Ryan Law: How To Defend Your Marketing Career in the AI Era | Ahrefs Evolve 2025

Hello. Hello. Hello. Hello. Wow, that was a hell of an introduction. I'm just a dude that writes blog posts. Okay, like don't get too excited. Uh, does anybody recognize this? I reckon a few of you probably will. Back in uh 2020, I got early access to the closed beta of GPT3. And this is what it looked like to start with. Now, at that time, I was at a content marketing agency called Animals. So, of course, the very first thing we did was use GPT3 to write a blog post. We wanted to see how much trouble we were going to be in. And this is that blog post. We used GPT3 to write about how you would use AI in the future of content marketing. And at first blush, it looks like a pretty good blog post. It's coherent, no obvious mistakes. But if you dig a little bit deeper into this article, there were a litany of quite insidious problems hiding in it. It was very salesy and hamfisted. The narrative structure was a total mess. It would start an idea over here and then never finish it. And GPT3 cannot do math to save its life. At this point, it was very tempting to dismiss AI completely. Right? I was the best writer you could imagine. I was a better writer than the best AI models created by the world's machine learning experts. Ryan won, robot zero. My career was completely safe. But the more I sat with that article, the more this kind of nagging sense of familiarity tugged at me. There was something I recognized within those words. It wasn't plagiarized or stolen, but I recognized it. And then it dawned on me GPT3 was me but 10 years ago. So this is me in my very early career with a glorious head of hair, RIP. Uh and this is one of the very first blog posts I ever published on my uh personal copywriting blog. The anatomy of a perfect landing page and very similar to that GPT3 article. At first blush, it looks pretty good. But if you dig deeper, it also kind of sucks as well. All the same problems that plagued that GPT3 content, you can find plaguing my early work as well. Uh the narrative structure was a total mess. I was very salesy and cliched. Uh yeah, generally just not a very good blog post. I had a realization at that point, something that I've carried with me to this day. It's probably a realization that many of you have had when you think about AI and its impact on your career. I realized I was basically a professional writer in a world where professional writing was about to become free. Uh GPT3 already that early version was as good as me in my early career. It took me thousands of hours, maybe even a thousand articles to get better. I thought AI was going to be a lot better a lot quicker. So, I had a uh two choices ahead of me at that point. I realized I could either scuttle off into the woods. I could abandon content marketing forever, never set my fingers to a keyboard ever again, or I could stand and fight. I could actually do something about this AI problem. I should actually take the fight to the machines. And after 3 weeks living alone in the woods, uh it was cold and lonely. I realized I should probably actually do something about this AI business. Uh this wasn't a good look for me in the long run. So I put together my AI defense plan. Uh I tried to be proactive and turn this source of potential weakness into a source of strength for me. And this talk is that plan. The very first thing I realized I had to do that I think we all have to do is embrace the threat posed by AI because it does pose a threat to many of our careers. The first risk of AI is that it will just do your job wholesale. There was this great bit of research from uh the good folks at Microsoft. They tried to see how effective AI would be in replacing certain types of professional roles. I was delighted to find myself represented at number five in the form of writers and authors. And don't worry uh lots of you all of your careers, your job descriptions, you will also find yourself in this research. Uh PR, sales enablement, translators, technical writers, uh we are all apparently at great risk of AI just doing that job for us. And it's not just Microsoft obviously. This is uh generative AI's greatest hype man, Mr. Sam Olman. Boo. Indeed. Uh, AI will handle 95% of what marketers use agencies, strategists, and creative professionals for today. That sent a deep shiver down my spine when I came across that quote. Now, even if AI doesn't get to that point, even if it doesn't replace your job wholesale, there's a very big chance that somebody else using AI will take your job, will outclass you. Uh, we did some research at the start of the year. We surveyed about 900 marketers to see how they actually use generative AI in their marketing. I know Patrick covered a bunch of this data, but just for emphasis, 87% of the people we surveyed were already using generative AI in their content marketing. That was way higher than I expected it to be. And what were they using it for? The thing I make a living out of, writing blog posts. By far the most popular use case for generative AI. Uh, a bit of a cold sweat broke out when I read that. And those people managed to publish a lot more content every month as well. 42% greater publishing frequency with AI. And that's a very big deal in content marketing. And just the final nail in the coffin, most of those people we surveyed actually thought the AI content performed just as well as the human written content. Uh, its ability to rank, to drive traffic. Most people thought AI was just as good as human content. The third risk is though we needed a third risk because I'm thoroughly terrified at this point. AI will just render your job completely irrelevant. You've yeah seen this kind of graph, the great decoupling. You may even be the owner of one of these beautiful charts like we are at Hrefs. Uh many businesses over the past year have seen a massive diminishing of the clicks they get from organic search and biggest culprit as far as we can tell are AI overviews. The research we conducted we saw a 34.5% reduction in clicks to the top ranking article because of generative AI because of AI overviews. And when I looked at the HF's blog AIO's currently appear for 21% of our ranking keywords. Quite often these are some of the highest volume keywords, the things that we care about the most. Suddenly we are competing with basically a Google branded blog post sitting at number one in all of these SERs. Crazy. Yeah, I was feeling pretty bad at this point, right? But we've done the hard thing. We've articulated the threat and now we can actually do something about it. The next thing to do is actually find a way to apply generative AI into your work to actually use it in a useful capacity. One of the reasons people get very stuck using generative AI, they ask too much of it. They ask very big complex goals that actually most people could not do well. Write my next blog post for me, automate my LinkedIn post, create my monthly performance report. Instead, a much better framing comes from this man, Mr. Will Reynolds of Seir Interactive. Uh, he had this AI survival checklist and he has this question he asks himself. How much of my job is a checklist? Checklist items are small, repeatable units of work that AI can do very well. It's both a threat to you if your job is entirely composed of checklist items, but it's also an opportunity. These are the parts of your work you can use AI to automate. And even in my ostensibly very creative, quite autonomous job, there's still a lot of drudgery. I still have a lot of checklist items that I do every day in the same way. Uh writing reax queries in Google Sheets, sharing posts on social media, reading tons of blog posts for the content we create. These are much better applications of generative AI. These smaller self-contained units of work. Instead of trying to get it to write a blog post for you in one go, use deep research to understand the article topic in a much faster and more effective way. Instead of trying to automate all of your LinkedIn content, maybe use it to surface quotes and ideas from the material you've published for you to respond to. Instead of trying to create the performance report, maybe get it to write the reax queries. And it may feel a bit underwhelming to automate these tiny little things. We want to do the big exciting stuff. But actually automating the small stuff is how you end up automating the big stuff. This is Gaul's law in action. Big complicated things generally have to come from smaller, simpler things that worked in the first place. And a blog post like this, this is a big complicated system, right? It's just words on a page, but there are a lot of competing goals and ambitions that go into those words. We're trying to answer the user's question. We're trying to promote our product. We're trying to be entertaining. That is very hard for AI to do off the bat. But even a big complex thing like writing a blog post is itself comprised of many smaller checklist items. This was an exercise we did. We introspected into what it actually takes to write a blog post at HFS. And these are the discrete steps. Uh content brief coming up with the idea, the intent, the guidance for it. Outlining where you pull all the key ideas together and work out the relative weighting and importance of all of them. Structural editing. And each of those steps is itself comprised of checklist items. Uh structural editing of an outline. This is probably one of the most important parts of good content. I could only think of five small mental heristics that happen during that process. This is stuff that AI can do very very well. And actually that blog post from earlier that is an AI generated blog post. We basically created a custom uh chat GBT project. We uploaded SOPs for each of those stages and we work through it using the AI to create what is pretty decent content. In fact, it's annoyingly decent actually. Uh it performs in most cases a lot better than the human written content I've created. Uh that one blog post over 21,000 visitors from search, from social media, even from AI search as well. So we're now using AI. We're automating small parts of our work. The next thing to do is to develop your sense of taste, your judgment. This is the perfect metaphor for generative AI, right? It is the ultimate spaghetti at a wall machine. We have basically democratized access to content creation. We can create unlimited number of blog posts of images of videos totally for free. And there is a big big problem with that because generally most of everything is bad. This has been true of human creativity. This is doubly true of AI generated stuff. And most marketers generally we've been judged on our output. Our ability to get done and make things. That matters far less in a world where AI can pump out content basically for free. What matters much more now is our sense of taste and our judgment. our ability to look at these myriad outputs and work out which ones are good, which ones are bad, and which ones are likely to perform well in the future. You may have seen this amazing video interview with uh legendary music producer Rick Rubin. The interviewer asked him, "Do you play instruments?" And he said, "Barely." "Do you know how to work a soundboard?" "No, I have no technical ability and I know nothing about music." And the desparing interviewer at that point said, "What the hell are you actually being paid for? The confidence that I have in my taste and my ability to express what I feel. That is the role marketers need to inhabit in this world where content creation is basically infinite." And this applies to marketing just as it does any other creative discipline. I love this tweet from Brian Hallagan. Overrated marketers with good analytical skills, underrated marketers with good taste. And taste is not this innate thing that you either have or you don't. This is something that you can develop and improve upon over time. I think the most important thing to do is to view your brain as a collection mechanism for inspiration. Uh everything becomes a source of possible inspiration, both good things and bad things. For me, I collect everything into my notes every day. If there's anything that triggers any kind of thought, any reflection, I save it to review later. And the next step is to critique that work to say, why is this thing successful? If it is successful and the inverse of that, if something doesn't perform how you expect, why is that the case? What do I like about these? Uh, and the most commonly used uh critique in my head, what do I hate about this thing? If I see something on the internet that I despise, I try and introspect and work out why so that I don't do the same in my work. Are there any techniques or devices in the creation of this content that you recognize that you can apply in the future? How would you improve upon this work? If you were to change one dimension of this, what would you do to give it a better chance of success? And most important of all, what does the MI version look like? If you take the general framework implied by whatever bit of content you've seen and apply it to your world, your business, your problems, what might that look like? The final step then is to copy, to imitate. In the same way that musicians play cover songs, athletes drill, famous plays, artists imitate all the great master works. The best way to understand how good things work, how they function, is to imitate them and build your own versions of these things. Another great quote from Rick Rubin here. As long as you're not plagiarizing something wholesale, actually trying to imitate another person, another brand's work is one of the best ways to understand it. And you'll end up creating something new in the process. It's impossible to imitate another artist's point of view. We can only swim in the same waters. And we do this at HFS a lot. Uh this is a real example from like a social media post we published recently. Uh so Tim, our CMO, he's fantastic at gathering sources of inspiration, things he likes and doesn't like on the internet and sharing with us the team to reflect upon. He saw this uh pretty viral LinkedIn post and he said, "Hey, there's probably a version of this that we could create given our own research." So we discussed it, we analyzed it, we tried to work out what happened in that Slack thread. Uh Tim came up with an intro hook for it. I said, "Hey, maybe I'll make an image of myself as a baby again." Bit of a long story, but it worked very, very well. Uh Tim suggested, "Hey, let's use a whiteboard." And then I imitated it. I made my own version of that successful post. We thought it would work well. We'd already vetted that and it did. It actually performed better than the original post. That was a process of taste acquisition. You can do that in your work too. Now, one of the reasons uh well, one of the biggest barriers to actually changing your work and adapting to this AI world is you. It's yourself. It's your own sense of identity. We get very wedded to the way we do things now because that's all we know. And that calcifies into your identity. In fact, uh at the start of my career, this is the type of content that worked very, very well. You probably recognize this type of content. the ultimate guide to whatever 10,000 words, all these longtail keywords crammed together on page navigation uh to what we would now call skyscraper content. And that became my identity. For the longest time, I was the skyscraper man. That was the content I made. That was how I thought about myself. Uh this was how I sold my services and communicated to other people. But over time, we see diminishing returns from that type of content. it doesn't perform as well as it used to do. Many writers today are still stuck with this identity. They still try and uh force this type of work into the world even though it doesn't work very well now. But when I joined animals, this was the type of content that was in Ascension. Uh thought leadership content. We're talking models and frameworks and 2x two matrices and very pretentious writing if I'm being honest with myself. But this also became my identity. I switched gears from being skyscraper man to being thought leadership man. This was my mental conception of myself. Uh sitting in a smoking jacket with a cigar and a glass of brandy talking about the unity economics of content marketing. But as with every other marketing strategy over time it works. People copy it. It becomes less effective. Your identity has to change again. This is the type of content that I see working very well at HFS today. Original research content. This is now my identity. I'm in that process of trying to change how I think about myself, develop new skills, do all this scary stuff because I'm trying to chase the tailwinds. I'm trying to do the type of content that works best. And I look pretty great in a lab coat as well, it seems. And what happens when you're willing to change how you think about yourself? You're willing to change the nature of the work you do, you get better results from it. Uh that blog post I mentioned, that bit of research into AI overview click-through rates. Uh we have about 2 and a half thousand articles published on the HF's blog. This entire year, the second most popular piece of content was that one short research study. 87,000 views, over a thousand domains, BBC, IBM, Wired, you name it. This is not a new concept, right? This is just Andrew Chen's law of shitty clickthroughs in action. Over time, every marketing strategy gets less effective as more people do it, as you get used to it, as the novelty wears off, as it becomes saturated. But the thing that's changing with AI, this cycle of change is accelerating vastly. Before in the course of somebody's career, you'd maybe cycle through successful marketing tactics every 5 or 10 years. Now it's much easier to test new strategies to scale them. We are going to have to cycle through different marketing strategies faster than ever. And our identity is going to have to change alongside that. So how do we do that? How do we change our identities? How do we adapt to new skills that we need in this world? AI is actually very good for skill acquisition. You've probably come across this idea of uh being T-shaped, right? The T-shaped marketer. This is the idea that great marketers tend to have deep expertise in one core area and then they have a series of kind of superficial related skills. You don't need much competence in them, but just a little bit is enough to make all of your work much more successful. And deep expertise is something that's actually very hard for AI to help you acquire. It's hard to do. Uh so many years of research and nuance and context that goes into that. But AI is very very good for helping you acquire superficial competence in these broad skills. It is amazing for getting to kind of 20% competency in whatever area you're interested in. And if we apply this idea, the T-shape to content marketing, maybe it looks something like this. Certainly for me in my career, writing has been my deep expertise. That is the thing I've spent all my time doing and thinking about. I've gone deep into search optimization as part of that. Uh editing and copy editing, interviewing, writing thought leadership content. But even a tiny bit of competence, just enough to be dangerous in areas like research and coding and data visualization and design, that is enough to vastly amplify the success of my work. Now, of all of these things, I decided I would use AI to try and acquire some of these skills. Uh, any guesses as to which one of these I picked? Ready? That's right. I picked coding. I thought this was a useful thing to learn with AI. Uh this is I think Tim talked about this briefly. This is the hideously ugly but quite fun Python script I put together for our monthly blog reporting. Uh abley assisted by hundreds of chat GPT conversations. Every uh view, every report, all the data I need every month, all the visualizations for our monthly reporting, it's all in this one ugly, hideous Python script. Now, am I a good coder at this point? Absolutely not. Uh, I do things that would make my team probably sobb into their keyboards if they saw what I was doing. But I've managed to build things that are very useful for myself with AI. I have just enough understanding and competency to be dangerous and to help understand and communicate with the rest of our team. I think it's quite dangerous to allow AI to guide your education wholesale. Uh, you'll end up learning things that may not be useful, that may not be vetted. I think much better is to find some form of professional scaffolding to start with. Find an entity or a person that you trust and use that to get started to provide the building blocks of this new skill acquisition. Uh I did this Python course by a company called Replet that was just enough to get me started in a useful direction. From there you can use AI to tutor and troubleshoot. Whenever I ran into a problem with a script I was trying to write, I would not only ask AI to solve it and fix the problem, I would also ask it to elaborate on the problem to reveal whatever underlying concept or framework I needed to understand to avoid this happening again. And most important of all, AI is amazing at helping you test and apply new skills. Uh this is always one of the most important and difficult parts of skill retention, actually applying what you learn. uh every day at 8 a.m. Chat GPT sends me a little pop-up notification with a Python test, something that I can choose to engage with every day and keep applying my skills. And in the same way, uh every time I sit down to do a bit of work at HFS, I have a chat GPT window open and I say, "What does the Python version of this look like?" It doesn't always work out, but I've ended up with 30, 40 different uh Python scripts that are really useful for all different areas of my life. So the final final point, possibly the most important. I think this echoes a lot of what uh Mark was saying earlier, playing like a rock star. I really like this analogy. Uh I think most marketers are very like session musicians as represented by sad monochrome Ryan on the left of the screen here. So session musicians are incredibly intensely skilled people. But fundamentally they are uh judged on their output, the things they make. They contribute uh melodies and riffs and solos to other people's work. They exist behind the scenes. They contribute to the work that other people are doing. And by necessity, they have to be stylistic chameleons. They have to adapt whatever other people are doing. That is a problem in the AI era. Session musicians are better uh AI is better at all of those facets than uh humans are. AI is skilled. AI is incredible at output. AI could adapt to any situation you want. Uh if that is the sum totality of our skills, if we are just the behindthe-scenes workers pumping out faceless content, I think we're going to be in trouble from AI. Instead, we need to be more like rock stars. We need to look like glittery tasselwearing Ryan on the right hand side of the screen over here. Like the session musicians, rock stars are also highly skilled. But there's much more to it than that. Fundamentally, they are judged on the experiences they create. You go to Wembley Stadium not to hear a particular melody. You go because of the ambiance, the smoke machine, the lights. They are public figures. You get to know them. They develop their own followings, their own personas, their own brand. And they are oures. They have a a known way of doing things, a recognizable style that you can uh understand and anticipate from them. In the age of AI, it is the rockstar marketers who will be irreplaceable. The people who don't just create faceless things, but actually attach themselves to those as well. So, how do you do that? How do you become a rockstar? Uh, and my colleague Louise would absolutely hate me for putting this slide up, but she's done a very good job at this. AI is amazing at detaching information from the source that created it. So, we have to actively fight against that and we have to plaster our faces on all of the things that we make. Louise could very easily just publish an article on the HF's blog and, you know, never think twice about it, work on the next thing. She does a lot more than that. She posts about her experiences on her personal social media accounts. She records videos talking through the processes she's done. And she even takes the ideas on tour. She goes on webinars and podcasts. This is going to become more and more important. We can't just be judged on the work we create. We have to attach ourselves to the work as well. You can share an opinion. AI is the ultimate hedge. Uh it loves to say broad uh you know non-opinated things. It's very reluctant to say something definitive and clear. We can cut through that noise by sharing very strong defensible opinions. Uh it's a really hard thing to do in SEO. People will jump down your throat if you share the wrong opinion, if you frame it the wrong way. But uh Despina from our team, very brave, she shared some amazing opinions. Uh LLMO is in its black hat era. Chunk optimization is overrated. Uh it definitely is overrated, by the way. And last of all, you can develop a style, a recognizable style. It doesn't even have to be a good style. It's just some way of signaling to the reader, to the audience, that hey, I was here. Uh, I publish these long, hideous textbased posts on LinkedIn. They go against every social media convention you've ever been told works, but people come to anticipate this and associate it with me. And it works very well. Now, these every three of these had over a thousand likes on these text posts. In the AI era, it is your uniqueness that is going to provide the longest lasting advantage. Instead of using these AI tools to smooth out all of your sharp edges and create a ton of homogeneous AI slop paste, you can use it to accentuate your weirdness to become not just a marketer, somebody behind the scenes, but a rock star as well. So, just a quick recap, this is the six steps of the AI defense plan. You got to embrace the threat. You have to be honest with yourself about the risk posed by AI because I think for many of us, it does genuinely pose a risk. You have to find small ways, small units of your work that you can automate and use that to ladder up to automating the bigger, more complicated stuff. Develop your own sense of taste and judgment. Be the person that can look at a thousand AI outputs and have a good sense of which ones are likely to perform best. Be willing to change your identity, to take on new forms of work that you wouldn't otherwise do to test yourself and stretch yourself. Try to become T-shaped. Use AI to uh actually acquire new skills in a more efficient, a more fun way. And lastly, play a bit like a rock star. Uh attach yourself and your identity and your weirdness to the work that you do. AI will not be able to take that away from you. Thank you very much for listening to me. It's been a pleasure talking to you all.