that can be set up to send out phone calls.
And there are tools, and ElevenLabs is a tool that does a lot of voice AI voice powered stuff. That's probably where it would start, but it's just not an area that I'm an expert in. So, I don't unfortunately, Bill have any additional information. other than that.
Hey everybody, we are at the top of the hour. I want to respect your time.
I've also got other meetings coming up today. This has been great and a lot of fun. And if you, we will send a follow up, if you have topics that you wish I got into today that I didn't, I'd love to know what they are because I plan on refreshing this webinar on a regular basis. If you have other digital marketing topics, like we're doing the SEO and coming up that you'd like to see us handle, please again, just reply to me when that email goes out, let me know what you're thinking. But I just want to thank everybody today for the time and for your attention and really appreciate it. And we'll be sending out if you missed any of today, or you just want to watch it again. We will be sending out links.
So when, once it's ready to be watched on demand, we'll share that with you and feel free to share it with your team, your friends, your dog walker, what have you. Have a great day, everybody.