There’s a lot of marketing people that don’t really get AI, there’s a lot of AI people that don’t really get marketing, and even for the people that get both, they often miss the bigger picture of the broader impact.
Most of the use cases I see for AI in marketing revolve around either; time saving/process automation, or shifting from SEO to GEO (or AIEO or whatever else Old Macdonald had on his farm); or pattern detection, allowing for better attribution.
I’ve taken a look at the ways AI is being used, and I want to give you my frank assessment of them. But I also want to explain what everyone is missing.
The world’s best slop
At ground level, many marketers are still talking about the best ways to use AI in their content creation, to save time and scale up the volumes of content they’re creating. Few admit to outright pumping out AI-generated content, however a quick scroll on LinkedIn posts these days reveals more than a whiff of slop. The em-dashes have disappeared, almost, and are being replaced with the ‘It’s not x, it’s not y. It’s Z - and that’s something or other…’
Similar traces of slop pop up in email marketing, blogs and so on. But the main issue here is not that AI generated content is bad. In fact, it’s already not that bad (I’ve read plenty worse written by humans before chatGPT was invented). The reality is it will become very, very good very soon. At the press of a button, you will probably be able to create a lifetime’s supply of excellent content as good as any human has ever written.
And when that day comes, who will be around to read it? The ability to create content at scale is fast becoming universal, which means it will no longer be a point of difference for anyone. It’s feasible that someday reasonably soon, everyone in the world will be able to give an LLM a few details on their life, and the LLM will create a powerful, moving, poignant novel based on this – the sort of book that would have won a lot of awards a few years ago. But if there’s 8 billion of these around, nobody is going to read them. People will still read something of course, but what?
That’s harder to predict – if I could tell where taste was going next, I would be there leading the vanguard. I heard someone on a podcast recently say (and I’ve tried without luck to track down this source to attribute it, but couldn’t find it, if anyone knows, please tell me in the comments and I’ll update this) that if generative AI had existed in the 1970s, it could have created some great prog rock albums that sounded like Pink Floyd, but it wouldn’t have invented punk.
The point is not that it wouldn’t technically be capable of creating it, but that AI can create great averages of what already exists, whereas human taste and cultural movements will shift, usually in unpredictable directions.
I don’t think this means we are facing a future where only things created “without AI” will be of interest. I think that distinction will fade. I used the “scare quotes” because what is “without AI” anyway? On some definitions, using autocorrect or grammarly to fix mistakes, or photoshop’s old background removing tool, is still using AI. Soon, people won’t care where AI is used, and it will be expected to be used where it’s useful and efficient. What they will care about is whether something is useful or interesting to them – i.e. the more things change, the more they stay the same.
If you’re relying on AI to supercharge your content creation, you’ll probably be able to arbitrage that for a little while longer, but it’s a very short-term strategy. In a world awash with unlimited content, only original and worthwhile ideas will stand out.
In marketing, the value shifts from “having lots of good content” which will be as ubiquitous as “having a website,” to originality, or synthesis, context, distribution, or trust.
When there is more content than anyone can read, then the platforms, the distribution, the attention of the buyer will matter more than ever. In a world where anyone can generate convincing content instantly, trust in the brand and the source become even more valuable.
The Wild West of GEO
Which leads neatly into the second usage, switching SEO tactics and protocols to work for generative AI. People are often asking their chatbots rather than Google when they want an answer now. So of course, making sure your content is readable to these programs makes sense. Last week, the best advice was to add a FAQ to your site, as this resonates well with LLMs and the way users ask questions.
It feels a bit like the early days of SEO where it was something of the Wild West. Some people seem to expect another arms race; where AI models and GEO greyhats race to keep up with giving people what they’re actually searching for on one hand, and gaming the system on the other.
It's another short-term play; I’ve been arguing for 20 years now that the only viable long-term SEO strategy is to have the information that people are actually searching for. With LLMs, it’s on another level – the brands with strong reputations already; good, real reviews etc, are the ones who will be recommended.
Stuffing sites with FAQs, or cramming reddit full of noise is not going to lead to long-term success.
But all this still misses the bigger issue – how much longer do we really think that the internet will be driven by content created by bots, just so it can be read by other bots? We’ve all seen the frustrating situation where someone uses AI to scale up their simple idea into a long and complex email or report, which the person receiving doesn’t have the time to read, so they use AI to summarise it again. It’s unlikely the summary bears too close a resemblance to the original simple message.
Eventually someone is going to realise you can cut out the middle-man. The question is whether the middleman is the AI or the person. But using AI to create a smoother internet experience aimed at humans, when increasingly it’s AI that’s actually navigating the web, feels like a very transitional stage. Soon the bots will just communicate directly with each other, they won’t need to mediate the experience through human marketers and content creators, they will be able to query and find the information they need directly.
What this means for B2B is that marketing as it currently exists is going to be fundamentally transformed, and the whole performative top layer will disappear. When clients realise that instead of making vendors write long tenders with AI and then using their own AI to summarise them, they instead can create a procurement process that finds the information and makes the decisions directly.
Enterprise procurement is already a highly templated and structured process, and this is the sort of decision making that is easiest to automate. Add in regulatory issues, concerns about undue influence, and an automated procurement future for utilities and similar bodies is no great leap.
So much of the commentary looks only at the first order effects of scaling up what you’re doing already, and misses the second order changes to come.
Making false precision more precise (and more false)
The third area where marketers talk about AI is broadly in the attribution space, i.e. enabling AI to track and measure and recognise patterns in your pipeline on a scale beyond what a human could do before.
This can work well in carefully-defined and closed environments where there is a very large amount of data to analyse. Within the Google Ads ecosystem, someone running a large enough campaign can use the data attribution and get some useful insights.
But for the rest of us – and even for the companies using that – the analysis is only as good as the data in the system. In B2B so many things happen in the messy, unrecorded, offline interactions between sales reps, at trade shows, or referrals between customers, or moments of brand recognition.
The problem with getting even more data, and making some parts of it even more accurate, is that it just increases the false sense of confidence you have in that incomplete data. I’ve seen this general trend over the past few decades in marketing generally, as digitisation allowed for precise measuring, there’s often a reliance, or an excessive focus, on what can be measured. And the most credit, and therefore resources, go into things because they are easy to measure, not because they are actually the best.
Imagine you’re counting calories, and you go to the fridge. The only two things in there are an apple and a Mars Bar. The Mars Bar has the exact number of calories written on the packet, the apple does not. You could guess, you could look up the average number in an apple, you could weigh it, but none of these things will give you as precise information as the Mars Bar packet. However if you’re on a diet, you will probably know that none of this means the Mars Bar is better to eat than the apple. I wonder how many data-driven marketers would go for the Mars Bar every time though.
When the data isn’t there, AI won’t fix attribution problems, it will just amplify those problems faster and more convincingly.
Legibility
Marketers, like many others, have both under and over-estimated the power and impact of AI. On the one hand, it will be the most revolutionary technology we have ever seen. On the other hand; the more things change, the more they stay the same.
Generative AI won’t give you any advantage when everyone is using it, you’ll still need original ideas and a strong brand. The answer won’t lie in continually adapting your tactics to keep up with the latest LLM iteration, but in still having what customers want.
To truly stand apart and be ready for a disruptive future which nobody can predict precisely, the best thing all businesses can do is to make sure they are legible.
What does this mean? You need to make sure your offering, your USP, your credentials etc are clear and known. Not just spelled out on your website, but properly defined and understood (and then spelled out) as it may not be long until this is what future decisions rely on.
Your internal systems need to be legible. This means it’s no longer enough for so many key processes and knowledge to sit only in the heads of key staff, or in offline documents, desk calendars and spreadsheets.
It’s not as simple as getting these key staff to document their knowledge, though that is a good start. You need to understand how your business really works. How customers really hear about you, where leads really come from, why some deals close and others don’t, what happens to information internally, where does feedback go, etc.
This isn’t an easy process, but if you map and understand this, you’re on the way to making your company legible. And this means that not only do you understand it, but AI can understand it as well. And then as disruption comes, and as you try to attribute and understand your marketing, or automate and improve your processes with AI, you can do so from the starting point of actually understanding how it all works.
When software, AI and other tools are added onto a fragmented system, in the hope that it will fix the problem, all it does is amplify the existing problems. Adding another tool won’t fix a problem, if you don’t understand the underlying problem you’re trying to fix.
Where to next?
B2B marketers, we are in for a wild ride in the next few years. By all means, save time and make the most of the tools at hand (just remember to clean up the em-dashes), but don’t lose sight of the bigger picture.
Churning out content, even if it’s good, and optimising your site for LLMs won’t save you either. Original ideas will still cut through, and people will make decisions based on trust and brand; distribution and attention will count for even more, at least to the extent that humans are still involved in the decision-making process. Beyond that, in B2B, AI will likely make decisions more on hard facts, which are harder to game.
The best thing any of us can do is make sure we understand how our business works now, so that it can adapt as needed, because change will happen faster than we think and in ways none of us can fully predict.
