How Conversational AIs Are Transforming Marketing and Brand Visibility

Emmanuel Dollé, CEO of Bubbling: decoding the invisible influence of conversational AIs on brands

Interview by Pascale Caron

In the digital economy, certain technological disruptions silently reshape value chains before companies even perceive their strategic consequences. After web search optimization, social networks, and then the algorithmic platform economy, another layer of intermediation is gradually taking hold: that of conversational artificial intelligence models.

For Emmanuel Dollé, CEO of Bubbling, this transformation is not merely a passing trend. It is already changing how consumers discover, evaluate, and select brands. A former executive at Renault, Google, and Meta, with experience in Paris, Dublin, New York, and several European markets, he now observes the emergence of a new strategic territory: GEO, or Generative Engine Optimization. His background gives him a rare perspective on the evolution of digital acquisition mechanisms, from traditional search optimization to generative conversational systems.

His analysis is structured around a strong conviction: large language models are now becoming a touchpoint in the purchasing journey. Unlike traditional search engines, they operate in relative opacity.

“Unlike Google or Meta, we don’t really know what happens inside them. They don’t communicate the information and sometimes don’t fully understand how the algorithm that produces the answers works themselves.” This opacity opens up as many risks as opportunities.

From an international corporate career to AI entrepreneurship

Emmanuel Dollé’s career path illustrates a typical trajectory of executives who have navigated several major technological cycles. His early career was built at Renault, where he spent a decade in B2B and B2C marketing roles, in France and internationally.

“When I left Renault, I was in charge of global marketing for the Mégane range.”

He describes this period as extremely formative. Post-war Croatia, Germany, Portugal, Paris headquarters: each position gave him the impression of “changing companies while staying within the same network.” But over time, another phenomenon emerged: intellectual stagnation.

“I realized that the learning curve and being challenged were extremely important to me.”

This constant quest for learning became a guiding thread throughout his career. Like many executives from large organizations, he also mentions the highly linear structure of internal career paths. This realization led him to join Google, after a lengthy series of interviews. There, he built the French automotive business before moving to Dublin to oversee Google France’s advertising operations. A few years later, he joined Meta to lead Southern Europe activities.

This experience represented a major turning point, in his view. He discovered an extremely international organization, a strong product culture, and especially a direct connection to the American innovation ecosystem. Israel, integrated into Meta’s Southern Europe scope, also gave him access to the world of high-growth tech startups.

Returning to France and questioning

For many international tech executives, the Covid period triggered a profound reassessment of personal and professional trajectories. While working again for Google in New York on global commercial strategy and product issues related to media agencies, Emmanuel Dollé returned to France for family reasons.

Settled in Saint-Paul-de-Vence, he experienced the difficulty of a hybrid model between Silicon Valley and the French Riviera. He then left Google with an existential question common among former Big Tech executives: how to transition from the comfort of large platforms to entrepreneurial uncertainty?

“After a career like mine, I was really afraid of launching and failing.”

So before launching his own company, he joined a French scale-up specialized in voice AI. This intermediate step allowed him to test his ability to operate in an entrepreneurial environment that was more unstable and less structured than American giants.

Bubbling: understanding what AIs actually say about brands

It was in this context that Bubbling emerged, founded in late 2025 with a partner from the financial data and tech entrepreneurship world.

The company’s positioning is based on a simple observation: companies no longer fully control how their image is constructed in generative AIs. For Emmanuel Dollé, mainstream LLMs, from ChatGPT to Gemini and Claude, are becoming decision-making intermediaries.

“The way I view mainstream LLMs is as a new touchpoint in the consumer purchase funnel.”

This progression follows an extremely rapid dynamic. In less than two years, conversational usage has shifted from an experimental phenomenon to a mass practice. According to figures mentioned in the interview, ChatGPT would now be approaching 900 million weekly users. Conversational interfaces are thus among the fastest-growing digital platforms in recent history. For Bubbling, this evolution undermines traditional digital marketing indicators. Companies must no longer only optimize their visibility in traditional search engines, but understand how generative AIs interpret, reformulate, and recommend their brands in increasingly complex conversations.

According to him, we are witnessing an extension of the historical search model, but with radically different dynamics. This shift is major: for years, the economic model of the web was precisely based on capturing and measuring traffic. Conversational AIs, on the other hand, tend to retain users in a closed environment.

“People will tend to have complete conversations within ChatGPT.” This shift challenges traditional marketing metrics.

“We’re seeing much less traffic to websites.” In other words, a brand can strongly influence a purchasing decision without the user even visiting its site.

A new economy of algorithmic influence

Bubbling’s technological core consists of massively simulating conversations between consumer profiles and different LLMs to analyze how brands appear in the responses. The approach relies on fine-grained modeling of marketing personas.

“We recreate the conversations of their target audience with the LLM.” The goal is to understand what prompts are actually used by customers and how AIs then structure their recommendations.

“We’re able to provide thousands of conversations about their customer target.” This logic transforms generative AIs into massive behavioral laboratories. The objective is not only to observe model responses, but to reconstruct the conversational logic that gradually shapes consumer preferences. Bubbling currently operates in a multi-LLM environment and simultaneously analyzes several conversational engines to identify perception gaps between platforms. This approach notably enables detection of narrative inconsistencies, recommendation biases, or reputational vulnerabilities invisible in traditional marketing tools.

The model’s scope extends far beyond simple conversational SEO. Bubbling actually seeks to build a dynamic map of algorithmic representations. In other words, understanding what AIs spontaneously associate with a brand when a user asks for advice, a product, or a recommendation. What attributes are associated with a brand? Which competitors emerge in responses? What criteria influence recommendations? And what biases appear in the models?

For Emmanuel Dollé, the challenge is both strategic and reputational.

“If, in a conversation about shampoo, Garnier isn’t mentioned, the person will probably buy L’Oréal, and Garnier won’t even know why it wasn’t chosen.” This decisional invisibility represents a major rupture, in his view.

GEO: beyond traditional SEO

Optimization for conversational AIs differs profoundly from traditional SEO. Traditional search engines relied on relatively explainable logic: backlinks, technical architecture, keywords, domain authority.

Generative models work differently. They aggregate complex, sometimes non-deterministic signals from multiple sources. This hybridization makes traditional strategies insufficient. Hence the emergence of GEO.

The idea is no longer just to be visible on Google, but to be correctly interpreted and recommended by conversational AIs. Bubbling thus seeks to identify gaps between the image desired by a company and the representation produced by the models. In some cases, results can reveal positive surprises. “Some companies possess genuine differentiating strengths without always measuring how positively they’re perceived.” In other situations, reputational risks become critical.

From conversational market research to algorithmic fact-checking

One of Bubbling’s most interesting aspects lies in the diversity of use cases. Beyond traditional marketing, Emmanuel Dollé mentions applications in market research, reputation management, information influence, or even detection of narrative interference.

He notably cites a project related to the hospitality sector around perimenopausal and menopausal women. The objective was to understand the expectations expressed by these customers through conversational AIs to build a tailored offering.

“Here are the words to use, and those not to use.”

The economic value is immediate: replacing certain heavy and costly qualitative studies with large-scale conversational simulations. But the most sensitive applications probably concern disinformation and reputation issues.

“We’re able to do fact-checking to see if LLM responses are biased.”

The interview also brought up another sensitive topic: Shadow AI.

The growing use of free generative AIs within companies now raises issues of information governance and strategic data protection. Behind everyday uses sometimes perceived as innocuous—document reformulation, content structuring, or writing assistance—some organizations are beginning to recognize potential risks related to unintentional exposure of confidential information in conversational models.

Throughout the discussion, this issue emerged as a particularly interesting avenue for the future evolution of solutions capable of analyzing responses produced by LLMs. These tools could gradually become reputational, informational, and security audit instruments for companies, potentially opening new application fields for players like Bubbling.

The transformation of marketing by generative AIs

Ultimately, the vision carried by Bubbling goes beyond the mere question of generative optimization. It announces a deeper transformation of the marketing discipline itself. For twenty years, companies learned to optimize their presence in search engines and then on social networks. They must now learn to exist in probabilistic conversational environments.

This transformation raises several strategic questions: how to structure a brand narrative intelligible to LLMs? How to control positioning across diverse models? How to detect algorithmic biases? How to protect confidential information integrated into generative tools? And how to evaluate their concrete impact on purchasing decisions? The challenge becomes even more complex as platforms evolve rapidly. Google now incorporates generative responses into its search engine. The boundaries between search engine, conversational assistant, and AI agent are becoming increasingly porous. In this context, companies risk losing part of their historical control over customer relationships.

SMEs, large corporations, and consulting firms: three key markets

Bubbling is currently structuring its development around three client types. First, SMEs and mid-sized companies with small marketing teams seeking to quickly understand their conversational visibility. Next, large corporations wanting to conduct advanced analyses on specific reputation, image, or competitive issues. Finally, agencies and consulting firms that can use data produced by Bubbling to enrich their own strategic recommendations.

This hybrid platform logic—both a technological tool and a strategic analysis layer—reflects a broader trend in the European AI ecosystem: offering verticalized solutions rather than frontally competing with large American models.

A still-underestimated transformation

The interview ultimately highlights a reality still poorly integrated by many organizations: generative AIs are no longer merely productivity tools. They are becoming intermediate cognitive infrastructures between brands and individuals.

This shift in informational power could profoundly transform digital strategies. The central question is no longer simply: “How to be visible on the Internet?” It becomes: How does an artificial intelligence talk about your company when no human is watching?

In this new conversational economy, lack of visibility can become invisible… until the moment when sales begin to decline without apparent explanation. For Emmanuel Dollé, this transition recalls the early days of web search optimization in the early 2000s. But the major difference is that LLMs no longer simply index information—they reformulate, prioritize, and sometimes synthesize it without complete transparency about their internal mechanisms. This evolution could shift part of brands’ influence power toward conversational systems, whose rules remain largely in flux.

For companies, the challenge is no longer just being visible on the Internet. They must now understand how artificial intelligences interpret, reformulate, and tell their brand story on their behalf.