Brent Weaver: Here from Vassie and Jerry, what an amazing session panel discussion, but buckle up, we are going to be talking to Andy Crestodina next, we’re going to be talking about trust and quality in the new era of content discovery, how to unlock AI secrets, practical SEO tips you can’t miss. Andy is a co-founder and CMO at Orbit Media, he is a digital marketing expert with over two decades of experience in SEO, analytics, and website optimization, he is the author of Content Chemistry, he frequently speaks at top conferences and is deeply committed to community service and environmental causes. Andy, I’m pumped for this session, I love content, I love all the stuff you’re talking about today and how to create more trustworthy and quality content, so Andy, welcome to our Cloudways Web Growth Summit 2024.

Andy Crestodina: Thank you for having me, I’m excited to be here. Let me share a screen and just confirm that you can all see it. How does that look, do you see a little curve on the screen, inflated expectations?

Brent Weaver: We’re seeing it, that’s it.

Andy Crestodina: Oh, hold on, there’s an extra step, I had to click the hide button, okay. This is a popular topic, AI search, this is a fascinating time, we are in July 2024, what a time to be alive and to do marketing. I have been a digital marketer since 2001, I’ve used analytics just that long, I’ve done SEO since prior to the days when Google became the dominant search engine, and have been a content marketer since 2007.

AI is obviously as big as social, like mobile internet, a mega trend, and like all mega trends we can plot it on the Gartner hype cycle. So first we start with inflated expectations, then there’s a trough of disillusionment, some of us may already be there, then that slope of enlightenment, and finally hopefully we reach the plateau of productivity, that’s the Gartner hype cycle for new technology, every technology follows this curve.

But it doesn’t really get to the feeling of being on this roller coaster, so to put us on this roller coaster, I’ve got an idea, this was a literal experience I had with my five-year-old, she’s six now, I’d like you to meet Ada Crestodina. She’s on an actual roller coaster here and I’ve got my camera in front of her to show her experience, you’re going to see the face of a five-year-old girl on her first roller coaster, and she’s at that peak of inflated expectations right there, she has no idea what’s about to happen.

This is where a lot of us are in AI, and the headlines show it, this is what all of our faces look like when you see stuff like, AI is going to lead to a 14% average productivity gain, that’s massive, that’s big news, PricewaterhouseCoopers says it’s going to add $5 trillion to the global economy, that’s good news. And then we all start jumping in and we start to do stuff like this, I prompt ChatGPT to create an article summarizing the top 10 roller coasters in the US, and it does, wow, it just wrote the whole article for me, what a time saver, I’m far more productive.

But wait a minute, let’s follow it, if I can write an article with one sentence in 10 seconds, doesn’t that mean that my audience, you feel that, you see that face, that’s where we are. If I can write an article in 10 seconds, my audience can also get that same information in 10 seconds. Here I am, I’m going to Perplexity, popular AI tool, it summarizes the internet like AI tools all do, I type in top 10 roller coasters in the US and it gives me a list, I don’t need to click on a website, I don’t need to go to Google, I don’t need to use a browser, the tool just did it for me, that’s basically the same thing ChatGPT gave me.

Wow, what is going to happen now, that’s a massive disruption, this could change user behavior, the use of browsers, certainly the job of content marketers and the value of a website, an article, a URL. So Gartner actually predicts that there could be a 25% drop in the use of search engines, just let that sink in for a moment. Barry Schwartz put that on X, and SEOs agreed that seems likely, 38% of SEOs said, yep, there might be a 25% global drop in search volume, maybe that’s plausible.

And then you see headlines like, this is the end of the web as we know it, flood of low-quality content is going to break the web, the algorithms can’t keep up, New York Times, Ezra Klein, will AI break the internet, creators will have no incentive to keep publishing, and that’s what I mean, that’s Ada’s face, here we go, we’re at the trough of disillusionment. But then, if you watch her face and if you look at our industry in that moment, she actually is starting to feel safe, like she’s internalizing this, she’s thinking about it, this is where a lot of us are.

So I want to step back and take the big view, and give you practical ways, I’m not really a keynote guy, I like to transmit skills and show methods, and I’m going to give you plenty of those. But I’m going to start with the big picture first by explaining, if you’ve been worried about Google AI Overview killing click-through rates, guess what, Google has been killing click-through rates forever. You start searching for something, top 10 roller coasters in the US, and you’re going to see a superpowered featured snippet that lists everything for you by summarizing the web, and where are the links, this is death to website traffic, but it’s always been like this.

I have this weird habit, this hobby, of collecting screenshots of Google search results pages, I have for 10 years been collecting them, it’s a weird idea but it’s interesting because you can look back and see what happened to the web, how SERPs have changed. This was a SERP from 2016 for the phrase modern wedding dresses, and the number one position was like 240 pixels from the top of the viewport, plenty of organic list snippets above the fold, seven ads and 12 images.

Try that again today, wow, that search results page is like a bridal magazine, filled with ads and shopping listings and pictures, there’s only eight organic listings and 77 pictures. So anyone that’s been tracking this, I’ve been tracking my rankings forever, you can see the rankings don’t change that fast, SEOs are very concerned with rankings, maybe we should be more concerned with click-through rates, because that’s changing more dramatically. You could rank number one every day for 10 years but your click-through rates are dropping by 50% or more and traffic is down by 40%, that’s the story.

You rarely see someone map it together where you can see, with the dates lined up, your rankings, your click-through rates, that’s Moz, Search Console, and GA4 in one screenshot, that’s the reality, and this goes back years. You can see it in one report just in Search Console by turning on average position and click-through rates, steady rankings, falling traffic, so this isn’t new, this isn’t weird, we all have to be adapting all the time, this is something we should have been expecting.

I can’t show you this in GA4 because it doesn’t have enough historic data, but if I go to Universal Analytics and plot 15 years of organic traffic you can see the rise and fall, but look closer, it was called segments, now it’s called comparisons, separate out traffic for information-intent visitors who land on blog posts from traffic for commercial-intent visitors who land on service pages. The service page traffic is actually far less impacted, most of the rise and fall of this organic click-through rate stuff is to articles, people that just want information, they were never going to become a lead anyway, that’s the reality of search.

You do it yourself in GA4 and check the conversion rates, comparisons, people who started on a landing page that included the word blog started on an article, the landing page that did not include the word blog started on a service page, 10x, sometimes 20x. I have access to hundreds of analytics accounts and I see this all the time, so not only is the most valuable traffic far less impacted, but I think AI can help you improve your click-through rates, does that sound weird?

I’m going to do some strange things, you’re going to see me use ChatGPT in a way that’s probably not typical, I’m going to export data to find ideas for having more compelling, higher click-through rate title tags. Yes, ChatGPT can help me with this, but I don’t write lazy prompts, I don’t do single prompts, I don’t expect value from typing one sentence, my best adventures into ChatGPT are from uploading data or giving it detailed information about my audience.

Here’s what the report looks like in Google Search Console if you want to export data showing the click-through rates of your articles, I filtered out my brand name, I’m including only blog posts, I click the export button and it gives me a CSV file, a bunch of this is noise, I’m going to delete all the things that have no clicks. But this doesn’t have my title tags, so I need to get another report out of GA4, these are my title tags, just go to the pages and screens report and change the primary dimension to page title, then add a secondary dimension, click the blue plus to see all their URLs, because I need the URLs so that ChatGPT can map together data from one report and another.

Anytime you export data from Google Analytics you have to delete the first nine rows, it just confuses the AI, it can do it for you but it’s cleaner to delete those first nine rows, I do this all the time. Then upload this data to ChatGPT, for this you need the paid account, the $20 per month account, I think it’s worth it. Claude, Anthropic’s AI model, will let you upload files without that, whatever, use whatever model you want, I’m not advocating for ChatGPT, but upload the data.

I ask it first, the Google Search Console data, can you analyze it, cool, I upload the GA4 data, can you analyze it, and by the way remove all the non-English language ones, I’m going with English only, use whichever language, we’re a global community here, merge the data with the other data set. It puts the data together, provide link to download, you have to tell it that, because you never know, it might have messed up, I don’t really trust AI, in fact I explicitly do not trust AI, I check it every time.

So when I click to download the link I can look at the data, and this is exactly what I wanted, this is all you need to have ChatGPT or any AI model recommend higher click-through rate title tags, because it’s got the title tags, the URLs, the search position, and the click-through rates. A data analyst could do this, it would take them a day and a half, but ChatGPT can do it for us in minutes.

So now, here’s my prompt, you’re an SEO expert skilled in analyzing search performance and the psychology of click-through rates, which of these pages has lower than expected click-through rates based on its search position, it’s all benchmarked against my own data. Then it comes back and tells you which of these has lower than expected click-through rates, and it’s going to start recommending better ones immediately, these are better.

Look at the ones I wrote compared to the suggestions, “14 unique ways to say thank you to your audience,” it’s got a better idea, “14 unique ways to say thank you and build your network,” look at that benefit statement. How about this one, “there are two types of marketers, the turtle and the hare,” see what its suggestion is, “turtles versus hares, which type of digital marketer are you?” that’s definitely better. It identified which ones were the best opportunities and, without even a subsequent prompt, went ahead and suggested better title tags, let’s all go ahead and batch upload brand new title tags to all our articles that have lower than expected click-through rates.

Moving on, I’ve got another concern, this is the roller coaster we’re all on with my daughter, number two, another concern, AI-generated content will be penalized. I think that’s a weird concern, I don’t really have it, there is no policy at Google, but I’d rather have us worry less about detectability, because after all AI content is not detectable, there’s scholarly articles about this, some of the big thought leaders like Ethan Mollick will tell us AI writing is obvious only because it was low-effort, prompting multiple times removes detectability, all the detectors have high false positive rates.

So I’d say forget about detectability for a minute, you’re looking at this wrong, instead obsess over quality, that’s what Google wants, that’s what your visitors want, that’s the game, that’s how you stay top of mind, that’s how you outperform your competition, that’s how you remain a brand that has exceeded the audience’s expectations about the utility you can provide. This slide summarizes my experience, it’s the 10x efforts that yield the 100x results, performance is always on an exponential scale, I will do anything possible to move even a little bit further over on this curve.

Time is linear, I spend an average of 10 to 14 hours on every article I write, does that sound high, not really, I do that for the sake of efficiency, because I want to create more of these super high performers because they get 100x results, I spend 12 hours on an article because I don’t have time. So let’s focus on quality, because it’s the high-quality things in any channel that get the 100x results, this is the job, Larry Kim said years ago everything we do is either a unicorn or a donkey, the job of analytics is to help us find the unicorn so we can go make baby unicorns.

I’m going to give you an example based on a random category, for the next few minutes we’re going to be a space launch service company targeting commercial satellite operators. Rather than the lazy prompting, I hate this, please let’s not do this anymore, write me an article about satellite launch best practices, if ChatGPT could write a good article like that, that’s exactly what you shouldn’t write, because your audience can just prompt to get that same response.

It also tastes like water, it’s boring, undifferentiated, I joke that AI might as well stand for average information, because that’s what it is, it read 85% of the internet and summarizes it, that’s what you get if you’re a lazy prompter. So instead I’m going to recommend that we first teach it about our audience and then do some brainstorming together, this is my persona prompt, please feel free to steal it, share it, improve it.

If I first teach the AI about who I’m targeting, all subsequent responses will be more on point. Build me a persona of a commercial satellite operator, that’s a job title, of a global telecom, that’s the industry, they’re responsible for planning launch missions, the challenges, they need help deploying satellites, they’re looking for new launch service providers, I want four things back, their hopes and dreams, their fears and concerns, their emotional triggers, and their decision criteria for selecting a company like mine.

If you don’t know the decision criteria for selecting a company like yours, you’re really not well prepared to even make a web page, your job as a digital marketer is to know a lot about why your audience makes choices, what’s important to them, what they’re worried about, what they’re hoping for, what the true story is. There’s a true story in the life of every visitor to every web page, articles, sales pages, all pages, you have to know that, that’s why we need to train the AI.

It makes us a persona, it’s probably pretty good, it might be partly wrong, it’s probably named Alex or Jamie, I don’t know why, I’ve done this 100 times, but don’t trust the AI, another joke, AI might as well stand for assume incorrect, always assume it’s incorrect, look for chances to improve it, don’t trust the AI, trust yourself. So with minutes of research I realized, hey, the AI cares about insurance and risk management, why wasn’t that in the persona, add that, add the following decision criteria for selecting a launch services company, I tell it to improve my persona, polish it up.

I’ll spend 10 minutes on this, or longer, or maybe I’ve got our battle-tested, sales-proven ICPs, ideal client profiles, just upload those if you’re not scared. Either way, now it’s going to know your audience, and now you’re ready to do some brainstorming, which according to my annual blogger survey is the number one use case for AI marketers, generating ideas, nothing wrong with that.

But before I just ask it for articles, I want to talk to it, it’s called ChatGPT because you have a conversation, I’m going to have a conversation about my audience’s information needs. Your job is to know your audience’s information needs, what information does the commercial satellite operator persona need to do their job well, what does Alex want to know, how could I be a content strategist without knowing that. Scan down, see any that align well with your message, your value proposition, your past content, “strategies for backup and recovery in case of failed launches,” when my audience reads that they’re going to know I understand the risk, they’re going to say, wow, this guy thinks like me.

Next I’m going to ask it for ideas, suggest 10 articles on the following topic, make them capture the interest of the persona, provide practical utility, make them compelling and memorable, those are my words, choose your own, make this prompt whatever aligns with your content strategy. What does it come back with, gold, now it’s much better than the lazy prompt, “bridging communication gaps, ensuring seamless team coordination during launches,” I don’t think I could have gotten that output without this input, garbage in, garbage out.

I give it as much as I can, but I might not like that either, I’m scanning down a bunch before choosing one, and here’s my final little joke, AI might as well stand for another input, because you are the strategist, you decide if it’s good for you or not. We don’t just take everything it gives us and assume it’s all good, keep using your brain, keep adding value, we’re not going to abdicate our job to a robot, that was the peak of inflated expectations, you’re going to crash if you try that.

So what did I do, started with persona, then talked about information needs, and then finally started talking about topics. Are you still worried about an AI penalty, you should not be, if Google were on this call I would show the exact same thing, what am I doing, data-driven empathy, because that’s my job, I’m trying to be a useful resource in the life of this commercial satellite operator, I hope Google themselves is doing this.

One more idea, this one’s a little less likely to be loved by Google, but it’s clever and I think it’s still perfectly ethical, I’m going to take GA4 and Google Search Console data and get it to make SEO recommendations, this is probably what most people would like to hear as advice for AI-powered SEO. Again I’m going to Google Search Console and export all the data for keyword performance, but this time for one URL, an article I wrote about how to make social media videos, I’ve made a filter to only show that one URL and export it.

This entire model is based on the fact that all URLs rank for phrases that don’t even appear on the page, so URLs have the opportunity to rank higher for those specific phrases by simply incorporating them in the content. If you’ve got a page that ranks for a phrase that isn’t even on the page, put that phrase in the copy somewhere and it should perform a bit better. I export this file, upload it, tell it I’m giving Google Search Console data showing the search performance, can you analyze, of course it can, and then clean it up, merge the rows with very similar queries into a single row with a representative keyphrase, combine the impressions and clicks, average the click-through rates and positions, provide a link to download, because I don’t trust you, AI, I’ve got to double-check your work.

I look at it, check it, looks great, these are all the phrases the URL ranks for, all the clicks, impressions, click-through rate, and position, not all of these even appear on the page. So here’s the next step, here’s the web page for the data you just analyzed, evaluate this content’s keyword usage and frequency based on the phrases in the data set, copy and paste in the article. It comes back to evaluate it, I’ve got to do some text pre-processing, sure, go ahead, some keyphrase extraction, whatever it takes, buddy, you got this, and then it starts making recommendations.

The prompt that leads to it says, focus on recommendations for including key phrases that are not on the page but are in the data set, and while you’re at it improve the flow and quality, I don’t want any keyword stuffing here. So it comes back, “social media streams are crowded, fast-flowing, and competitive,” I wrote that sentence, perfectly fine, but read the revision, “navigating the crowded and competitive social media video landscape can be daunting,” that’s just as good and it’s more keyword-focused.

Let’s try this one, “it wins for the algorithms and wins for attention,” what’s the suggested revision, “this is why learning how to create captivating social media video content is crucial,” look, it noticed that the key phrase was in the data set exported from Google Search Console but not in the copy. I think that’s a bulletproof on-page SEO recommendation engine, you turned ChatGPT into SEO software, but you first had to give it the data, no lazy prompting, and if you’re in the same conversation where you first taught it the persona, it’s even more likely to be on point. It’s like a salad bar, just pick the tips you like, leave the rest, I’m just showing you things I found interesting and useful, this is early days still, but I thought that was fun.

Next big concern, search itself is threatened, Gartner’s prediction, 25% drop in global search volume. Watch this, I’m going to capture my screen while I go to Google, browser plus search engine plus website, to get a recipe, guacamole, it’s a sauce for chips. Wow, a whole bunch of guacamole recipes, they all look the same, how about this one, okay, there’s a lot of share buttons and a favorite heart, some ads loading at the bottom, a video on top and more share buttons, and where’s the guacamole recipe, scroll, a popup window, a captcha, wait, what, all I wanted was a recipe.

Instead I got a video ad, a sticky banner, a favorite button, share buttons, a popup window, captcha text, this is not good, this is a nightmare, let’s all admit in this moment, July 2024, that the mobile internet experience is bad, very bad, the internet’s a mess. I did this again on a desktop and looked to see what I had downloaded, 17 CSS files, 72 JavaScript files, 48 tracking tags, 43 images, all I wanted was a recipe and instead I got two megabytes of data transferred.

Try it again using ChatGPT’s app, open the app, type in how do you make guacamole, next thing that happens you get a guacamole recipe and that’s it, it just gave you exactly what you wanted. Probably every day thousands of people are realizing there’s a better way to use the internet, that’s all you wanted, that’s all you got, 128 words. So maybe we’re in an era now when the guacamole recipe is just basic human knowledge and nobody, Google or that website, should be able to monetize that visitor, I’m fine with that, because two megabytes versus this, 12 ads and offers, six modal windows, 14 tracking tags, 2,000 words, really, all I wanted was a guac recipe.

These are tough times for the web, so what does that mean for content discovery, for the content marketer, for the use of browsers and Google itself, I don’t know, but I’m not going to overreact, I’m not going to panic, because guess what, AI can also be a source of traffic. Go ahead and create an exploration in GA4, here’s the segment I made, this won’t even capture all of them because it doesn’t get app traffic, but this is a regex that shows me traffic from AI to my website, wow, I’m getting traffic from AI, no search engine involved.

So it’s not unlikely that content discovery will change, but that doesn’t mean the value of creating awesome content, publishing original research, helpful, useful, timely, detailed pieces, goes away. So let’s answer this question, this is the most important marketing question of the year 2025 probably, we can start on it now, how do you get AI to mention your brand and your content, how do you get yourself into the training data, how do you teach ChatGPT-5 that you are a good resource on these topics.

I don’t know, it’s not SEO, what do you want to call it, I’ll leave it to you, you guys discuss amongst yourselves and decide what this new category will be called. But what you want to have happen is someone says, well, what’s the best web design agency in Chicago, Orbit Media Studios, that’s my company, I took the screenshot on a good day, that’s not a consistent outcome, but you get the idea.

Step back and first consider what was the training data for ChatGPT, GPT-4 we don’t know, but GPT-3 is published, it was trained on the Common Crawl, like 85% of the public internet, 250 billion pages, so we want to be in the Common Crawl, in many places. WebText2 was trained on web pages that were linked to from Reddit posts with two or more upvotes, that sounds weird but I think it’s clever, Books1 and Books2, this is a bunch of subtitles, they had to teach the AI grammar and vocabulary, and then Wikipedia, which was heavily weighted.

So the idea would be to appear in the training data with your name, brand, value prop, differentiation, in many places. Here’s my quick suggestions, and this is mostly speculation, but it’s time for us to ask how do you get the models to know you are legit and relevant. Deep content and original research, those are what the citations are in Perplexity and the AI Overview, YouTube videos appear in tons of AI Overviews and they’ve got transcripts that can be crawled, and the strategy for that is called content marketing, same strategy, dude, for a different channel.

How about list posts, appear in list posts, press releases appear everywhere and are definitely crawled and ingested by AI, podcasts, but only if they have transcripts, we don’t care if they’re popular. The authority isn’t the point here, a podcast with a detailed transcript, your elevator pitch, and a description of your business on a very unpopular new podcast might be better than a famous podcast without any transcription, of course I’m only talking about training the AI, I’m not talking about brand visibility and influencer relationships, there’s more to life than search and more to life than AI, rule number one is to connect with your audience.

Finally, directories, review websites, local SEO, aggregators, forums, discussion boards, all the things, be in all the places, in other words the biggest digital footprint will win, that’s how you’re going to train the AI. Okay, watching the time, we’re going to move on, the internet’s going to collapse, this is my fourth and final concern and that moment of despair, the disillusionment, we’re going to come back up.

Oh no, the end of the web and search on the computer as we know it, you read this article back in 2013, they’ve been predicting the end of the internet for a very long time, that literally is an article from 2013. The algorithms are going to be overwhelmed, social media and search want people to handle it, that is algorithm marketing, but what about all the other types of marketing, your audience still needs help, and there’s more to life than search.

Think about this, we need to diversify, SEO and social media marketing, there’s an algorithm in between you and your audience, big tech sits right between you and your target audience. But relationship marketing, email, referral partners, word-of-mouth marketing, influencer marketing, that poll question we had in the last session, how many channels do you use, if the answer is two and they’re both algorithms you might want to reconsider. Events, offline marketing, community building, webinars, these are direct, this is disintermediation, this is how you connect without big tech in between.

So let’s broaden our skills, if AI could make us more efficient, maybe we can add a new skill set, this is called the T-shaped marketer, these are a bunch of marketing channels and skills, the best marketers in digital have wide breadth of knowledge across many areas but then they go deep into a certain channel. If SEO, that’d be keyword research, technical SEO, performance tracking, E-E-A-T, and site architecture.

So allow me to suggest, if you’re an SEO considering expanding your skill set, focus maybe on conversion, my concern about the SEO community is that it’s all cheese and no mousetrap, if you could get better conversion rates you’ll have better conversion rates from any traffic source, not just search, this works for everybody, 100% relevant to all brands, startups, Fortune 1000s, B2C, B2B, high consideration, low consideration. So yeah, I’m an SEO and have been forever, I’m also a conversion optimizer, my hope is to be a dual-threat marketer and focus on improving that number, remember from the first slide, the session conversion rate for that high-stakes visitor on the key service page.

Here again maybe AI can help us with conversion optimization, example, a service page targeting commercial satellite operators, a bad page, weak, thin content, 250 words, no supportive evidence, ranks low, the video is weird. So I’m going to give this web page to AI and upload the full-page screenshot, that’s my favorite way to have AI analyze a URL, I just wrote about this on LinkedIn, that way it can evaluate the visual hierarchy and see awards, trust seals, badges, certifications.

I upload that with this detailed prompt, a lot of my prompts look like this, here’s a page, you’re a conversion optimization expert, you’re good at evaluating pages against their ability to both inform and persuade, that’s our job, and you know that the best pages have headers that clearly indicate the topic, that the copy answers visitors’ questions, that the order of the messages aligns with their priorities, and that the page uses lots of supportive evidence like testimonials and statistics, this is a conversion optimization audit prompt.

It comes back and evaluates the page, oops, the page does not address concerns such as reliability, ChatGPT just found a gap, keep going, the page does not prioritize their information needs like cost-effectiveness and tech capabilities, gap, the page does not add any testimonials, partner logos, case studies, or statistical evidence, gap, the subheads are very vague, it says “our technology,” it probably should say “proven launch technology based on decades of success,” far more credible.

What did I just do, I did that all in the persona prompt, but I do not use AI to be more efficient, I don’t care about efficiency, that’s not the point of marketing, the point of marketing is to make something 10x better than your competitors and beat your visitors’ expectations, that’s how to win and stay top of mind. Only a few brands in every category can be top of mind, this is how to do it, exceed their expectations, blow it out of the water on quality, the visitor lands like, what, this page is reading my mind, I’m not looking for efficiencies, I use AI to find deficiencies and then go make it a better page, AI-powered, persona-based gap analysis, very powerful.

So I’m going to suggest one last thing as we climb up the curve to that point of productivity, this is my daughter at the end of the ride, look, we made it, good job, I can do this. I’m going to leave you with one final tip which should be comforting to humans, it starts with this quote from Jay Baer, he says the problem is not that AI is doing what marketers can do, the problem is that marketers keep doing what can be done by an AI, just let that sink in, if an AI can do it you should stop doing it and let the AI do it, and instead do something that can’t be done by an AI.

Most of the things you say in day-to-day conversation, AI could never write, that’s absolutely true. Here’s an article I wrote, GA4 versus Universal Analytics, this is the other disruption from last year, a side-by-side comparison plus input from 10 angry marketers, that timing is perfect based on my knowledge of my audience, because I published this right before the GA4 cutoff date in July, so I know them better than the AI, I talk to my audience every day, AI does not.

Then scroll down, we get to this sentence, GA4 is better than Universal Analytics, that’s my point of view, I actually believe that’s true, I know I’m in the minority, but that’s a sentence AI could never write, because it has no point of view, no opinions, no perspective, no life experience. Coffee, if you put milk in your coffee you wrecked your coffee, that’s bad coffee, coffee is supposed to be dark and smoky, that’s a trivial example, but it’s another sentence AI could never write, in the end it doesn’t care about anything, it doesn’t understand what it’s saying.

So thought leadership content cannot be done by AI, only by humans, that personal video with me making my case in front of a camera, AI can’t do that, it has no face, that visual metaphor comparing GA4 to Universal, it can’t do that, that’s my own creative input. I reached out to my angriest friends, Mike King, Justin Rondo, Court, Alex, Zanti, and Amy, because AI has no friends, it cannot do good influencer marketing, I built relationships over years that it did not.

So do what the machines cannot do, you are still a valued human, this is not going to come close to hurting your job if you just move to things where the AI made you more efficient or improved quality, and now you’re focused on thought leadership, taking a stand, drawing a line in the sand, and getting down to the beliefs and values of your brand, AI has no opinions, AI has no friends, and AI has no face, ChatGPT can’t throw a punch.

One last tip, this can help you find provocative topics your audience might love, listen to these prompts, they’re very special. What are some relatively mundane, almost trivial space industry topics that professionals have very strong opinions about, I’m not trying to be controversial, I’m trying to trigger a conversation, and it comes back, units of measure, metric versus imperial, everyone’s got a point of view, naming conventions for spacecraft and celestial bodies, rocket design philosophies, these are all things my commercial satellite operator might care a lot about, and AI helped me find those.

Maybe 80% of them are boring or obvious, fine, skip those, but a few might be really interesting for you to come out and have a point of view, which AI has no point of view, you say, yeah, it’s ridiculous that the United States hasn’t converted to metric, it’s embarrassing for Americans, AI can’t write that sentence. Try these prompts, what questions are people in my industry afraid to answer, what false things do people in my industry think are true, what are the most common assertions people in my industry make but are least likely to be supported by evidence, go publish that original research, there’s your missing statistic.

What are the most important topics least likely to be covered by the popular blogs, do gap analysis on my whole category and entire industry, this is everything my industry’s ever published, there’s a hole, AI finds the hole, you plug it. Or better yet, what counter-narrative opinions are least likely to be discussed by thought leaders, counter-narratives, these are powerful words to add into prompts.

Okay, we made it, we’re at the top, look, she’s ready to go again, look how happy she is, there’s my wife, she’s happy, my son in the front seat, that’s it, you got it, we can do it, it’s all going to be fine, channels will change, quality will not, the brand that best understands their audience, solves for their information needs, builds trust, answers questions, and grows awareness and top of mind is the brand that will win, that will be forever true. Okay Brent, I’m going to tab out, we left just enough time for a question or two, I promised to leave 10 minutes.

Brent Weaver: Perfect timing, Andy, dude, I’m blown away, I just want to say, before we get into questions, thank you for this talk, I know for myself I’m going to ask my whole team to watch this. Let’s give Andy a little virtual round of applause for an amazing keynote packed with so much great content. I’ve heard that phrase, don’t be scared that AI is going to take your job, be scared that somebody using AI is going to take your job, and I’m thinking right now, I’m like, damn, Andy’s going to take my job, he’s going to take all of our jobs, because this was phenomenal. Leanne said this is the best SEO talk I’ve seen in a very long time, absolutely brilliant, best of the day, lots of applause here. So if you have a question for Andy, let’s get that in the Q&A panel, I’m going to put this one up first, there are new threats out there like deepfakes, if LLMs get information from fake sources how reliable will it be, at what point does the AI start consuming the AI’s false information, any thoughts?

Andy Crestodina: Yeah, I separate all conversations I have about AI into two groups, there’s the between nine and five, I want to do my job better than I did yesterday, and then the political implications, accuracy, bias, labor market impact, these are things I don’t have a practical perspective on but I have lots of concerns about. That question is about AI safety, and the engineers need to work very hard on that. People are just really bad at checking sources, so my hope is that as marketers we can be the source, be the source of truth, be a better answer, so that when someone’s looking up a statistic or a framework or best practices they go read us.

But when they don’t do that, they’re accepting something without validation, and I’m scared, because we should not trust AI, and when that thing they’re accepting has political implications, or it’s plagiarized, or inaccurate, or a bad actor trying to manipulate us, there will be bad implications from this. But if it helps put it in perspective, social media has bad implications and we keep going, we teach our kids, let’s all be diligent about checking sources and teach our children to check sources, Ada will get this lesson when she’s ready.

Brent Weaver: This is a question from chat, and I have this same question, how can we get some of your prompts, do you have a resource, a blog post, on how to become a better prompter and leverage data in your prompts like you’ve done so eloquently in your talk?

Andy Crestodina: Yeah, I’ve published them all, as I get to them I publish them all, I’m going to put it in the chat, this is every article I’ve written about AI, every one of those articles has all the prompts, there’s probably 20 that we didn’t touch on today, you can copy and paste them out of there. I don’t see any value in trying to hold back, some of these have such high value you feel like you just wrote software, it takes hours but you iterate a prompt that really works well. I’ve had moments where I’m like, oh, this one’s really effective, should I share it, but that’s anti-marketing sentiment, you should share all your best, we’re past the era of secrets as Mark Schaefer would say, so my philosophy is just to try to be the most helpful human on the internet, so all the prompts are in all those posts.

Brent Weaver: Very cool, I love that. I’m not 100% sure what the last part of this question is, but when I search, Google Gemini comes up on top, why would Google put Gemini ahead of their search query, any thoughts on their strategy?

Andy Crestodina: Oh, right, I wondered that too, at first I was like, whoa, where are the ads, what’s going to happen here. I think they’ve acknowledged that accessing that AI summarization of the internet is a better user experience than browser plus search engine plus a noisy mess of web pages, so they’re just trying to stay relevant, there is actually some risk of them losing market share for the first time since I’ve been using Google, this is a brand new thing.

It’s hard for me to say what the strategy is, in the end I think Google has a very strong play to become a dominant AI platform, because information is how AI gets trained and they have vast amounts of information, all the computing power, all the pieces in place. The problem is it’s a bit in contrast with their business model, because part of the reason why AI in Perplexity and ChatGPT are such great sources of information is because it’s clean and fast and private, and the mobile browser and the experience on websites after you click is not great, it’s a mess, so they’re going to figure it out, I don’t know if in the end they’re just a slightly less valuable company.

Brent Weaver: I feel like with Google though, you’ve got the whole Google Workspace, they’ve got everybody’s email, business documents, they probably have one of the largest collections of data on planet Earth.

Andy Crestodina: Yeah, Maps, YouTube, they’re so well positioned, so dominant, try to not use Google for half a day, they’re the most popular operating system, browser, search tool, analytics tool, map tool, video hosting and streaming platform, leaders in productivity software, their search is where their revenue center is, but I think there’s a lot of strategies they could play.

Brent Weaver: Awesome, Andy, one quick last question, how has your experience been participating in this year’s Web Growth Summit at Cloudways?

Andy Crestodina: Maybe unique, but I’ve been a fan of Cloudways for a long time, 10 years ago I met a marketer who was there and didn’t know anything about the brand, and have gradually come to have huge respect for the way this program has worked out, clearly there’s tons of interaction. I would just say, let’s all connect with each other and keep in touch and keep sharing, it’s kind of sad when conferences end but hopefully this can be a new beginning, I’m honored, I would never miss the chance to be part of anything Cloudways does, it was a small thrill to be invited to be part of the Web Growth Summit, so thank you, Brent.

Brent Weaver: Awesome, well, Andy, based on our user reactions to your keynote today, I’m sure we’re going to be tapping you to come back to our future web events like our Prepathon or a podcast, so amazing talk today. I think we have your blog pinned in the comments, so for our audience out there, if you haven’t yet checked out Andy’s blog over at orbitmedia.com, go check that out. I love so many of the takeaways today of focusing on quality and not necessarily efficiency, which almost goes counter to the poll we heard earlier where 45% of people said time was their biggest constraint, the thing I heard from Andy today was, slow down, spend more time, be more thorough, bring your own unique assets and unique voice, have an opinion, have a position. So thank you, Andy, again, our audience will follow up with you and check out your stuff, thank you again for being here.

Andy Crestodina: Thanks, Cloudways, thanks, Brent, have a beautiful day everyone.

Brent Weaver: And this is going to wrap up day one of our Web Growth Summit, if you haven’t yet completed the survey for your participation, we’d love for you to do that so you can give us the feedback that helps us improve. Tomorrow is day two of the Web Growth Summit, we have another packed day of amazing content, speakers, discussions, activities, so definitely come back, we’re going to be focused on how to maximize traffic, we have sessions on website reliability, website UI/UX, and a live website performance review, which I always love seeing. Website speed is a huge topic still and has a huge impact on conversion and visitor retention, so come back to us for day two, make sure you complete our survey, and we are looking forward to kicking off tomorrow. Thanks again for being here, awesome hanging out with you all in chat today, amazing sessions, amazing speakers, and that is our program for day one, until next time, I’m Brent Weaver.