Related Articles
View AllDecember 15, 2025
GEO: The shift from linking to talking
Miguel Pereira is co-founder and Executive President of Darwin & Verne, a Spanish brand consultancy and creative agency with an ecosystem of solutions designed to connect brands with audiences, including a business unit specializing in AI solutions applied to marketing. He also serves on the ICOM Board, where he leads the Innovation practice, helping independent agencies harness new technologies to drive growth and creativity.
Lucía, 31, is organizing a weekend getaway to Portugal with friends. She used to type into Google: “cheap Porto weekend ideas” and get lost in blogs, links, and three-year-old comparisons. Today she opens ChatGPT in voice mode and says: “I want a weekend in Porto with winery visits, viewpoints that aren’t overcrowded, a good coffee spot, and a budget of €150 per person, and give me alternative plans in case it rains.”
Within seconds, she receives a living itinerary, with routes, schedules, and a checklist. Then she follows up: “Any historic bookstore? A place to watch the sunset?” and the conversation sharpens the plan.
It’s not that Google no longer works; it’s that, for inspiration, the first stop has shifted from the link to the dialogue.
The data confirms what users already know
Beyond discrepancies across different data sources, one thing is clear: real user habits are changing. McKinsey speaks of a “new gateway to the internet” and notes that “a substantial share of consumers already use AI search,” and that this behavior is beginning to influence how people discover and decide. As a result, many websites —particularly publishers— are experiencing significant drops in traffic.
In parallel, usage data from ChatGPT and other assistants shows that conversational queries are rising, and a relevant percentage of users now start their search there. Adobe captured this in its 2025 survey among ChatGPT users; some questioned sampling biases and population reach, but the trend is unmistakable. The Wall Street Journal, citing Datos, also points to double-digit year-over-year growth in desktop traffic to AI assistants (ChatGPT, Perplexity…).
This isn’t a sorpasso — it’s a gradual transfer. A trend.
But here comes the nuance that matters for the marketing industry. As Enrique Pérez of Digital Angels clarifies:
AI wins at the top of the funnel; Google still rules at the bottom.
Indeed, for generic, inspirational, or general-information queries, the conversational format reduces friction and solves questions in a single dialogue, without leaving the tool’s environment. Behavioral studies (like Pew Research’s 2025 analysis) already suggested that when generative synthesis is part of the experience, users click less on external sites: if the answer satisfies, they stay where they are.
Conversely, the closer we get to purchase, the less the habit changes: users ask for direct links, compare prices and logistics. In that terrain, traffic referred by LLMs remains marginal and converts worse than organic or affiliate traffic.
Simple translation: conversation captures inspiration; links govern the transaction.
Voice makes the conversation feel human
Another reason for the rise of conversational formats is voice. Classic voice search never exploded as predicted, but conversational usage skyrockets when voice is paired with LLMs. In this context, ChatGPT’s voice experience (asking, re-asking, requesting nuance) normalizes the habit of “asking the machine” in natural language because it resembles talking to a friend. How many of you have “talked” with ChatGPT about any random doubt recently?
To understand why this experience is so convincing, it’s worth looking at how LLMs work underneath. Beyond “writing,” many assistants combine generation with real-time semantic search to pull in relevant context and citable sources. Tools like exa.ai act as a radar: they scan the web with vector-based queries, prioritize relevant and recent pages, and return data that the assistant reads, summarizes, and links. This bridge between generation and retrieval reduces the “black box” effect and turns the chat into an experience anchored in reliable sources.
But what does this mean for brands?
From ranking to being cited
It’s simple, and critical: the old game of ranking with SEO is no longer enough; you need to be citable, as is often said in the world of GEO (Generative Engine Optimization).
The question stops being “what position do I appear in?” and becomes “who cites me, and with what authority?”
In GEO, what matters is your presence inside the answer: that the model incorporates you, cites you, recommends you —and does so based on high-quality sources.
The practical approach for brands comes down to three ideas:
- Be citable: citable-first content on your own site, with clear definitions, useful comparisons, relevant FAQs, visible update dates, and authorship.
- Build authority beyond your site: presence in key vertical media, wikis/directories, expert repositories, UGC forums, and comparison sites that models consult and link to.
- Think by engine, language, and country: ChatGPT, Gemini, and Perplexity draw from different sources and weight them differently; paraphrasing in your content changes results.
It’s worth remembering something that is becoming obvious in the GEO world (still called SEO of “Search Everywhere” by many). The fundamentals are nearly identical to SEO: most brand mentions and citations go to those who already have strong SEO, because models still rely on the same pillars.
The equation hasn’t changed much: quality content —now also adapted to how AIs consume and process information— and external authority, meaning digital PR and strong linkbuilding. In the end, LLMs search the internet, and they tend to trust those who already have enough authority to be considered.
How we’re helping brands stay ahead
At Darwin & Verne, we’ve decided to get ahead. From Phileas (our AI specialist unit) and with Digital Angels as partner, we’ve structured a GEO service for brands that don’t want to wait for the adoption curve to run them over. The service has three layers: monitoring (with tools), expert consulting, and content/PR activation.
For new clients, we run a structured onboarding (2–3 weeks) with objectives and markets, tool set-up, and dashboards. For agency clients, we integrate naturally into their account and content teams: we extend their plan with a GEO layer that measures, interprets, and activates, supported by tools and expert knowledge.
Our goal is simple to state (and not so easy to execute): that the next time someone asks ChatGPT about your category, your brand appears in the answer.
Managing expectations is also important. In SEO there’s a gradient: from position 20 to 1, every step adds visibility. In LLMs, that scale doesn’t exist; it’s far more binary. For a given prompt, you’re either there or you’re not —you get cited, or another brand does.
Additionally, mentions and citations can vary depending on the user’s own history and conversational context, where personalization has much more weight than in classic search. That’s why no team can guarantee appearing in a model’s answers; what they can do is maximize the odds, as in SEO, but with more uncertainty and fewer intermediate positions to measure progress.
The clear conclusion: GEO doesn’t replace SEO —it elevates it. Because in this new battle for attention, the truth is not the ranking; it’s what gets cited.
Enjoy Porto, Lucía.
