Brent Weaver: And welcome to day three of the Cloudways Prepathon 2024. Hopefully y’all have had a great time the last two days. I want to give a special thanks to all of our speakers and panelists and all the Cloudways team so far, and of course our amazing audience. Maybe let’s give ourselves and our speakers a great round of applause. Thank you, Solem Sol, for being here. So great to have you guys here the last few days. How many of you here were with us over the last two days? Laden’s back, yep, loving the chat. Maybe we can just kind of light that up, let’s welcome everyone, say hi.

We’ve got an amazing day ahead. We’ll talk about some of the things we’ve got on the docket for today. Yesterday was absolutely fantastic, revenue and growth, those are two of my personal favorite topics. Today we’re going to be focused on AI and innovation, so we’ve got some amazing activities planned for you today, we’ve got some awesome speakers, we’ve got some additional games. Hey, what’s up Christopher. Hey, what’s up Lindsay. Sebastian, been here for most of the talks, love that. Let’s keep those emojis and memes flying today, let’s keep the internet interaction going.

We’re going to be doing some fun AI innovation challenges and quizzes and activities today, of course in between our amazing keynotes. And we have a special guest for us after we announce the contest winners this afternoon. The team has been so amazing to give us a little bit of a Cloudways Prepathon afterparty. We have a comedian joining us by the name of Tara Brown. She performs clean and family friendly comedy across the country, she’s been on some of the stages with the best comedians of our time. She’s been recognized by some of the major news organizations out there, Charlotte 5, Newsweek, New York Post, Charlotte Observer, Lancaster News, all that good stuff.

So we’re going to have an amazing day, we’re going to end with some fun after we announce the contest winners at the end of the day with a little standup bit from Miss Tara Brown. So I’m personally looking forward to that. It’s going to be a good time to kick back, relax, really enjoy all this amazing content that we’ve had here at the Prepathon.

One other thing before we go into our first speaker, Mr. Andy Crestodina, who’s going to be giving a fantastic presentation on AI. He’s been one of my favorite speakers recently on AI. I’ve had the chance to host him on our podcast, the Digital Agency Show. But before we get into Andy’s presentation, I just want to give a quick call out. If you haven’t yet gone by the booth feature within the platform, make sure you do that today. We’ve got double the points and double the prizes for folks checking out those booths today, so we’ve got some additional gift cards, hosting credits, all that good stuff, and you’ll get additional points for the overall leaderboard contest if you hit up those booths today. So make sure you guys do that. You’re going to get double the points, double leaderboards today, so explore those booths while we’re hanging out here over the rest of day three of Prepathon.

All right, I think I’ve covered all of our announcements for the day. I’m excited to be hanging out with you all in chat throughout the day. We’ve got some fantastic guest speakers coming up, but we’re going to get started here today with our first speaker, with Mr. Andy Crestodina, here in just one moment.

Day three of the Cloudways Prepathon, so good to see so many agency owners, e-commerce businesses, SMBs engaging with us over these past couple of days. Really excited to be introducing our next speaker hanging out with us. We’re going to be continuing our theme on AI and innovation. We are really excited to be welcoming Andy Crestodina to our stage. Andy is a co-founder and CMO of Orbit Media, an award-winning 50-person agency based in Chicago. He’s helped over a thousand businesses with his digital marketing advice, he’s written over 600 articles on content strategy, SEO, GA4, and AI. If you guys were here with us on our Web Growth Summit, Andy gave a game-changing presentation on the changing landscape in SEO with the invention of AI. So we’re really excited to be welcoming Andy back to our program here. Andy, good to have you here on Cloudways Prepathon.

Andy Crestodina: Thanks for having me, Brent. I’ve really been looking forward to this. I’m a fan of this event and excited about this topic. We’re going to do AI for B2B social media, which is super interesting, lots of angles, and we’re basically going to just rip through a bunch of specific use cases and AI methods for getting better results in social with AI.

Brent Weaver: Very cool. I love when you share your stuff, man, it’s like real hands-on actionable tips and hacks and techniques. So I’m ready with my pad of paper and notes, I’m going to be taking some notes and implementing stuff for my own business, so really looking forward to it. Andy, you’ve got the stage, man.

Andy Crestodina: Great. Let’s hope not to disappoint you, Brent. If you like it, it’s good, because you’re a pro at this and you know what you’re talking about. Okay, I’m going to click this share button, share screen, share entire screen. And this is recorded, of course, so I’m going to go kind of quickly, but there’s a lot. You’re getting all this content anyway, and you can go back and watch it in slow motion or copy and paste the prompts, whatever you need.

I want to start with this because AI puts us in kind of dangerous territory with social media, because it reduces the cost to do certain things, like generate content, including things like comments. And I am not fuzzing out the names, these are actual comments, and you can see these are people who just bought a tool that is automatically jumping into streams and leaving the same comment again and again. Look at that comment spam. This is AI generated comment spam. I don’t even know if this needs to be AI, honestly, an everyday algorithm could do stuff like this. Not good, not happy, not healthy, not good for the internet. This is antisocial media, it’s the opposite.

I mean, this is clearly spam. If I could say that there’s one thing that should not be done by a robot, it’s the comment, because that’s supposed to be your point of view, that’s your chance to have a conversation. Outsource tasks, but don’t outsource your voice, please. And I’ve seen it getting a little bit more sophisticated. People are using tools that generate a bit more, including this copy and paste thing, it’s clearly just filling in the blank. I’m seeing this one after the other and they’re just kind of all the same in a stream, and it’s not good. Okay, so let’s not do that.

I’m going to go through methods that are much more ethical, sustainable, effective, and you’re going to notice that it’s not mostly about generating words. I’m going to use it for analysis, I’m going to use it for ideas, I’m going to use it to make me a better marketer, from two perspectives: one, to get results for the brand, two, to get results for yourself. The last third of this is going to be mostly about personal branding, which I think we should all pay some attention to here.

So first, how AI can discover topics. This one will take a second. This is the biggest one, and I’m starting with it because I believe strongly in this. There’s good research on what formats for content drive results. This article from years ago, published by Steve Rayson, co-founder of BuzzSumo, changed my content strategy forever. Here’s what Steve discovered after analyzing a million articles. He’s like a statistician. He said if you want to create content that creates a high level of both shares and links, we’ll call those key performance indicators for social and search, then you should concentrate on one of two things: opinion forming authoritative content, or well researched and evidenced content.

I’m going to paraphrase Steve here for a moment: strong opinion and original research. Now I’m going to say a strong opinion is what gets the best traction in social. I’m going to say original research is what gets the best traction in search. Nothing attracts links like original research pieces, it’ll always be part of my content mix. But strong opinion triggers response, and responses trigger social algorithms.

Here’s an example. I wrote an article about what you remove from your website. I make a case here: press releases don’t belong on websites, they’re not digital content best practices at all. “For immediate release,” what does that mean, am I not supposed to read this yet? That little crappy blurb about the “about” section at the bottom, this has nothing to do with digital content, this is a press release, turn it into a story, you can do a better job. PDFs, why do people put PDFs on websites? Maybe as an alternate version, but everything should also be HTML. In other words, I’m taking a stand on some little mundane things, and one of them, I said don’t put dates on blog posts. You can put the date in the text, but don’t use the blog template for the date.

Okay, I make a video for this: 13 things you should remove from your website immediately. This video goes on YouTube. 775 mostly negative comments, people are very worked up about this. But that’s what thought leadership is, that’s what strong opinion is. I’m going to use Seth Godin’s definition of thought leadership: it always creates attention, it’s about making assertions, you have to be willing to be wrong, you can be certain that some people will disagree. Think about that for a minute, if you can’t disagree with it, it isn’t thought leadership. That’s the point. Draw a line in the sand, plant a little flag. That’s what the best content marketers, the best social media marketers are doing, because that’s when people believe, that’s when people can really connect.

Okay, so this goes on YouTube and people start calling me sleazy and untrustworthy. I’m most secure in myself, so this isn’t bothering me too much. This guy’s got all caps, hates my recommendation. This person said I told people to make low information, useless blogs. “Do not remove the date,” in all caps. Someone said their beautiful wife won’t even watch it, that the comments told her the advice is terrible. It goes on and on. This person says I don’t know what I’m talking about, I should let a real experienced UXer talk. I’ve been part of the planning process for more than 1,000 web projects in my career, 24 years, I’ve made hundreds of analytics accounts, I do have some experience here. Anyway, it goes on and on. Some people give me good feedback, some people give me irrelevant feedback. This person thinks I dye my hair. I don’t dye my hair, for what it’s worth, just for the record, I don’t know why you’re going there.

But my point is this: YouTube thinks that it’s worth watching, and they showed it to 300,000 people. That’s more than 100x my typical YouTube video. I’m happy to get 2,000 views, that’s good for me. 300,000, why did that happen? Because the engagement, the topic triggered the engagement, the engagement triggered the algorithm. So my first advice here, and the best thing that you can do in social media, is to understand how to take a little stand on something without necessarily going into controversy. And that differs from AI, because ChatGPT can’t throw a punch, but it can help you find provocative topics.

So I’m here to share with you now my topic research, mundane but provocative topic research prompt. I’m going to do it for this company, I don’t know this company, it’s a spaceport. Okay, we’re going to be a spaceport for a minute. We’re looking for topics to connect with our audience, commercial satellite operators. Ready? What are some relatively mundane, almost trivial space industry topics that professionals have very strong opinions about? These answers are fantastic. Imperial versus metric units, oh, that’s going to start a big conversation. How do they name spacecraft and celestial bodies, that will divide people, people will take a stand, like, is Pluto a planet. Rocket design philosophies, reusable versus expendable. These are going to trigger my audience to have bigger conversations. Just use that provocative but mundane topic prompt, and you’ll find out in your category what people really care about.

This one’s for a company that targets office managers. You know what everybody cares about? The temperature in their office, the thermostat, they’re going to get in fights over this, that’s a provocative topic. Open floor plans, are they really that good? So you can just imagine what this does to your marketing, what this could do in social media. Again, we’re not talking about geopolitics and social justice and gun control, I’m trying to just trigger a little interaction here, I’m trying to give people something to talk about.

I’m also not trying to be correct, in a way. That comment I said, where you shouldn’t put press releases on your blog, I know a woman who totally disagrees with that. So I got her to provide to me a contributor quote, I put it right next to that piece of advice. Andy, why would you want other people to dissent, disagree with you in your article? Because I’m trying to start a conversation. It’s like moving the comments from under the blog into the blog. I’m just trying to have a conversation, I’m trying to give my reader something to talk about.

Here’s some more prompts. What are people in your industry afraid to answer, fantastic insights. What false things do people in your industry think are true, fantastic insights. What are the most common assertions that are least likely to be supported by evidence, might find some research related topics. What are important things that are not covered, humans can’t do this, that’s an AI only method. What are important topics that are unlikely to be covered by the big blogs, the only way you could do that is by reading every blog in the world, impossible. What counternarrative opinions are least likely to be discussed by thought leaders, that word is powerful.

This is how an experienced content strategist can write good prompts that maybe a junior person can’t. The junior person isn’t thinking about counternarratives, but the pro social media marketer might. So write the prompt and give it to your junior marketers. Here’s an example: what common assertions are least likely to be supported by evidence, what counter narratives, and it comes back with these things, these are topics I could potentially cover.

I learned this from Brook Ellis, in the social penetration theory. This is a well researched thing, it’s like an academic topic, that is how people connect, that is what’s memorable, that is what creates an emotional bond between two individuals. It’s not cliches, it’s not facts, it’s opinions and core feelings. That’s what you’re going for in social media, that’s what matters, at least some of the time.

Okay, an even better way to do this is to first give AI your persona, which we’ve talked about, and then talk to it about topics. You can learn the fears and hopes and secrets of this audience. Here’s what a persona prompt looks like, you can deconstruct these and use them for yourselves. Better yet, upload your battle tested ideal client profiles. Build me a persona of a job title, commercial satellite operator, at an industry, global telecom, company size, billion in revenue, responsibilities and challenges, planning launch missions, needs and requirements, they’re considering a launch services provider, they need to deploy communication satellites.

Okay, AI, give me four things. I want to know the hopes and dreams, I want to know the fears and concerns, I want to know their emotional triggers, and I want to know their decision criteria for selecting a company like mine, how they buy. This is the core of a persona, and it’s made from a simple prompt, breathtakingly useful. Once you have a synthetic member of your target audience, that’s amazing, I don’t know if there’s a better use for AI than this.

Oh, write me a blog post in 10 seconds, that’s the worst use of AI. If you use AI to write a blog post for you in 10 seconds, guess what, your target audience can also write that same prompt and get the article. That’s the last thing you should write. Opinion, perspective, point of view, beliefs, that AI cannot write. AI has no opinions, it can’t do it. Every day, most of the things you say could never be said by AI. I’m going to speak some sentences AI can’t write. Brent, I’m digressing here a tiny bit. Marketers overspend on technology. Pop-ups are bad for trust. Rice is better than pasta. Red licorice is not licorice. There are a million things that we all say every day that AI could never say, it doesn’t have a point of view, it has no background, it has no story, it has no beliefs.

Go ask AI what it really cares about. What does AI care about? Nada. What do you care about? I care about a lot. We can connect, that’s what social’s about, that’s what social interaction, online or off, is really about. Okay, I digressed, I’m back on track.

It gives us a persona, it’s probably wrong, fix it. Of course, don’t trust AI, it’s going to be incorrect, you’ve got to tell it what you know, tell it to improve the persona based on what you know to be correct. Now try these prompts. What widely accepted best practices does my persona secretly reject? What common advice does this persona think is harmful? What seemingly unrelated topics does this persona think has crucial lessons for their industry? What could the movie Titanic teach us about marketing automation, you get the idea. What commonly held fear or concern does this persona think is overblown or misguided? This is catnip in social media, this is it, this is decoding the psychology of a larger audience for triggering engagement, which triggers algorithms.

I think it makes sense. What unpopular view do they hold about a trending topic? I got these from Lisa Adams, great person to follow on the topic of AI. When I meet someone who really knows stuff I want to learn, I ask them if they wouldn’t mind doing a call with me every month or so. Lisa and I are constantly trading ideas, this is the best way to learn AI, I found.

Okay, moving on, moving on, we’re only at number two, it speeds up here, I promise. AI can generate images for you. These images are useful for social because they’re visually impactful, but they’re not very specific. Everything that I do is in ChatGPT because people have it. I’m not advocating for ChatGPT or Copilot or Claude, use whatever you like, I don’t care what you have access to and enjoy using.

So here, watch. Hey AI, make me faces of a small, diverse group of satellite operators in the control room during lunch. Wow, that is impressive, that is crisp and vivid, that is colorful and beautiful, that’s a beautiful image that didn’t exist 10 seconds ago. But can we pump it up? So this is where, my takeaway here is not that you can use it for images, but that you can keep pushing it to create more interesting images. Now they’re excited, it was a successful launch, look how happy those people are, they’re thrilled that the rocket launch was so successful, tears in their eyes, that’s a compelling picture.

Legally, you can’t copyright this in the United States because it had no substantial human involvement, and I’m not a lawyer, but there’s interesting aspects of AI generated images. They’re celebrating the most successful rocket launch of all time, they got out the champagne, the rockets are launching in the control room itself, there’s other rockets launching, what is going on here. The happiest rocket launch party of all time, filled with ecstatic satellite operators, the rocket launch was totally awesome, I mean super rad. See, I’m just throwing in words to try to push it in a different corner, you get the idea, this is what prompting is, it’s about the little verbal cues you’re giving it. Totally awesome, super rad, whoa, it went in kind of a cartoon direction, you can see what that did, look at how excited these people are.

I lost my diversity though, can we get that back, more diverse, I guess diversity means vaguely ethnic party hats. The rocket is launching inside the control room, people are being blown backwards, there’s a safety risk here, it’s making up new flags. Anyway, you get the idea, iterate on those image generation prompts, and don’t expect to get anything, you can’t control it, don’t try to fine-tune it, keep it broad. These are what I consider to be long shot prompts.

Moving on, method three, we’ve got 10 of these, let’s keep going. Find an AI to produce sound bites. If you took the transcript of this entire session and wanted to promote it somehow, you’d want to find the juiciest, most compelling, unexpected thing and use that in the social post. I’m generating a ton of language right now. Which of these things I’m saying are the most surprising, counterintuitive, compelling, memorable? Very hard to do, you’d have to watch it and pay close attention the whole time. Instead, get the transcript and give it to AI, what’s the most provocative thing in this sound bite.

I had a conversation with Paul Roetzer, founder of the Marketing AI Institute, and Chris Carr, founder of Farotech. Here we are, fun conversation, it went up on YouTube. Scroll down, I’m promoting this on social somehow. Grab the show transcript, copy and paste the show transcript into a text file, upload the text file, I can repurpose that very easily. Suggest articles based on that, interesting, AI search engines, rise of generative AI, navigating privacy, totally based on the article, not based on Bluesky, but the article.

Now I want to promote on social. Which statements in this video are people most likely to be surprised by? That’s what’s going to work in social, I need that sound bite, that quote, that counternarrative, that surprising quote. Which sound bite makes the best video clip? If I want to cut 10 seconds of this out and use that in social media as a short form video to promote the piece or get broader engagement, that’s exactly how to find it. AI is amazing at this, it doesn’t get tired, it will sit there and read every word in the transcript and compare it to the rest of the internet and highlight for you which things are more surprising. And it’s actually telling me on which lines, if you leave the time code in on YouTube it will reference that, and then you can make these surprising statements.

Some of them are pretty good, and some of them are garbage, and that’s fine. If 80% of what it gives you is trash, no big deal, the 20% might be gold, you only spent three or four minutes on this exercise, we’re not super invested. Don’t dismiss AI because something it gave you wasn’t good, nothing in this life produces something 100% perfect. And don’t get too excited about AI and think that everything it gives you is great. We need to calibrate our perspective with this technology, we need to have a relationship with it on some level that isn’t either “I drink all the Kool-Aid, everything should be done with AI now,” or “I don’t trust the accuracy, bias, labor market impact, copyright law, I hate this.” We’re just looking for value here, people, you’re just trying to get some utility.

I’ve done things in AI that were worthless. I spent 45 minutes trying to teach it to recommend internal links for articles, and it still doesn’t work, I’ll get it right. But many of these methods aren’t that much of a time save. I’m not looking to save time, actually, I’m looking for performance, I have no interest in efficiency, really, I’m trying to get 10x, 100x results. Sounds strange, not strange. The goal isn’t to be 10 minutes faster, the goal is to drive a marketing outcome. The top 1% in any category, in any channel, in any topic, in any format, get 100x results. Do whatever you can to get greater performance, don’t look to find deficiencies, look to find deficiencies and then fix that thing. I digress again, Brent, I’m back on track.

AI is going to analyze some social media data for me here. This is also kind of not a shortcut at all, but you’ll get good insights from this. I’m on LinkedIn, it’s almost my only social network anymore. Analytics and tools on my profile, I click, I land, I’ve got to choose a long date range, and I’m going to export this. Now, when you export a file, you’re going to look at that file and you’re going to see either structured data or unstructured data. Unfortunately, LinkedIn is a lot of unstructured data that is not that helpful. This needs a ton of work, and I want this to show topics, I want this to include the post, I wanted to do analysis on my LinkedIn engagement, and this is just links and dates, and not very good.

So I’m going to spend a couple hours, literally, I spent, let’s see, it was before breakfast, my kids were asleep, I sat there copying and pasting all the posts, tagging and assigning to categories, until I’ve got this, and this is perfect, this is exquisite, this is structured and ready for analysis. I’ve got the topic for each, the text for each post, I’ve got the topics and categories, I’ve got the format, is that a newsletter, is that a video, I’ve got all the social metrics, impressions, engagements, comments, and reposts, very useful.

Upload this to AI and it’s going to give me insights. You’re an expert social media marketing data analyst, you can find insights with data, which topic gets the most growth, engagement per impression. It comes back and does some nice analysis for me, it’s looking pretty good. Deeper, I need more. Draw a heatmap matrix showing the normalized averages of impressions, engagements, and comments for each topic. That chart was surprising to me. This is a normalized, color-coded heatmap matrix showing the relative performance of every topic I talk about in social media, topics on the left, the metrics on the bottom.

Look at what this found. When I talk about web design and UX, I get a lot of engagement, but those people never follow. When I talk about analytics, I get very low engagement, but those people are the ones, that’s where I’m getting my growth, that’s where I’m getting followers. Fascinating. A lot of social media strategists have known this for years, it’s counterintuitive, but the things that you do in social media that get lots of visible engagement may not be driving the actual business outcomes you were hoping for, they may just be like big conversations, provocative topics.

But the things that get very low engagement look quiet, you hear crickets, like what happened, no one’s engaging. That one actually might be the one where people see it, not much to talk about, but they say, I want more of that, I want this guy in my streams, I’m going to follow this person. So that’s true in other channels as well. Don’t get too hooked on these visible metrics. Social media platforms have gamified engagement, and they’re trying to get you to pay attention to certain metrics by making them really big at the top. It’s called the availability heuristic. Marketers get really enamored by these top-line metrics, that’s not what we’re supposed to be doing here, people, be a bottom-line marketer. Okay, I’m not going to digress anymore before I run out of time.

Okay, other interesting ideas. What three-word trigrams appear in top performing posts? Junior marketers may not think about three-word trigrams, a black belt social media chessmaster level genius might think a lot about trigrams. But actually, in this case, I didn’t get that much insight from this. I know why that one’s there, it’s because I was talking about that, explain, diagrams, interesting, look, diagrams, visuals. So sometimes you get more insight, sometimes you don’t.

Let’s try some other prompts. How does word count correlate with performance numbers? How do emojis correlate with performance? Not best practices on someone else’s blog, for your blog, your audience, your content, your social metrics, I exported this from my profile. Perform a semantic distance analysis to find subtopics that are not in my data set but would work well. Semantic distance analysis, another combination of words for a prompt that maybe a junior marketer wouldn’t think of, but quite compelling. You can find that there’s a hole in your content, because it’s measuring the distance between the relevance of these related topics.

Using topics, tone, and voice, create a LinkedIn post that would drive high engagement, let’s see, you make a draft, AI, I’ll probably rewrite it all, but let’s see what you got. Summarize the content strategy. Oh wow, if you copied and pasted in 40 different LinkedIn posts, don’t leave that conversation with AI until you ask it to summarize and give you a style guide, or give you a content strategy based on what you gave it.

My buddy Tony, I wrote about this, I think I made a video about this, we could share it. My buddy Tony did this, and then wrote me this email, and I’ll read it to you. I tried Andy’s method of using AI, or ask him for this quote, that’s why I put my name, I asked. I tried the method of using AI to analyze my LinkedIn posts and adjusted my content based on what I learned, then I did it again six months later, impressions went up 13x. I think Tony was doing some other good things in LinkedIn at the time, I saw him do this, but his engagement just exploded, and he attributes it to this approach, using the data to see what worked and then tuning up the content strategy based on performance. These topics lead to this outcome, I want that outcome, I’m going to use those topics. He says, crazy what some data driven insights can help you accomplish, thanks Tony, love the feedback. I’m doing this myself and results have been quite good. The trick is to get your data structured, it’s hard.

Okay, number five, let AI make a style guide. This one’s short because we just did it, basically. Once you gave it all the text, just one more prompt: create a prompt that can work as a style guide and can be given to ChatGPT so AI can effectively write future posts in the same voice, tone, and style. And what it gives back, there’s your style guide, thanks AI, copy, paste, save as PDF, upload to your shared prompt library, and now anybody on your team can go grab that anytime.

This is my shared prompt library, just a Google Drive folder, a bunch of documents inside, and inside each document there’s tons of prompts that I’ve hammered on for months, I know that they work. There’s a different folder for prompts that are kind of in beta, I’m not very confident in them, there’s some that don’t work at all, like that one I mentioned, that’s an alpha prompt right now. But anyway, once they get promoted to this level, anyone can use them anytime, and I welcome my team to contribute and collaborate here. This is a great place to upload that PDF persona, that PDF style guide, a data export, things that you want to be able to give the AI easily, very useful. That’s scaling up, that’s AI on teams without a fancy tool.

Number six, moving right along, number six, let AI discover what to promote. Okay, here’s my thesis for you. You have articles that have 10x engagement far down the funnel, people who read that one are so inspired and you built so much trust with them that they will actually give you an email address, and that takes a lot of trust. People get way too much email, if they gave you their email address after reading that one, they’re very likely to want more of that, that’s the post that gets the best results for you. But guess what, there are no metrics in GA4 showing the conversion rate from visitor to subscriber per blog post. Where do you get that?

The visitors who subscribed, here’s an article, it’s got a signup call to action, it’s a sticky footer, you can get to the email signup thank you page from any URL, from any article. Now, which of these articles drives the greatest percentage of people to that page? What is the calculation to find the conversion rate from visitor to subscriber per article? This is it: you have to take the key event, the number of visits to the thank you page when that article was the page referrer, and divide that number by the total number of visits to the article, that’s the conversion rate of that article from visitor to subscriber. Make sense? How many times did that article send visitors to the thank you page, divided by how many times that article was seen, that’s a conversion rate from visitor to subscriber.

I’m not going to do the math in analytics, but if analytics did have it, it’d be called the key event rate, it’s a calculated metric, GA4 doesn’t make it for you. A Data Studio or Looker Studio Pro could do it for you, I’m terrible at that. Excel could do it for you, yeah, but I’m bad at that too, so I’m going to use AI. Here’s where you get the data. I’m on the pages report, choose the long date range, add a filter so that you’re only looking at blog posts. Ideally, every article on your website is in a common folder called blog or resources, there’s a lot of reasons to do this, easy analysis is one of the good ones.

If some programmer decided one day that all of your blog posts should be at the root instead of in a folder, they made a mistake, that’s not good for this kind of approach, you’ve got to do some fancy programming to try to get those out of there. Some people, this is like an SEO superstition, that things closer to the root have greater share and more of the domain authority of the overall, never seen it, doesn’t matter, don’t worry. So yes, I’m going to filter just to see blog posts, now I’m just looking at blog posts, there they all are, click apply, and then export that. I’ve got half the data, traffic to blog posts.

This is also not great structured data. When you export anything from GA4, you need to delete the first nine rows, they are simply comments, and anything else, like you see the thank you page in there, the blog main page, get that stuff out of there. You also don’t need any of that data, all I want is sessions, you saw my calculation, I just need traffic to these blog posts. Looks good, half the battle, that’s what I needed.

Now I need traffic to the thank you page when these were the page referrer, a different report. I’m back in here, so now I’m getting the page path, same date, it’s got to be the exact same date range. I’m going to add a secondary dimension, you have to click the blue plus and then search for and find page referrer and choose that as the secondary dimension. Now, the filters for this: add filter, I only want to see the newsletter thank you page, so page path exactly matches the newsletter signup thank you page. That’s the first column, also known as the primary dimension, only has the email signup thank you page. And then there’s another condition, it’s a two-part filter, the second condition, I only want to see when the page referrers are blog posts, the page referrer contains blog, that’s it.

I hope this makes sense, that’s actually pretty simple, show me every time this article sends a visitor to the thank you page. Apply, export again, remove the top nine rows and any other noise, and also you don’t need those other columns, you don’t even need the first column, you don’t need the thank you page column, this is all you’re looking for. Also, there’s a bunch of ultra low volume stuff, like zeros or ones, that’s not good for analysis, it’s false positives, oh, there was one reader, one visit to that blog post, and that reader subscribed, so that blog post has a 100% conversion rate, not useful, so remove all the things with ultra low data.

I also now can give this to AI, first I actually change the headers so that they match the other report, upload both files, and here’s what it sounds like. I’m giving you two GA4 reports, one shows traffic to blog posts, the other shows conversions from those same blog posts, merge those as you would using a VLOOKUP, provide a link to download. Anytime you ask AI to edit a spreadsheet, provide a link to download, because it might remove every row and there’s only one row left and it doesn’t look right, you have to always double check it, again, don’t trust AI. Okay, looks good, that’s just what I wanted.

I could actually do the conversion right here, but AI already has the data, so it’s easier to do a prompt than to do a calculation, but that’s it. Sessions and conversions, conversions divided by sessions equals the conversion rate. Create a bar chart showing the conversion rates for the top 20 blog posts. That chart is very, very valuable to the content marketer, because these are the articles that, regardless of the traffic source, grew your list. And don’t be surprised when there’s a small number of them that have 10x results, if I took this out to the top 30, the 30th would be very, very low, some of these things might have 100x results.

The average, now you know what to put in social streams, now you know what to keep in heavy social rotation, now you know what to put on your homepage, now you know what to put on your email signature, now you know what to send again as a newsletter, now you know what to search optimize. Put on your traffic driving gloves, you just found the best mouse traps, now give them better cheese, I’m mixing my metaphors, gloves and cheese, but you get the idea. Very few marketers know the conversion rate from visitor to subscriber per blog post, ask the hardcore, most serious marketer you know if they can tell you this one, most cannot, but isn’t it valuable, don’t you want to know that?

You can promote anything, the social media marketers spent years building an audience, they can promote anything, but which do they promote, now you know. So this is not technically a social media marketing strategy, it’s a data analysis for conversion optimization strategy, once you know these things, you know what to go do. I love that one, do it, and you repeat it every few months, because it will change.

Okay, moving on, I’m going to let AI review my social ads. We have never bought an ad, I’m not that kind of marketer, not actually that strange if you do content marketing long enough and hard enough, but this is an 8 million dollar agency with 55 employees, never bought an ad, zero ads, never gated content, never used a popup window. There’s lots of examples of businesses like mine, you can grow a business without using those methods. But because I don’t know anything about ads, I’ve got a skills gap, that’s where AI helps. If you’re good at what you do, AI will make you a little bit better, if you’re not good, AI will make you a lot better, it’s a skills leveler, that’s what AI is doing here for us.

Okay, so I’m going to make an ad, which I’ve never done, I’ve never made an ad, let’s try it. Here’s an ad, it’s promoting a webinar, free webinar, B2B lead generation tips. Now, here’s a lazy prompt: suggest five ways this ad could be better. I don’t believe in lazy prompts, I believe in detailed prompts and multi-step methods. So what it gives me back in this case, I’m not very surprised, it’s terrible, it suggested changing the headline to “unlock your sales potential.” Of course, it didn’t know what to do, it doesn’t know my audience, I’m not targeting salespeople, I’m targeting marketers, so it’s off, that’s okay.

So what are we going to do next? We’re going to ignore most of this and give it a better prompt. My prompt this time includes my persona, I’m uploading the draft, the creative, the ad, the persona, and I’m going to give it a much more detailed prompt. So there’s three things happening here, the asset, the persona, and the prompt, and the prompt is like a mini set of best practices. I don’t know ads, but I can do some research and I can find the best practices. You’re a paid social strategist skilled at crafting ads that drive clicks and conversions, the best ads have a certain approach to messaging, a certain approach to the headlines, a certain approach to the visuals, a certain approach to the CTA, they trigger urgency and exclusivity.

Yeah, I spent 15, 20 minutes researching best practices for paid social ads, partly I want to learn, partly I want the best results I can get. I’m not trying to save time, I’m not giving you any shortcuts, the risk of me not doing this well is I waste tons of money, you’re buying traffic using an ad that would have low engagement. So I will always spend a little more time, and by the way, one way that you can research best practices is by talking to the AI as well.

So here’s the actual prompt: rate the extent to which the ad does or does not align with best practices, recommend changes. That’s a common approach that I use, rate the extent to which it does or does not blank, that’s me asking it for perspective and keeping it broad. It comes back, my original copy said “struggling to drive B2B leads,” its suggestion, the audience you told me about cares about conversion rates, why don’t you talk about conversion rates, that’s a great tip, I like that advice. Actionable insights to boost your lead conversion rates, that actually is smart, it’s more specific, and as we probably know by now, specificity correlates with conversion. Drive leads, or boost lead conversion rates, demonstrate ROI. My persona, I discovered through persona research that my audience as individuals want to specifically show their performance to their teams, so demonstrating ROI is actually a very smart thing to put as a CTA. B2B lead gen tips, it says, proven strategies to boost B2B conversions, that’s better, that’s just clearly better, that’s empirically better.

Okay, let’s fix it, I’m making a new ad, here’s a second ad. Create a color-coded heatmap matrix scoring them against best practices and alignment with the psychology of the persona. Okay, it likes the second ad better. Caution, I’m asking AI to check its own recommendations, of course it likes what it recommended, I did what it told me to do and it said I did what it told me to do, so you get the idea. But creating heatmap matrixes, if you’re going to a meeting, AI is great at not just doing the analysis, but creating some kind of visual asset for you to go have a conversation with your boss, or your manager, your executive, or your client.

Moving right along, 15 minutes, Brent, we’re going to end on time. Your LinkedIn profile, I’m going to move now into the personal branding side. AI can create headshots. Quick tip about headshots: that’s a bad headshot because you’re panned out too far, dude, you’ve got to show your face, it’s called Facebook because people want to see your face. If your headshot is panned out that far, that’s less engaging, there’s only 5% of this image that’s the face. Whoa, whoa, okay, back up, that’s too much face. So proper framing, we want to do that, and also don’t try to put your lifestyle and interest into your profile picture, that belongs in the background. I used to do a whole thing about social profile best practices, if you’re like this guy, he’s like in a witness protection program, what’s up with the silhouette.

There’s a tool, one of many of these, it costs something, I can’t remember how much, maybe 50 bucks, upload tons of images, you have to give it a lot of pictures, I think it wants like 50 high-res images to do it, but then it comes back with some pretty good likenesses of yourself. None of these are photos, there was no camera involved, it just gave it lots of training data. It’s also a lesson about AI, you always have to give it more than you would have expected, lots of training data comes back with lots of pictures. This is like analysis paralysis, what do they call it, marketers know this, the paradox of choice, give someone too many options, they get stuck. But scroll down, whoa, these all kind of look a lot like me, the lighting is a bit perfect, they’re actually sort of obviously AI, I think we can all understand AI now.

Change my eye color to blue, I don’t have blue eyes, I didn’t know what to choose, in the end I just chose this guy down here, the orange one. Why did I choose the one with the orange background? To me they’re all kind of the same, I chose the one with the orange background because that is a complimentary color, it’s the opposite on the color wheel, orange is a warm color, LinkedIn is a cool colored social network, all the interface elements are blue, you probably know this by now. Orange, red, and yellow are warm colors, purple, green, and blue are cool colors, so in the context of a cool color, a warm color will pop, so I’m standing out. That’s Kate, she stands out because she threw on a colored jacket, that’s a best practice.

If you want data for it, you can upload these photos to Photofeeler and then people will rate them and review them, and after you tell it, is this a dating profile picture, is this a business picture, of course I’m doing business here, it says, I paid to get 40 votes and it scored quite well, it was very high for competence and also not bad for influence, likable was, oh no, this one was more likable. Anyway, I’m not using my AI generated picture anymore, you can choose your own adventure.

Nine, getting there, nine, let AI draft a new headline in the About section, update your LinkedIn bio. Okay, you’re going to copy and paste the whole thing, which is tricky, because you have to expand, LinkedIn is filled with these sort of accordions, expandable content areas, you have to expand them and copy and paste the text out. The following is a LinkedIn profile, an About section for this person, suggest five alternate headlines, I put in the character limit, that probably doesn’t work, language models are terrible at counting, image models are terrible at spelling. We live in an era right now, this is 2024, where these models are very separate, the future of AI will be combined models, these will all get sort of closer together, you won’t be, these will all merge together a bit better someday.

Anyway, suggest five alternate headlines, write a new draft of an About section, structure it with a clear narrative showcasing achievements, quantifiable results, include a call to action, there again, my little best practices, these are not short, simple, lazy prompts, I’m going to tell it more specifically what I need, then I give it this giant copy and paste of all the text. Here are the headlines, I think these are pretty good, here’s the draft About section, I like that too, I think that this did a good job. Again, I don’t love AI for writing, but as a draft, as a point of view, as an editor, as a starting point, it’s interesting and it might be useful, but again, never trust AI. You’re the strategist, you’re the expert, you know your audience, it hasn’t had 500 conversations with your target audience, but you have, it hasn’t had your life experience and doesn’t know all the things about your background, but you do, so filter all these through your own point of view, please.

And then edit. Let AI audit your entire profile based on the point of view of your target audience. What if you gave AI the persona and your LinkedIn profile? What if you’re a frontline rep, what if you’re doing influencer marketing and you’re out in the world, or you’re a customer service person who people look up, or you’re an executive, or you’re a founder looking for funding, or whatever. Your LinkedIn profile is a web page, have you optimized it as a web page, have you audited it from the perspective of that most valuable visitor to that web page, who is considering working with you, collaborating with you.

Maybe you’re a job seeker, build a persona of the hiring manager and ask it, give it the job. If I were looking for a job, I would take the job description of my ideal job and the persona of my hiring manager and the detailed, full, every word of my LinkedIn profile, and I’d start talking to AI about where their gaps are. Is that time consuming? Yes, but I’m not trying to find efficiencies, I’m trying to find deficiencies and make my LinkedIn profile better, because when this person lands on it, I want them to see the promise and hope of me in that role, so they don’t have to use their imagination that I’m in there. You could even have it do a SWOT analysis, AI, here’s my LinkedIn profile, here’s the job description, do an analysis of the strengths, weaknesses, opportunities, and threats of me in that role, it’ll do it for you, and then you’re ready for the interview.

LinkedIn is the cornerstone of your personal brand, I can’t overstate the importance. So here I’m going to my final method, I’m giving it my LinkedIn profile, I’m giving it my persona, and I’m telling it, you’re a networking and personal branding expert skilled in evaluating LinkedIn pages for their ability to build trust and credibility. All my prompts start the same way, you have a role, you have skills, why, just context. I’m giving you a persona and a LinkedIn profile, list five ways in which the page could be more compelling to the persona, list five of the least important and persuasive elements on this page, similar to the “rate the extent to which it does or does not meet.”

LinkedIn profile, persona of my target audience, detailed prompt, and what it comes back with, my persona’s named Alex, you can see it here, it knows my audience, my buyer is looking for evidence, they want to reduce the perceived risk of working with me, I totally agree, AI, I should add more case studies, show data. My audience wants to know that we’re not a one-size-fits-all WordPress template theme websites, it’s customized, personalized for them and for their brand and for their target audience, highlight more projects and solutions that are specifically tailored to their unique challenges. My audience is not secure in their content strategy, they were attracted to us because they see we’re doing this at a certain level, my LinkedIn profile doesn’t really talk about that, describe your approach to content that leads to these high conversion rates, focus on areas that they’re struggling with, like messaging and personalization. Wow, I think those are all great tips, even if it didn’t write them for me, I can go back and revise.

Too much detail can overshadow critical achievements, AI, I totally disagree with you, I don’t think that’s true at all, I think that for a LinkedIn profile or any web page, my job is to make the most detailed version of that thing possible, I want to be exhaustively thorough, and that’s my philosophy about digital, so I disagree with that point, I’m not going to listen to it. How about this one, prioritize your endorsements so the most relevant ones, B2B web development, B2B marketing, data driven strategies, are at the top, great idea, I am embarrassed that I just made this slide, I haven’t updated my endorsements yet, but that should totally be changed.

You get the idea, now if you want to go ahead, the next prompt almost writes itself, but if you wanted it to give you a hand, rewrite sentences for the About section of this profile, see how I’m asking for specific sentences, and to better target the persona, show both the original and the recommended rewritten sentences. This is me working with it, I’m going back and forth, it’s a long conversation, and it’s like an editor, it’s like a perspective, it’s kind of interesting, sometimes I like it, sometimes I don’t, but I don’t want it to rewrite the whole thing again, that’s a, add value, add your strategic insights, don’t just have it do one big monolithic thing and control C, control V, that is going to taste like water, it’s not differentiated, it’s an amazing summarization tool, it read the internet, I joke AI might as well stand for average information, we don’t want the average information, we’re trying to be different here.

So I’m actually looking closely at its recommendations, I lead our award-winning team in delivering customized, data driven digital solutions that drive measurable results for B2B tech companies, okay, first of all, we don’t just work with B2B tech, secondly, I hate the word solutions, I think the only people who use the word solutions are salespeople, that’s not a good word for marketing, I’ve done hundreds of hours of keyword research as an SEO and I’ve never seen the word solutions appear in keyword research. So some of that is awesome, some of it is not, but I use my skills and expertise and my years and decades in digital to take what it gave me and go make it better, you get the idea, it goes on.

Brent, we’re going to share these, I’m going to stop, hopefully we’re all still on, I’m going to stop sharing now and come back for a quick conversation, and just say thank you, Prepathon and Cloudways, for letting me be part of this.

Brent Weaver: Andy, dude, virtual high five, round of applause. I feel like I’m always drinking from the fire hose on your talks, man, they’re value packed, there’s so much there in terms of specific strategies, prompting, ways for our amazing audience to leverage AI to more effectively promote their business, their agency, their client businesses, that was an absolutely amazing talk, so thank you for that.

Andy Crestodina: My pleasure. I think it’s fun, these were like use cases I had and I kind of finally put them together, and I hope what came through is the ethos about it, like invest more time in it, I think you get better results, that was hopefully the main point that came through.

Brent Weaver: Yeah, and I mean, I think even now that I’ve gotten a chance to get to know some of your material and just your perspective on how to approach AI, I think it’s really amazing advice for people to go in and learn how to do detailed multi-step prompts, how to seed AI with your avatar, with your own profile, with your own content, and then be able to pull insights out of that that then informs your content, bringing in that specific perspective, beliefs, points of view, really being able to make your content stand out with AI, versus what I think a lot of entry-level people to AI go, oh, I’m just going to have AI write my content, which, so are your users, they’re not even going to your website if they feel like it’s written by AI, which I think is such a brilliant point.

Andy Crestodina: Yeah, there’s a famous quote, or I don’t know if it’s famous, it should be, Jay Baer once said, the problem is not that AI can do what marketers do, the problem is that marketers keep doing what can be done by an AI. If an AI can make that thing for you in 10 seconds, that thing should not be made anymore, because your audience can have it made for themselves in 10 seconds.

So the first thing we did is the one I maybe love most, consider using AI to help you research provocative topics, it’ll give you tons you don’t care about or don’t agree with, but the one or two where you can take a little stand, victory favors the bold, if it triggers one of those ideas and you can go into that and make that kind of part of your little platform for your brand, that is really what people remember. And I know there’s no metric or data point or GA4 report for whether you are remembered, but that’s actually what we’re trying to do here, is to be top of mind in our category, only one or two companies can be top of mind in every category, and if there’s one outcome that you’d want more than any other, it’s that, even though there’s no shortcut or data for it, that would be the ultimate outcome.

Brent Weaver: And I love that. I imagine that some folks watching this session might have, I mean obviously they’ve gotten some very specific tactics that they can implement, but I know at least for myself, I’m like, man, I need to get more into this, I need to learn more about the type of stuff that Andy’s doing. You mentioned your kind of growth hack of finding somebody like Lisa Adams and saying, hey, let’s talk about AI on a monthly basis. For business owners, e-commerce businesses, agencies that are kind of just dipping their toe into AI, they’re just starting to leverage it for their business, any other suggestions that you’d provide them in terms of starting points, besides just opening up and starting using ChatGPT or another tool?

Andy Crestodina: A lot of people have said this, if you can document all of your most important SOPs, your standard operating procedures, and then go through them breaking down the tasks and asking if there’s an AI use case for that thing. A lot of what I’m doing in AI, and some of my friends are doing, is just doing it to help them do more comprehensively things they were already doing. So rather than think, how can I use AI, look at what you’re doing all the time and say, could it help with that, could it help with that, could it help with that, topic research, oh, data analysis, these are things like keyword research and on-page SEO, these are just things that I’ve been doing forever, so I don’t ever start by asking, what can I do with AI, I’m trying to drive outcomes through methods I’ve always used, and sometimes, not always, but sometimes, discovering ways that AI can help.

Brent Weaver: Very cool. Andy, I’m sure we’re going to have you come on board on future growth summits and Prepathons. For those that can’t wait that long till your next keynote with us, how can folks find out more about you or check out some of your content?

Andy Crestodina: orbitmedia.com, I write an article there once every two weeks. If you want to connect, you can find me on LinkedIn, that’s my best social network, and then also, I counted the other day, I’ve done 60 videos on YouTube, so at least maybe once or twice a month I’ll put something there, so if you go to YouTube and search for Orbit Media Studios, you should be able to find me there too.

Brent Weaver: Awesome. Well, Andy, thanks again for being with us today, and that concludes our keynote from Mr. Andy Crestodina. Stay tuned with us here at the Cloudways Prepathon 2024, today we’re going to have more content focused on AI and innovation, so stick with us and we’ll be right back here with Cloudways Prepathon 2024.

Andy Crestodina: Thanks, Brent. Thanks, Cloudways.