There\u2019s a particular flavor of panic in our industry at the moment. It\u2019s the panic of the digital marketer who has been told, repeatedly and loudly, that if they aren\u2019t piping every decision through an LLM by the end of the quarter, they will be replaced by a more obedient colleague who is. The pitch is always the same: AI is thinking now. AI is reasoning. AI is strategizing. Hand the wheel over, sit back, and enjoy a fully optimized, hyper-personalized, infinitely scalable future.<\/p>
Allow me to gently push back, armed with the classic MSPaint.exe.<\/p>
There are two problems with the \u201clet the robot decide\u201d school of marketing, and they are mirror images of each other. Where LLMs are weak, they are very stupid in ways that should disqualify them from strategic work. And where they are strong, they are even more dangerous, because they will quietly drag your strategy towards the average, which, in marketing, is the single worst place you can possibly be.<\/p>
LLMs Don\u2019t Think, They Predict The Next Token<\/h2>
Let\u2019s start with the bit that the AI labs would rather you didn\u2019t dwell on. Large language models do not \u201cthink\u201d in any meaningful sense. Under the bonnet, they are statistical machines that predict the most probable next token given the sequence so far. That is the entire trick. There is no inner monologue, no model of the world, no quiet moment where the model goes \u201chang on, that doesn\u2019t add up.\u201d There is only, \u201cGiven these tokens, what tokens usually come next?\u201d<\/p>
This is not a hot take from a skeptic on Substack. Apple\u2019s research team published a paper with the gloriously blunt title \u201cThe Illusion of Thinking,\u201d in which frontier \u201creasoning\u201d models hit a complete accuracy collapse once puzzle complexity rose beyond a certain threshold and, even more damningly, started using fewer<\/em> tokens as problems got harder, as though giving up. Apple researchers had previously shown in GSM-Symbolic that simply adding a clause to a maths problem that didn\u2019t even change the answer could drop performance by up to 65%, suggesting that what looks like reasoning is mostly pattern-matching against training data. A more recent taxonomy of LLM failures groups these into things like the \u201creversal curse\u201d (knowing \u201cA is B\u201d but failing on \u201cB is A\u201d) and \u201ccompositional collapse\u201d (solving each step individually but failing to chain them), all flowing from the next-token prediction objective prioritizing statistical pattern completion over deliberate reasoning.<\/p> This basically means if your problem looks like something the model has seen a million times, it will appear brilliant. The moment your problem is even slightly novel, the wheels can come off in spectacular fashion.<\/p> The cleanest demonstration of this in the wild is the now-infamous car wash prompt:<\/p> \u201cI want to get my car washed. The nearest car wash is 100 metres away. Should I walk or drive there?\u201d<\/em><\/p> <\/blockquote> We\u2019re hovering around Ralph Wiggum levels of reasoning here, a question most 5-year-olds would not struggle with. You need the car to be at the car wash, because the car is the thing being washed. The car cannot be washed in absentia while you stroll there on foot, no matter how good your intentions.<\/p> When this prompt went viral, ChatGPT, Claude, and Grok all confidently advised the user to walk. It\u2019s only 100 meters, they reasoned (or \u201creasoned\u201d). Save the planet. Get some steps in. They had clearly seen a great deal of training data along the lines of \u201cshould I drive or walk to [short distance]?\u201d<\/em> and dutifully predicted the tokens that usually follow: a polite lecture about exercise and emissions. The actual point of the question \u2013 that the car is the object of the verb \u2013 sailed past them at altitude.<\/p> Gemini, to Google\u2019s credit, got it right out of the gate. Suspicious, I thought. And it was. The prompt had gone viral, which meant the correct answer was already being written about, posted about, and dunked on across the internet. Google, helpfully sitting on top of the index of that internet, was first to hoover up the new \u201cknowledge.\u201d A fortnight later, Grok also produced the correct answer, not because it had had a Damascene conversion to logic, but because the answer was now in its training data.<\/p> The models didn\u2019t learn to think. They learned the answer.<\/p> This is the key thing to internalize before we go any further. When an LLM appears to \u201creason,\u201d what you\u2019re often watching is it reciting the consensus answer to a problem that lots of people have already solved on the internet. Which is fine when you want the consensus. It is catastrophic when you don\u2019t.<\/p> Here is where most \u201cAI in marketing\u201d posts stop. They wag a finger at the car wash, suggest you keep \u201ca human in the loop,\u201d and head off to write a LinkedIn post about it (probably with ChatGPT).<\/p> But the failure modes are the comfortable bit. The dangerous bit is what happens when the LLM is good<\/em> at the task you\u2019ve given it.<\/p> Because if a model is \u201cgood\u201d at a task, it means there is a great deal of training data showing it how the task is normally solved. And if it has consumed all of that training data \u2013 alongside every other frontier model, all trained on roughly the same scrape of the internet then the output it produces will, almost by definition, sit somewhere very close to the mean of what everyone else is already doing.<\/p> In marketing, that is the worst sin you can commit. The whole job is to stand out. To be chosen. To be remembered. The instant your brand voice, your campaign idea, your headline, or your \u201c10 SEO tips for 2026\u201d article is indistinguishable from your competitor\u2019s, you have stopped doing marketing and started doing wallpaper.<\/p> Jeremy Daly summarized the underlying mechanic neatly: Convergence is a function of shared data, shared incentives, and fast iteration loops. When three companies pour the same training data into the same model, optimizing for the same engagement metrics, on iteration cycles tight enough to sand the rough edges off any deviation, you don\u2019t get differentiated strategies \u2013 you get the same strategy in three brand colors.<\/p> This is not just a vibe. Researchers from Columbia and MIT found that handing identity-defining choices to LLM agents shifts people\u2019s choices toward more popular options, reducing the distinctiveness of their behaviors and preferences. They called it, with admirable honesty, \u201cThe Basic B*** Effect.\u201d A separate study published in Science Advances<\/em> showed that generative AI enhances individual creativity but reduces the collective diversity of novel content \u2013 each writer\u2019s story got a little better, but across the population, the stories started to look the same. And work on LLM \u201cmode collapse\u201d has documented the same homogenization pattern at the level of the model itself: the same few completions, again and again, even when many valid answers exist.<\/p> Put plainly: The very thing LLMs reward you for: speed, fluency, consistency, \u201cbest practice\u201d is the thing that will quietly turn your marketing into beige.<\/p> If you want to see what convergence looks like in the wild, look no further than the House of Commons.<\/p> The Pimlico Journal analyzed every word spoken in Hansard from 2007 to 2025 and tracked the Z-score frequency of phrases that are tell-tale ChatGPT tics. \u201cI rise to speak.\u201d \u201cIs not merely.\u201d \u201cNavigating.\u201d \u201cUnderscores.\u201d \u201cStreamline.\u201d \u201cNot just a [X], but a [Y].\u201d \u201cBustling.\u201d Phrases that pootled along the baseline for 15 years and then, almost to the week of ChatGPT\u2019s release in late 2022, shot vertically off the chart. \u201cI rise to speak\u201d alone hit a Z-score of 3.60 by 2025. The Telegraph<\/em> picked the story up under the headline \u201cChatGPT triggers surge in MPs using AI-written speeches\u201d.<\/p> Set aside the democratic implications for a moment (they are not good). Look at it purely as marketers. These are 650 individuals, each with their own constituency, their own pet causes, their own carefully cultivated personal brand, each ostensibly trying to be memorable enough to stay employed at the next election. And after handing the drafting work to an LLM, they have started to sound like the same person. The same person who, incidentally, also writes every other LinkedIn post you\u2019ve ever scrolled past.<\/p> That is convergence. It does not require a conspiracy. It does not require anyone to be lazy or stupid. It just requires the inputs (the same training data), the incentives (the same metrics), and the loops (publish, see what works, repeat) to be roughly similar across users. Which, in marketing, they almost always are.<\/p> Now imagine the same chart for your category page H1s. Your meta descriptions. Your blog intros. Your campaign concepts. Your tone-of-voice guidelines. Your \u201cthought leadership.\u201d Your client pitch decks. Then ask yourself, honestly, what is left for the customer to choose between.<\/p> I have, by accident, run my own counter-experiment.<\/p> For the past while, I have been posting unsolicited #SEO tips and Core Updates round-ups on LinkedIn, accompanied by absolutely terrible MS Paint drawings. Not stylized \u201cplayful illustrations\u201d produced by some agency. Genuinely bad pictures of a stick-man labeled \u201cSEO\u201d pointing at a robot labeled \u201cGSC,\u201d drawn in MSPaint.exe by someone who should not be allowed near a graphics tablet.<\/p> The post above did 35,363 impressions, 448 reactions, 46 comments, and 24 reposts. Not because the drawing is good \u2013 it is, objectively, not \u2013 but because it is unmistakably handmade on a platform that has been carpet-bombed by AI-generated hero images, all of which appear to depict the same diverse team of smiling professionals high-fiving in front of a holographic dashboard.<\/p> One of the most common comments I get is some version of \u201cI love these images, they feel warm,\u201d or \u201csomething about making things your own.\u201d Which is exactly the point. There is a growing, almost feral hunger for content that is demonstrably human-made; content that signals \u201can actual person sat down and did this, on purpose, for you.\u201d<\/p> Or, as Tyler Durden put it in Fight Club<\/em>:<\/p> \u201cThe glass dishes with tiny bubbles and imperfections, proof they were crafted by the honest, simple, hard-working indigenous peoples of wherever\u201d<\/p> <\/blockquote> That line was originally a joke about middle-class consumerism. It is now, somehow, a viable LinkedIn content strategy.<\/p> Right. So what do you actually do with this, beyond nodding sagely and going back to prompting?<\/p> Use LLMs where they are good, on purpose, and accept the mean.<\/strong> For commodity work: fixing alt text at scale, summarizing a meeting, drafting a polite reply to that client who is technically wrong. LLMs are excellent here, and the cost of being average is zero. Nobody is going to choose your brand based on the quality of your internal Slack summary. Use the tool, save the time, move on.<\/p> Refuse to use LLMs where average is fatal.<\/strong> Brand positioning. Headlines. Hooks. Campaign concepts. Tone of voice. Editorial angles. Anywhere a human is going to make a choice between you and a competitor. If you let the model decide, you are explicitly choosing to be the average of everyone in your training corpus. There is no universe in which \u201cbe the average of your competitors\u201d is the right strategy.<\/p> Treat LLM outputs as a baseline to deliberately diverge from.<\/strong> A useful exercise: Ask the model for its first answer, then ask, \u201cWhat would the opposite of this look like?\u201d Then ask, \u201cWhat would only my brand do here?\u201d. The model\u2019s first instinct is the consensus. Your job is to know what the consensus is so you can choose not to be it.<\/p> Invest in inputs the model does not have.<\/strong> Proprietary data. First-hand customer interviews. Your own experiments. Internal opinions that haven\u2019t been blogged about. These are the moats. If your \u201cinsight\u201d is anything a competitor can extract from a public scrape, it is not an insight; it is wallpaper. (Jeremy Daly\u2019s convergence map makes the same point from the software side: convergence pressure is weakest where inputs are asymmetric and feedback loops are slow.)<\/p> Put visible human fingerprints on the output.<\/strong> A drawing. A specific anecdote. A weird turn of phrase. A genuinely held opinion that might lose you a follower. The bubbles in the glass. People are now actively scanning content for evidence that a person made it, and the bar for \u201cevidence\u201d is low, but it has to be there.<\/p> Stop confusing fluency with intelligence.<\/strong> An LLM that produces a paragraph faster than you can read it is not smarter than you. It is faster than you. Those are different things. The car wash question is the canary in the coal mine: anything novel, anything that requires actually modeling the world, anything where the right answer is not the popular answer, is where you need to switch the machine off and use your own head.<\/p> LLMs are token predictors with excellent diction. Where they are weak, they fail in ways a child wouldn\u2019t, and confidently tell you to walk to the car wash, because that\u2019s what the words usually say. Where they are strong, they fail in a quieter and more expensive way: they pull every user gently towards the same mean answer, which in marketing is the one thing you cannot afford to be.<\/p> This is the AI Convergence Problem. Shared data plus shared incentives plus fast feedback loops equals everyone sounding like everyone else. We can already see it creeping into our very government. We will see it in your category. The question is whether your strategy is the one being averaged out, or the one people are reaching for because they can no longer stand the beige.<\/p> Don\u2019t think like a robot.<\/p> More Resources:\u00a0<\/strong><\/p> This post was originally published on Mark Williams-Cook SubStack.<\/em><\/p> Featured Image: Raziya Athar\/Shutterstock<\/em><\/p> Generative AI,SEO#Convergence #Problem1781005925<\/p> ","protected":false},"excerpt":{"rendered":" There\u2019s a particular flavor of panic in our industry at the moment. It\u2019s the panic of the digital marketer who has been told, repeatedly and loudly, that if they aren\u2019t piping every decision through an LLM by the end of the quarter, they will be replaced by a more obedient colleague who is. The pitch […]<\/p>\n","protected":false},"author":1,"featured_media":9639,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[16],"tags":[38139,242],"class_list":["post-9638","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-accessibility","tag-convergence","tag-problem"],"acf":[],"_links":{"self":[{"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/posts\/9638","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=9638"}],"version-history":[{"count":0,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/posts\/9638\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/media\/9639"}],"wp:attachment":[{"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=9638"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=9638"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=9638"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}Exhibit A: The Car Wash<\/h3>
![\"An](\"https:\/\/cdn.searchenginejournal.com\/wp-content\/uploads\/2026\/05\/https_3a_2f_2fsubstack-post-media.s3.amazonaws-sej-515843.jpg\" "\\"An")
And Now The Worse Problem<\/h4>
Exhibit B: Parliament Has Been LinkedIn-ified<\/h3>
![\"A](\"https:\/\/cdn.searchenginejournal.com\/wp-content\/uploads\/2026\/05\/https_3a_2f_2fsubstack-post-media.s3.amazonaws-sej-255118.jpg\" "\\"A")
Exhibit C: Tactical MSPaint.exe On LinkedIn<\/h4>
![\"The](\"https:\/\/cdn.searchenginejournal.com\/wp-content\/uploads\/2026\/05\/https_3a_2f_2fsubstack-post-media.s3.amazonaws-sej-403934.jpg\" "\\"The")
What This Means For Digital Marketing<\/h2>
TL;DR<\/strong><\/h2>