A new French solution aims to fill the blind spot of SEO in the age of generative AI.
For two decades, measuring a brand's visibility meant monitoring its Google ranking. A few tools dominated the market, methodologies were well-known, and KPIs were widely shared. But since the emergence of ChatGPT, Perplexity, Gemini, and Grok in everyday use, a question that no one was asking three years ago is now being discussed in executive committees: What do generative AIs say when we talk to them about us?
Most tools on the market remain silent on this point. While established SEO platforms have started adding AI monitoring modules here and there, their logic remains based on that of traditional SERPs. It is precisely this blind spot that [this tool] aims to illuminate. Cockpyt AI, a young French solution that has entered the MVP phase, designed from the outset to measure a brand's visibility in the responses of major language models. We took a closer look at the product.
A new territory to measure: the GEO
The subject now has a name: Generative Engine Optimization or GEO. The term refers to all the practices aimed at optimizing the presence of a brand, product, or service in the responses generated by LLMs. While traditional SEO targets search engine results pages, GEO directly targets the sentences produced by AI. A brand can very well dominate the first page of Google for a query, and yet never be mentioned by ChatGPT when a user asks the same question. The reverse is also true.
The problem is that this new visibility is technically difficult to measure. LLMs are non-deterministic The same query can produce different responses from one call to another. They dynamically query the web via embedded search engines, triggering invisible subqueries (the famous fan-out queries), and sometimes hallucinate sources that do not exist. Seriously measuring a brand's visibility in this environment requires a radically different methodological approach than keyword tracking.
This observation is what motivated the creation of Cockpyt AI.
Cockpyt AI: a tool designed GEO-first
Cockpyt AI is a SaaS solution developed by a French co-founder duo: an SEO/GEO expert on the product and strategy side, and a developer on the technical side. The project is driven by a clear conviction: to add a tab “AI tracking” Using an existing SEO tool is not enough. Geo deserves a native tool, designed around its specific constraints.
“Measuring visibility in LLMs is not the same as measuring a position in a SERP. You have to test each prompt multiple times, on several models, with several variations, and cross-reference the results. It’s a different approach.”explains Florian Zorgnotti, one of the co-founders.
The product targets three profiles: the SEO agencies who want to expand their offering to include GEO, the GEO freelancers who need a dedicated tool for their missions, and the ambitious SMEs who want to manage their visibility in the new search engines without hiring a full-time agency. The positioning is deliberately narrow: Cockpyt AI does not seek to compete with general SEO suites, but to become the benchmark in a specific segment.
The product is organized around four pillars, each of which deserves closer attention.
1. The Cockpyt Score: a unique indicator to manage your GEO visibility
Rather than overwhelming the user with a dozen metrics, Cockpyt AI has opted for a single synthetic score, Called Cockpyt ScoreIts formula combines three dimensions:
- la activation of the brand in the response (40%)
- la position of the brand in the list of brands mentioned (40%)
- the presence of a contact or a link to the official website (20%)
A notable feature: a blocking rule The score is reset to zero if the brand is not present at all in the response. This choice may seem drastic, but it avoids a common pitfall of modern SEO dashboards: displaying flattering scores when the brand is simply invisible. It's a deliberate choice in favor of clarity and honest measurement.
In addition to this main score, the platform calculates a Visibility Score (ratio between the number of mentions and the number of tests performed) and a SOV Global which weighs mentions and URL presence to give an overall view of share of voice.
2. Multi-LLM Deep Scan: 7 tests per prompt
The technical core of Cockpyt AI is called the Deep ScanRather than sending a prompt once on a model and considering the response as the truth, the tool executes seven behavioral tests per prompt, by playing with temperature and seed parameters to explore the natural variability of LLM.
Each scan actually triggers a sequence of five steps:
- Automatic generation of onboarding prompts based on brand information
- Variant generation to test different formulations
- Execution of the 7 behavioral tests with active web search
- Batch extraction key elements (mentions, positions, sources, competitors cited)
- Hallucination testing on the sources returned by the model
This multi-test approach makes it possible to produce indicators statistically robust, which withstand the inherent uncertainty of LLM. For local brands, a field city is injected into the first prompt and propagates naturally through the chain, an attention to the use case that betrays a true knowledge of the business.
The fan-out queriesThese subqueries, which the models trigger in the background to enrich their response, are extracted directly from the API call metadata, at no extra cost to the user. They are currently visible in the detailed view of each prompt and will be the subject of a dedicated section in a future version of the product.
3. Competitive analysis: who is stealing your citations
Knowing you're invisible is useful. Knowing who takes the place The market share we'd like to occupy is exploitable. Cockpyt AI offers two competitive analysis tools directly within the dashboard for this purpose:
- Le SOV Global measures the share of voice of each actor across all tracked prompts
- Le Threat Radar identifies the competitors who most consistently dominate the responses, prompt by prompt
For each prompt analyzed, the user accesses a detailed view that combines three dimensions: the positions occupied by its brand and its competitors, the competitors mentioned with their frequency of occurrence, and the sources that the LLM actually consulted to produce its response. This last piece of information is probably the most valuable: it literally says where to go to work on your content to fall within the scope of the model.
A bar graph visually compares visibilities, and a tech-style radar signals detected hallucinations with an immediately readable OK/KO code.
4. GA4 integration: combining AI visibility and real traffic
This is undoubtedly the product's strongest differentiator. While most GEO tools stop at measuring visibility in responses, Cockpyt AI connect directly to Google Analytics 4 via an OAuth connection, and identifies traffic originating from large chatbots: chatgpt.com, perplexity.ai, gemini.google.com, grok.comA daily CRON job retrieves the data into a dedicated table and cross-references it with the visibility scores.
In concrete terms, the user not only sees si His brand is cited by AI, he sees How many visits? This generates results on their website. For GEO, it's the equivalent of the integration of ranking and analytics data that every SEO expert has mastered for the past ten years at Google.
Where is Cockpyt AI headed?
The team is already publicly documenting upcoming projects, drawing in particular on the RESONEO study on variations between ChatGPT versions (5.3 vs. 5.4). Several advanced metrics are announced for V2:
- Le "Bigfoot Effect", a monitoring of the sometimes sharp drop in the number of domains cited from one response to another
- Un Concentration Score measuring the degree to which a few actors monopolize citations
- A dashboard Trusted Domain Authority identifying recurring domains in LLM sources
- Un monitoring by model version (for example, GPT-5.3 vs. 5.4) to track how a model update reconfigures the visibility landscape
- A distinctive conceptual framework parametric visibility (from training) and dynamic visibility (from web search)
- A section Model Impact in the dashboard to compare the performance of a brand according to model variants
- Le Survival Rate, a metric that will measure the persistence of a mention over the course of tests
The ambition is clear: to make Cockpyt AI the GEO visibility benchmark barometercapable of keeping up with the rapid evolution of an ecosystem that changes with each new model release.
To remember
Cockpyt AI arrives in a market that is still relatively unequipped, at a pivotal moment when brands are beginning to concretely measure the effects of zero-click search powered by generative AI. The product has three main merits: a serious methodology (multi-tests, multi-models, hallucination verification), a strong differentiator with GA4 integration which brings the measurement back to the reality of traffic, and a readable product philosophy which prioritizes usability over display performance.
The product is currently in its MVP phase. For SEO agencies looking to get ahead of GEO, freelancers seeking a dedicated tool to offer on assignment, or brands starting to wonder why their organic traffic is declining despite not having changed their SEO strategy, this is a project to keep an eye on.
