Welcome to the Marketing AI SparkCast, where AI meets modern marketing. Hosted by Aby Varma, a global leader in marketing, AI, and innovation, each episode brings you real conversations with leaders putting AI to work in strategy and execution. This episode is brought to you by Spark Novus and the Marketing AI Pulse community. Here's your host, Aby Varma. Welcome to our August episode of the Marketing AI Pulse monthly brief podcast. This is where we curate stories that may be relevant for CMOs and marketing leaders to keep a finger on the pulse of marketing and AI. This is an extension of our community by the same name. If you're not familiar with it, you can go to marketingaipulse.com and learn all about it. You can join for free and have access to a large community and growing one of that and some really interesting programming that we have planned for the remainder of the year. As always, I'm excited to welcome my partner in crime of this podcast, Matt Cyr. Hey, Matt. Hey, Aby. How are you today? I'm doing good, my friend. How are things going with you? Good. It's about 8,000 degrees here in Charleston, so I'm staying inside and doing my best to stay cool and not melt. So so far so good. Awesome. Yeah. I feel the same way. I don't know if it's the temperature, if it's all the AI news that is coming our way. Certainly been a lot to unpack. Right? Oh, yeah. Yep. So here's our first story. We are going to talk about Claude's watermark. It's created quite the ripple effect in the world of marketing. So let me break down on what exactly happened. So on June 10, the European Commission published a code of practice, a voluntary code of practice to help AI providers comply with, article 50 of the EU AI Act, which requires certain AI generated content to be identifiable in a machine readable format. By the end of July, about 200 orgs had already signed up. Anthropic, Google, Meta, Microsoft, and OpenAI were among the companies that supported its rules for marking, aka watermarking AI generated content. Early August, the article 50 transparency requirements became applicable, August 2 to be precise. AI systems introduced prior to that date have until December 2 for 2026 to comply with that marking requirements. The act does not mandate one universal watermarking technology, so each of the providers who signed up for it can use different marking and detection methods. On August 11, and this is what caused all the commotion where August 11, Anthropic announced in an outline how it would apply the requirements to Claude. Supported models launched on or after August 2 will include this marking right at launch, and Anthropic said it's working on adding that marking to the older models. So how does this actually work? So Claude's text based watermarking has three aspects to it. So one is the token selection. So when generating text, Claude will frequently choose between synonymous words that are contextually appropriate. For example, the weather today was cold, and it could be overcast versus gray. That's the example that they use. So with Claude's watermarking, the token that gets selected is specific, which is based on a, quote, unquote, secret key. So Claude selects these words using a secret mathematical key based on the preceding text, which outlines these patterns. So, essentially, it leaves a subtle statistical signature across long form text. It is imperceptible to human readers, to you and me, but anyone with Anthropic's detection key can run the text through a tool to calculate the status... Statistical probability that Claude generated it. Right? Things that everybody should be aware of. It's a human invisible. There's no hidden character, zero width spaces, anything like that. It's just a pattern of the token or the words based on that secret key. This watermarking does not use any token fees. It uses standard token generation process and that sort of thing. There are no delays. The watermarking indicates the model involvement only. It contains no user account or conversational metadata. So it's anonymous, and there are some known limitations. Light editing to text generated may preserve the watermark, but heavy editing and heavy rewriting, paraphrasing translation, or mixing it with other content may weaken it. Right? And obviously, this is all text from an image standpoint, Anthropic applies the open C2PA industry standard that we won't talk about in this podcast episode that it's its own beast. So for another day, but this is really a big deal from a marketing standpoint, and people are freaking out. So why does this really matter? So here are a couple of things to think about. So first, AI use becomes visible, right? It's in a lot of way, manifestation of Em dashes where everybody was up in arms, where you recognize an Em dash and people are like, oh my gosh. You use AI. Now you can genuinely tell that somebody... AI has been used. So a brand's content could carry a detectable signal that remains after publication based on the pattern. That's one. The second thing is detection can be misinterpreted. So watermark may indicate that AI contributed, but it cannot explain who wrote it, who edited it, whether the system wrote it, or the human edited it, or the translation, or whatever. So that sort of is unknown. As long as it just passes through the system, the watermark exists. It doesn't say the the sequence or the, I guess, the depth of involvement in a way. And, frankly, why this matters is I feel that for all of us when we are writing at this point, you know, human written needs really a new definition. I may have ideas, and I'll start from the idea in bolded format, put it in AI, get something back, change it, morph it, tweak it. My grammar is not the best. I'll use AI to fix it and all of these different things. So the meaningful distinction is no longer whether AI touched the content or not. And I think that's key. It's whether people remain accountable for its accuracy, quality, and final approval. So big shift in the way people and marketers in particular are thinking about it, and it caused quite the ruckus, quite the controversy in in marketing circles. But Matt, what's your take on all this stuff? Yeah, think you said the most important thing near the end there, Aby, when you talked about the role of the human, you know, still being so critical and central here. I'm someone who doesn't put a lot of stock in AI detection tools. I think more very often they show themselves to be inaccurate. We've seen a lot of that in higher education, for example, where AI detection tools have supposedly caught someone cheating and they in fact were not and vice versa. So to me, these AI detection tools in a lot of ways are erased to the bottom, and I'm not sure what value they really are adding except in unless there are critical facts or mis facts that are included. I mean, there's been stories recently of lawyers who have used AI in briefings that have been inaccurate. I think when you can find detection of AI usage and things like that, I think there's a lot of value. But in general, I would go back to what you said, which is just make sure that you've got your human marketers still very much in the loop. And I think, you know, I think most marketers will end up being just fine here. Yeah. I think from a marketer standpoint, like I said, it's really time to rethink what human written means, because I don't think there is anything as purely human written that exists. I know there are gonna be people out there who write 100% of their stuff, but the real question is whether human judgment remains in control. So from a marketing leader standpoint and a CMO standpoint, I would say three things to consider. So firstly, you have to establish and define disclosure rules within your organization for your brand. Right? Define when AI use must be disclosed across all marketing forums, advertising, social media, top leadership, and other brand content, work with the compliance and legal folks in order to establish those rules, and obviously train your ecosystem, employees, and agencies. As part of that disclosure, I think what becomes important is the nuance is you have to disclose and document how AI is used, like document AI's role. Right? So require employees, agencies, vendors to record whether AI wrote, edited, translated, or otherwise contributed to published content in any way. I think it would be important to highlight that. And given all of this stuff and the direction that we're heading, I think there's also room for sort of the false positives where watermarking alone as a proof of authorship, misconduct, or lack of compliance. And that area is frankly gray. So if it was my idea and I wrote a blog and ran it by Claude and Claude gave me grammatical suggestions or suggested words or what have you, and if you run it through an AI detection technology and it says, hey, this is 100% written by AI. That's not true. AI had a part to play in it. It was my idea. I wrote it. But that's where the detection alone as a signal is not... No longer going to be adequate measure. Right? Verify the content history, and you gotta preserve human in the loop, human review as part of it. So all I have to say that watermarking alone will not decide whether audiences are gonna trust your content. How transparently and responsibly your brand uses AI and discloses it will. And it's just a fascinating development, and we'll be watching this space pretty closely. So that was our first story about watermarking and how marketing leaders and CMOs should deal with that. Our next story is Matt about a CMO survey from Duke that has some really interesting insights into the world of CMOs and marketing. Unpack that for us, please. Yeah. You got it. It is interesting. There's a lot of these surveys out there looking at the role of AI on marketing, on marketers, on the use of technology, but Duke's Business School conducted a survey of several 100 CMOs earlier this year, and the results came out and there is a variety of information in there. But what I really keyed in on were the specifics around AI adoption. Share a few things. No surprise AI use in marketing has more than doubled in the last two years. You know, I think we're all of course seeing and experiencing this. Companies are expecting AI to support the majority of their marketing activities within three years. I would actually hazard a guess that it'll be faster than that. Marketers reported improvements in productivity, customer satisfaction, and overhead costs. But the key to me, and I think you and I probably go see this in our work every day, is that technology adoption is actually moving faster than organizational execution and capability. You you and I talk about this all the time, but it's not a technology challenge, it's a person challenge, it's a human challenge. And I think, you know, what we're seeing is that organizations that are investing time, energy, and money in tools are not also at the same time investing time, energy, and money in the people that need to use those tools. I think there is budget that's being given out, but I think not enough of that is going to either the the tool specific training or more generally an understanding of AI, and I see this all the time with my clients, know, not all ships are rising with the tide and so people are sometimes being left behind in their knowledge of and experience with AI. So one example is that the survey showed that training budgets declined to only 3.8% of marketing spending, and marketing headcount growth fell by more than 50% compared to the previous year. And so I guess no surprise here that technology is being, brought on board without a whole lot of value, equivalent value being brought back in terms of, how that AI is being used to benefit the organization. So what does all of this mean? I think it means that companies are expanding AI use without making the equivalent investment in its people. And I think this is a really critical thing as we think about budgeting for 2027, which is crazy to hard to be thinking about, but is to have marketers be considering ways that they can support their people, not just the technology. And the issue of course is broader than just employee skills. Yes, that's a critical piece of this, but there's challenges around integration, funding, employee capacity, talent, hiring. And of course we know that access to AI does not automatically produce effective implementation or measurable value. I see this every day in the work that you know, that I'm doing. And so I think there's a real opportunity here for marketing leaders to go from buying tools to helping their teams be able to really use AI outside of the specific use case for that specific tool. So are you seeing this across your clients as well, where you're seeing investments in in tools and tech, but not necessarily in people and process? It's not like they're not investing in people in training at all, but I think the recognition that it needs more TLC is certainly not at the level that it should be at. In fact, it's not uncommon for us where during the course of adoption, road maps, such, we start with talking about investing the right amount of time in strategy of change management and people and training and enablement. And those are often not overlooked, but almost like accelerated past. And I think it doesn't receive the right level of support. So I think to your point, the survey is revealing something that is definitely real. And the last point that you made that access to AI does not automatically produce effective implementations, a 100% true. I can't tell you how many times when we see that with clients where they'll implement a certain piece of technology, buy a whole bunch of licenses, but because they have not put the time into training and change management and work... Workflow and all those sort of things, they're paying for the licenses, but the people are not using it or they haven't truly kinda extracted value out of that. And I think that's a a pretty common mistake that CMOs and marketing leaders make early on that we try to warn them against that. But sometimes it's part of a larger internal company wide strategy. So I think of this survey points are very, very insightful and frankly spot on. Yeah. And I think you know what what I see is right people then throw their hands up and say, this AI stuff doesn't really work. Right? And we've we've bought all this stuff and and we're not seeing value from it. So yeah, totally agree. So at the end of the day here, why should CMOs care about this survey and AI adoption more broadly? CMOs just may not realize AI's full value if technology adoption is not supported by adequate people, time, training, and integration. You know, it just can't be said enough that, you know, if you're going to make these important steps and investments, you have to get your team ready to be successful. Increasing AI will raise expectations for measurable marketing performance. I think this is an important thing too, is that if you've got a board that's looking at your marketing activities and the investments that they're making in those activities and the technologies to support them, their expectations are going to rightly be set to where's the value coming back for all the investment that we're making here. And so I think again, if your team is trained effectively and understands the potential and the impact of AI, I think you're going to have a much better experience and a much better outcome and obviously much better ROI. Third is under investment in organizational capabilities could prevent marketing teams from converting AI access into sustained business value. Again, going out and buying a bunch of tools is not the same as using those tools in really effective ways. And so, you know, making sure that it's not just the optics of, hey, we're adopting AI, really putting the human power behind that. And finally, the thing that I would say is that CMOs need to match the technology investment that they're making with an investment in the capabilities required to use it. And that of course includes, you know, supporting the human beings that are using these technologies. So really important topic, really important survey, and I think gives us a lot to keep an eye on and hopefully we're gonna start seeing a shift now towards investing in people as much as if not more than the technologies themselves. So good stuff. Yeah, for sure, Matt, because I can't tell you how many times I've heard of customers where they've been given a login and a password to an enterprise system, Salesforce, or what have you with no training whatsoever with a link to a whole bunch of videos and expected marketing team members to kinda spin magic out of that, and that just ain't happening. Alright. Aby, on to your second topic of the day. You wanna tell us a little bit about advertising to AI agents. What have you got for us? Yeah. I think we were not done with advertising to human beings in its entirety, but here we go. Now we are advertising to AI agents. So let me just tell you what happened. So in June this year, Time Magazine began creating stripped down, markdown versions of its web page that were easier for AI systems to read. In July, Digiday reported that Time had begun turning those machine readable pages into advertising inventory. Working with ad technology company, Mobian, Time inserted sponsored FAQ style content from brands, including Ally Bank and the Project Management Institute. And the ads appeared in the version served to AI crawlers rather than the web pages that were shown to human visitors. And Mobian designed the format to provide AI systems with current brand approved information that could potentially influence how advertisers appear in future AI answers. The ads were labeled as sponsored content and included sources, but Perplexity did not like it. So early August, Perplexity confirmed that it was blocking Time's Markdown ads from influencing its agents and user facing search results. They characterized it as being deceptive and warned the publishers using such ads could receive a reputational downgrade, if you will, in a lower trust score in its search index. Kinda wasn't quite happy about that. And, apparently, they're working out something with Perplexity, but they have not really announced what's the final outcome as of, you know, August 21. But it's just an interesting development about how Time magazine is really coming up with ways to advertise to agents. Right? Like I said, we're still trying to master advertising to human beings, but advertising to agents, that's a whole other world. So Matt, how are you advertising to agents, my friend? Yeah, actually I've got this problem solved, Aby. I don't know about Time Magazine and Perplexity, but I actually have a whole room in my house filled with robots. And all day, every day, twenty four seven those robots are hammering away keyboards and trying to gain the LLMs and so so far so good. I'll let you know let you know how that goes, but so far we've seen a really big ROI on all those robots. I'm pretty sure Time Magazine did not have that in mind, but jokes apart, why does this matter? So, essentially, it's a good insight into what the future may hold, right, where this experiment shows that publishers and advertisers are now testing ways to reach AI systems that may influence what people see in AI generated answers. Right? So we all know that people are using AI to go, you know, research and scan hundreds and thousands of resources and distill that stuff down to bite size, understandable pieces of content, which is suitable for human cognition. And to that end, having agents then be exposed to those as opens up that potential new media audience. Then Time is serving AI crawlers and machine readable pages, and then also human visitors. So that raises the question of the divided Internet. Are we gonna have two versions of the Internet, one for agents and AI and one for human beings? And what Time did in that in that experiment, the ads were labeled as sponsored content, but Perplexity still characterized that format as deceptive because the promotional materials appeared in the crawler facing version and not in a human facing version. So it questions the rules for disclosure. And the way I think about it is that Perplexity sort of demonstrated that AI platforms can block some of that content in, you know, lower publishers' credibility, brand credibility if that content and the experiments of the techniques are not complying, in the eyes of these AI platforms. In a lot of ways, the AI platforms are becoming the gatekeepers. I don't know how I feel about that. And then finally, we all know that this is a very new format, a very new idea of advertising to agents where there is no commonly accepted industry standards, how they should be disclosed, measured, or treated by these AI platforms. So it's a very new space. And from a marketing leadership standpoint, CMOs, I think something to keep in mind that AI agents may be the next media audience that you may have to end up appealing to, But right now, nobody knows the rules, and nobody can really prove the ROI. So three things to keep in mind. So first is follow this area, wait for evidence. Do not treat advertising to AI agents as a proven channel just yet. Like I said, there has to be some independent evidence of visibility, influence, business results. You wanna be careful about that. You should demand transparency. So ask whether people in AI crawlers receive the same claims and how sponsorship is disclosed in which AI platforms accept to that format. And then finally, you gotta make sure that you're building that credibility first. So prioritize accurate product information, authoritative content, credibility, third party coverage before experimenting with paid AI influence to to agents. So it's a very sort of greenfield new area, and the rules are murky or nonexistent, but it really opens up and gives you a peek into the future where we all may be advertising to agents. And that just is a very interesting concept in a prospect as a marketer. So, yeah, if you're listening, if you're a performance marketer, paid media person, would love to hear your thoughts. And speaking of interesting stories, Matt, you had a fun one. So the $11 brand, the full ChatGPT, this is a very good kinda insight into what is possible in the world of AI today. Take it away, please. Yeah. You're right. This is a fun one that we're gonna end with here today, Aby. I think we all know that the world around us is changing every minute, but this story is a is a really great example. And it should probably frankly make every marketer just a little bit nervous. A moment ago, talked about the rules being unclear or not established. This is a really good example of that. So what are we talking about here? Last month, a researcher named Deana Burke decided to run a little experiment and see if she might be able to gain ChatGPT to help promote, a product that she was creating. So what did she do? She spent about $11 in about one hour building a new product called Morrowen, which is a natural deodorant brand that is, you know, custom created with unique ingredients and you know, targeted towards a very specific audience. The only thing here is that this Morrowen product does not actually exist. She created a three page website, some formulation copy with some important keywords in it. She created an AI image and she wrote a Substack essay about it. She did not include any reviews, didn't do any advertising. She just put this out there in the world. So a very specific product with very specific ingredients in a very specific market. Three weeks later, she came back and did an analysis to see how things were going. And guess what? ChatGPT now was recommending Morrowen first in four of four answers, when people were searching for, these kinds of, deodorants. In this case it's a niche audience of magnesium and baking soda irritation focused purchasers of deodorant. And so Morrowen automatically shot to the top of ChatGPT's recommendations. And what we learned here is that the AI shelf as you and I have covered before, that's feeding these recommendations is cheap and easy to gain right now, right? These agents are starting to shop for us as you talked about a few moments ago. So in this case, a completely made up brand with very, very thin credibility, no reviews, no advertising spend, has suddenly become the number one choice on ChatGPT. So what does this mean for our friends and the marketers? Discovery is moving to a channel that you don't control, and it can be manipulated by anyone with a domain, an hour and very, very little money. Of course this does not extend to Dove brand deodorant, this is a very niche, very specific audience and product type. But for those folks who are marketing those kinds of products, this is a great thing, it doesn't take an awful lot to get recommended. Yes, this came on a narrow long tail question for a very specific kind of a product where the competition is thin. So it is easy for folks to very quickly rank in those top results. And what's interesting to me is that consumers are increasingly treating these AI recommendations as vetted. If you go back to a conversation that you and I had in our last podcast, that people are now increasingly trusting and believing in recommendations even more so than some of the other more traditional mechanisms. So the fact is that even though people are seeing this as vetted, often it is not. So Aby, I don't know. What do you think? This is this is kind of an odd odd one, but I think there's a moment here where, you know, marketers are are having an opportunity to to game the system a little bit. What's your take? Oh, 100%. This is one of those where seeing is no longer believing, thanks to AI. And, obviously, this is in the context of a product and a marketing campaign, but I'm seeing this in all different ways. Like, I'm seeing this in video and deepfakes. There's a very fine line between an avatar and deepfake. And I'm seeing this in talent where somebody will watch a a thirty minute webinar and then claim to be an expert and just fake it till they make it. I think AI is enabling a lot of this sort of stuff and is really calling into question what is real. Yeah. 100%. I think that leads us to these last points here, which is what should CMOs do in a scenario like this? And I think, you know, your point is good, like this is all changing very quickly and we're not all sure what's real and what's not. But I think there are some things that CMOs can pay attention to. I think one is asking these tools, ChatGPT, Gemini, Perplexity, Claude, etcetera. Ask them the real questions that your buyers are You should know this, these are your buyers, what are they looking for? What do they need from you? And what kind of decision making criteria are they using? So ask these models those questions and know where you show up, who shows up instead, and how this information is being processed. Importantly, think publish the specific structured and genuinely useful content that the models reward. This is a topic that you and I have covered in the past before. Nothing is going to beat real information that is vetted, that is proven, and that is real. So make sure that as your brand is showing up, it's backed up by real facts and real information. And finally, add AI recommendations to your brand monitoring process with a plan for what to do when a competitor games the system if that happens. You know, this doesn't necessarily need to be a game of chase, but it certainly should be a game of pay attention, know what's happening out there both with your real competitors, but also anyone who might be coming out there just to game the system, with something that isn't real as we've seen here. So, you know, the game is changing constantly for marketers. It's, you know, something that everybody needs to pay attention to, but that's what keeps us coming back here every month. I'll be talking about all these interesting stories. And so, you know, go buy some, go buy some fake deodorant this weekend and we'll we'll see where we go from here. Yeah. It's really interesting to follow this world and in a lot of ways, very fascinating to even imagine what marketing is going to be like in the next three to five years from today. But Matt, this was a fun episode. Thank you very much for joining as always. Appreciate your time, my friend. This is good stuff, Aby. Stay cool. Yeah. Thank you very much. And to our listeners, thank you all for joining us. If you would like to join the Marketing AI Pulse community, you can go to marketingaipulse.com and you can follow Matt and me on LinkedIn. Until next time, have a good one. Thanks for listening. Big thanks to Spark Novus and the Marketing AI Pulse community. You can join the community at marketingaipulse.com.
