Three years ago, Francisco Vigo went to Google looking for training shoes and could not recognize a single brand on the page. Around half of the ten blue links were ads, with so many advertisers bidding on the keyword that the result was useless. He put the same question to ChatGPT, then still on GPT-3.5. It knew who he was, and it answered with one clean recommendation for exactly the product he wanted.

That contrast has since become a market. Consumers have moved product research into AI assistants, and brands have lost the ability to see, let alone shape, how those assistants describe them. Vigo is co-founder and CEO of geoSurge, a London company helping enterprise brands measure and improve their visibility inside large language models. In this Founders on the Record interview with Dedale Intelligence, he explains why customers no longer search, how geoSurge splits the problem into a retrieval layer and a model-memory layer, why it gave its tracking product away, and what investors should ask about portfolio companies whose organic traffic has collapsed.

From Ten Blue Links to One Clean Answer

"That was my aha moment," Vigo says of the training shoes query. "I realized that the whole world was about to adopt this technology. Every brand, product, service, and company would need to measure and understand this new channel, and would want to be present in it." He frames the stakes in a way any founder recognizes: "Imagine how you would feel if someone asked ChatGPT for the best business podcast and you appeared in the answer."

What made it a company rather than an observation was the difficulty. Solving it needed "technology, ML, data, and an intersection of skills." As he puts it, "that combination is very rare."

A Lost Contract and a Game of Magic: The Gathering

The founding story is less tidy than the thesis. Vigo had left his role leading the data science team at a London fintech unicorn and was bootstrapping, selling a product to banks in Latin America. On the day he was due to sign, the client pulled out. "It was a very hard day, so I went to play to think about something else."

The game was Magic: The Gathering, in a London pub. His opponent that night is now his co-founder, a programmer Vigo calls the most gifted he has ever met, who built the second-largest functional programming consultancy in the world and who "has never had a job." Both were having a bad month. Dinner after the tournament turned into the conversation that led to geoSurge.

Customers Do Not Search, They Ask

Vigo's summary of the past three years is blunt. "I usually say that customers do not search; they ask. This has been the most fundamental change we have seen in customer behavior." His own buying is the evidence. geoSurge doubled its headcount in three months and needed HR software for global contractors, so he used ChatGPT's deep research feature, got a shortlist of four, and picked one.

Five years ago, marketing measurement meant a Google Analytics console and a Facebook Ads Manager, with Google holding 92% market share. "I am not the regulator, but that sounds like a monopoly to me." AI is structuring itself differently. ChatGPT competes with Claude, Gemini, open-source models, Grok, and now Muse. Vigo, an economist by training, thinks the competition is healthy, but it fragments the problem. "What ChatGPT thinks about your brand might be different from what an open-source model thinks," he says. "Your visibility on Claude might also differ from your visibility on Gemini."

The deeper issue is that the old scoreboard is gone. On Google there was a results page and a rank: fifth for "credit cards," say. "LLMs do not have an index. They are probabilistic, with trillions of parameters." Scaling laws explain why the models keep getting smarter, but not what any one of them believes. "Nobody can tell you why a model has learned what it has, or why it recommends a particular brand."

The Text Paradox in a Video-First Decade

For five years, brands moved budget into short video. Then language models arrived. "ChatGPT is a large language model, so it learns from language," Vigo says. "The paradox is that the companies that kept updating their blogs, kept writing, and had a digital footprint are the ones that ChatGPT learned about when it crawled the entire web."

He is careful not to write video off. When a model cites a YouTube source it is not watching anything: Gemini pulls the transcript through the YouTube API and reads the text, because it has to answer in under a second. Keep the aesthetics and the hooks, Vigo says, but "understand that there is a new audience using that content, and that this audience is using its text and script."

Budget ownership tends to get settled the hard way. "The CEO is on the sofa on a Tuesday night and asks ChatGPT for the best providers of whatever the company does. The brand is not in the answer, so the CEO calls the CMO and asks why the company is not showing up." The work lands with whoever maintained the website or ran SEO, a community Vigo says was initially terrified. He sees an opening instead: a chance to "stop being isolated in the SEO department, and to grow into the head of traffic."

Explore more founder interviews and market intelligence in the Dedale Intelligence article library.

Two Layers: Retrieval and Model Memory

Plenty of vendors run prompt tracking: ask Gemini, ChatGPT, and Claude the same questions, parse the answers, compute a visibility rate. geoSurge does it too, and gives it away. Visibility, Vigo says, "is like a thermometer of how your brand is positioned. The question is what you do about it and how you improve."

geoSurge splits the answer in two. The first is the retrieval layer, closest to classic SEO: the model writes a fan-out query, finds sources, and builds an answer from them, so the play is to win placement in those sources. "That is what most of the market is doing right now." The second is how models learn in the first place, where geoSurge puts its research effort and its proprietary technology, corpus engineering: creating training data good enough that the labs' models absorb a correct picture of the brand. "In my view, data is becoming the most important marketing channel in the history of humanity," Vigo says. Put simply: "There are two ways to think about AI visibility: being found by AI, and teaching AI about your brand."

In practice the work often starts lower down. "We go to large enterprises, and I am talking about companies that are listed on stock markets. When we look at their websites, the main thing we see is that they are blocking ChatGPT's crawlers. They come to us saying that their AI visibility is low, and I tell them that they are not allowing ChatGPT onto their website."

Why geoSurge Gave Tracking Away

The free tier is a deliberate read on where the value is not. "We always knew that tracking was not the hard problem to solve," Vigo says. Strong teams were already coding visibility dashboards themselves in a few days. "That is when I realized that the tracking layer was going to be commoditized: a race to the bottom, with no defensibility and no moat."

geoSurge still invests in the analytics product, including per-model granularity, but does not charge for it. Against the sub-$100 tools in the market, Vigo draws a sharp line: "At geoSurge, we are in the business of solving the problem, not of showing you that you have one."

He is equally firm about staying a software company. geoSurge has turned down large consulting contracts, and treats agencies as partners, many using the platform as the backbone for the GEO services they sell on. On whether the category stays fragmented, he reaches for the ten principles of good design: less is more. "You just need the best tool."

Agentic Commerce, and What Investors Should Not Miss

As Vigo reads the Cloudflare Radar data, only around 30% of web traffic now comes from humans, with the rest from agents, crawlers, and bots. He expects agentic commerce to arrive because consumers will want it, in the same way they eventually trusted putting a card online once Apple Pay made it safe. ARK Invest's Big Ideas 2026 report puts agentic commerce ahead of traditional discovery within about two years.

This is where the memory layer pays off. geoSurge has published research showing that brands held in a model's memory are searched three times more often than those that are not, and the mechanism is simple once stated: before ChatGPT searches, it writes its own query, based on what it already knows. "If the model does not know about your brand, it will never write a search query that includes your brand. Therefore, even the dynamic layer, the retrieval layer, is going to be skewed toward your competitors."

His message to investors cuts against the SaaSpocalypse reflex. geoSurge could build its own CRM and pays HubSpot roughly $25 per seat instead, because that is not its business. The more urgent point for funds is operational: several private equity firms have already partnered with geoSurge because a portfolio company that relied on organic traffic watched it fall away.

Not a Feature, a Channel

Asked whether GEO survives as its own layer, Vigo answers by pointing at the room. The studio he is being filmed in was equipped with products discovered and compared using generative AI, and nobody involved plans to stop.

"I personally think that this is going to become the most important marketing channel in the history of humanity," he says, because of the intimacy of the relationship and the context these systems hold about us. "In my view, it is the best recommendation engine ever created." Today, humans and their assistants decide together. He expects that to tip, and his test case is flights: find the most convenient one, check the calendar, confirm there is a train to the airport. "I just want to know that my agent took care of it."

For more on how AI is reshaping software markets and go-to-market motions, explore the Dedale article library and the October 2026 Software Market Trends report, or read our previous Founders on the Record interview with Lightbringer on AI-native patents.

Contact Dedale to learn more about our research coverage.

This interview was conducted by Fu Ye. Connect with Francisco Vigo on LinkedIn to follow geoSurge's progress.