Why we look at the algorithm instead of the user: SEO in the age of LLMs

There’s a curious, and I’d almost say inevitable, phenomenon in the way companies are approaching the generative-AI tsunami. Brands, marketing teams, content departments, and consultants all seem perfectly aligned in the same reflex: wondering what’s going to happen to SEO. What will change. What will disappear. What algorithm needs to be decoded. What hack needs to be uncovered.

The paradox is that while everyone is obsessing over algorithms, almost no one is looking at the user. It’s the same pattern we’ve repeated with every tech wave: a new tool appears, the focus shifts toward understanding its mechanics, its shortcuts, its metrics, its “tricks,” and the audience — real people — gets pushed to the background once again.

But this time, it’s different. Not because the technology is more sophisticated, but because, for the first time, the interface is no longer a search engine… it’s an interlocutor.

And now that we know what each one is good for, the real question is: how do we use them? Search is no longer a transaction; it’s a conversation. And when search becomes conversation, SEO stops being engineering and inevitably becomes brand strategy again.

This is the part many companies are missing: LLMs are pushing brands into a space where the priority is no longer to optimize the machine, but to exist clearly in the minds, and in the words, of users.

But before we get into what this means for brands, it’s worth pausing on the cultural shift that makes it possible.

Search is no longer a query; it’s a dialogue.

Since Google arrived, we’ve all searched the same way: by typing keywords. “Cheap car insurance.” “Restaurant reviews.” “Gym near me.” “The order of the Star Wars movies.” We weren’t talking, we were telegraphing. We learned to write like machines so machines could understand us.

The wild part is how normal that became. So when the possibility of speaking to an AI naturally appeared, we felt the difference instantly. We shifted from keyword chains to full sentences:

• “Which restaurant in Madrid would you recommend for an anniversary without the usual crowds?”
•  “I need ideas for my history assignment — around 500 words.”
•  “Make me a weekly meal plan and a list of ingredients to buy.”

@hot_town_

This is by far my favorite prompt when working with LLMs (ChatGPT) when coding or writing. I call it the possible solutions thinking prompt. What prompts do you find to be super effective when working? #vibecoding #chatgpt #prompting #aiassistedcodong

♬ original sound – Vinny

What changed wasn’t just the form: it was the level of context. When a user speaks in full sentences, the semantic load multiplies. In Spain, over 50% of the population already uses AI to guide purchases, 50.4% use chatbots for customer service, and 21.6% receive post-purchase recommendations directly from generative AI. Meanwhile, 52% trust AI more than traditional influencers, 75.8% say AI has transformed how they compare options, and 59.8% make decisions faster thanks to AI. 

This shift isn’t technological, it’s cultural. People aren’t changing how they search because AI allows it; AI allows it because people want a more human way to ask questions. And that’s what blows up the entire SEO paradigm.

Brands keep staring at the algorithm because it’s easier than looking at the user.

SEO has always offered a comforting sense of control. You could do things: optimize titles, tweak keyword density, restructure headers, get backlinks, improve site performance. You could invest, measure, tweak, you could play the game.

LLMs shattered that illusion.

• There is no page 1 of ChatGPT.
• There is no keyword to “rank for.”
• There is no public leaderboard.
• There is no Search Console.

And that is uncomfortable. 

Because it forces brands to stop thinking about the algorithm and start thinking about people. It forces them back to questions they’ve avoided for years thanks to the engineering layer of SEO. While teams argue about keeping a “120-post long-tail SEO blog,” the user is asking ChatGPT: 

“What shoe brand would you recommend if I want a modern vibe without that old-school, outdated look?” 

It doesn’t matter how many generic articles you have: if you don’t exist in the cultural conversation, the AI will not recommend you, not because it’s ignoring you, but because it doesn’t know you. LLMs don’t work like search engines. They don’t retrieve; they interpret.

@maverickgpt

How to use ChatGPT to find the best Black Friday deals 🤫 #chatgpt #blackfriday #deals #fyp #chatgpttips @ChatGPT @OpenAI

♬ original sound – Maverick | AI & ChatGPT

The new SEO: you don’t optimize content, optimize perception.

Here’s the real shift: relevance is no longer built for an algorithm, but for a language model trained on millions of human interactions. 

Meaning your placement inside LLMs depends not only on your pages, but on your cultural footprint. LLMs don’t decide by keyword density they decide by coherence,
don’t prioritize technical structures and prioritize linguistic patterns. To exist inside a model, a brand needs to show up across multiple layers: 

1. Clear, consistent content.
2. Cross-network presence.
3. Recognizable reputation.
4. Strong semantic associations.
5. A stable narrative over time.

LLMs don’t crawl. They learn. And that changes everything. Voice matters because the model also “has” a voice. A brand without a defined voice, tone, or coherent presence has a fundamental problem in a conversational-AI world: it doesn’t exist as a coherent entity. If the AI can’t infer a recognizable way to talk about you, it can’t meaningfully talk about you at all.

@marcusaureliusdigital

🚨 You’re doing AI SEO wrong, here is how to do it right. Rank your pages using this SEO method utilising Chat GPT and a custom GPT. #chatgpt #aiseo #seo #searchengineoptimization #seotips #seotricks

♬ original sound – Marcus-Aurelius | SEO + AI

Dark Social as the hidden mirror of conversational SEO.

Dark Social has been a measurement blind spot for years: thousands of private conversations happening in WhatsApp, Telegram, DMs, closed groups — real people talking in places brands cannot track. 

In the LLM era, this “dark” space takes on new meaning. Not because AI can read private chats (it can’t), but because it learns from the public conversations that behave exactly like the private ones. The model doesn’t read WhatsApp, but it does read open communities where the same patterns emerge.

Brands aren’t losing visibility because they’re disappearing from search, they’re losing visibility because they’re disappearing from conversations. Between 47% and 65% of online brand/product conversations happen in Dark Social, and 49% of related web traffic originates from these private channels.

This is the new SEO: not winning positions, but winning meaning inside a recommendation..

So, as a brand, what can you do now?

Brands that want to exist in the era of generative models don’t need a new set of hacks. They need clarity, presence, and cultural consistency. They need to make sure that if someone asks in their kitchen, on their phone, or at work: “Who would you pick for this?” their name comes up naturally.

That means:

1. The brand must have a real, recognizable story. A proposition that isn’t interchangeable, a narrative that can survive being explained by others.
2. Content cannot exist for SEO; it must exist to answer real questions. Human questions, with human nuances, in human context. No empty articles, no replicated lists, no texts designed to please a machine.
3. Authority cannot be built solely on your own website: it has to be transversal. In interviews, collaborations, social networks, open projects, communities. Models learn from the diversity of signals, not from the repetition of a single voice.
4. Brands need to understand Dark Social 2.0: not as a darkness to illuminate, but as an indicator of the true depth of their cultural relevance.

For this, you can consult our AI Rank report or the Bank SEO report. 

The greatest irony of this revolution is that it brings us back to something that has always been true, but we had forgotten. Generative models don’t reward shortcuts; they reward meanings. They don’t recommend empty content; they recommend clarity. They don’t amplify noise; they amplify patterns.

For the first time in a long while, SEO stops being a collection of tricks and becomes a strategic discipline: what you contribute, how you exist, and what you represent to the people who mention you. This is not a technical change — it’s a cultural one. And the brands that will succeed are not the ones who understand the next algorithm, but the ones who understand something far more complex: their audience.

Because, at its core, the future of SEO isn’t about models, datasets, or metrics. It’s about returning to the oldest question in marketing: Why should anyone talk about you?

When you can answer that clearly, generative models take care of the rest.