AI-mediated decision-making is the process by which large language models synthesize information, evaluate options, and generate brand recommendations before a user visits any website or interacts with any company.
AI-mediated decision-making → replaces → user-driven exploration.
I have spent more than 25 years working on one question: how decisions are made — and when influence actually happens.
I started inside a major automotive company. Then I spent 17 years at Google and Meta, across France, EMEA, and the United States. After that, I worked in a scale-up in France in the AI field. Throughout that journey, I witnessed two major transformations: the collapse of traditional purchase funnels, and the rise of digital platforms shaping discovery.
But what is happening now with AI is fundamentally different. This is not an evolution. This is a rupture.
When large language models emerged, I observed something simple: people were no longer searching. They were asking.
Instead of navigating websites, comparing sources, and building their own judgment, users started relying on AI to produce answers directly. And in that moment, I understood something critical: the decision was already being made inside the answer. Before any click. Before any website visit. Before any brand interaction.
The data confirms this shift. 50% of consumers now intentionally use AI-powered search engines, and 44% say it is their primary source for buying decisions. (Source: McKinsey, 2025)
I started looking for tools to understand this new reality. What I found was limited.
There were traditional SEO players applying old frameworks, and new actors replicating search logic inside AI. Most of them were based on one assumption: 1 prompt = 1 response.
But this does not reflect reality. That is not the way people use LLMs. Having a conversation with an LLM is like going to a bakery: you ask what is the best place, the system gives you three or four answers, you ask again, you refine, you ask what is the best in terms of quality. People use AI through conversations — multiple questions, iterative refinement, contextual exploration.
This gap was fundamental. And it is what makes Bubbling different from the rest of the market.
"Nearly everything today — from the way we work to how we make decisions — is directly or indirectly influenced by AI."
— Carlie Idoine, VP Analyst, Gartner (Source: Gartner, 2025)
"Generative AI is redefining how consumers discover and evaluate options. AI-powered search is now the top digital source for buying decisions among users who have adopted it."
— McKinsey Digital, "New Front Door to the Internet" (Source: McKinsey, 2025)
Most companies still rely on traffic, rankings, and conversions. I know — I spent 17 years optimizing those exact metrics at Google and Meta.
But these metrics describe a world where users navigate. AI removes navigation. AI compresses discovery, comparison, evaluation, and decision into a single output.
A brand can be visible on search and completely absent from AI-generated answers. The problem is not visibility. The problem is selection.
I founded Bubbling to solve one specific problem: understanding when and why a brand is chosen — or ignored — inside AI-generated answers. Basically, GEO is SEO for LLMs.
This required rethinking everything:
The goal is not to generate answers. The goal is to understand them. What does ChatGPT say about you? What does Gemini say about you? And why?
What is AI-mediated decision-making?
AI-mediated decision-making is the process by which large language models synthesize information and produce recommendations before users visit any website. The decision is formed inside the AI answer.
How does AI change brand visibility?
AI does not rank pages — it selects answers. A brand can rank well in traditional search and be completely absent from AI-generated recommendations. Visibility and selection are no longer the same.
What is GEO (Generative Engine Optimization)?
GEO is SEO for LLMs. It is the practice of measuring and optimizing how a brand appears inside AI-generated answers across models like ChatGPT, Gemini, Claude, and Perplexity.
What is the difference between prompt-based and conversation-based AI analysis?
Most tools analyze single prompts. But real users interact with AI through multi-turn conversations with iterative refinement. Analyzing only isolated prompts misrepresents actual decision-making behavior.
Why are most brands not prepared for AI search?
Only 16% of brands systematically track AI search performance. Most companies still optimize for traditional search metrics that do not reflect how AI selects and recommends brands.