AI Architecture: I build what makes a brand visible when AI answers
I don’t do AI. I build the infrastructure that makes a brand visible at the moment AI systems are the ones answering. The work presented here forms a single ecosystem, organized around one area of expertise: being cited by answer engines.
An AI only cites a brand if someone has built what makes it citable, then what verifies that it is. That architect's work is what I do.
One area of expertise: being cited by AI
Getting cited by ChatGPT or Google’s AI Overviews isn’t about a trick. First, you need content an answer engine can reuse without distorting it. Next, you need to know whether it actually gets reused, and against which competitors. Finally, the site hosting it must be as readable for machines as it is for people.
Each of my projects addresses one of those conditions, which is what makes them a whole rather than a collection of projects.
These are proven in-house systems, not products I sell. I can, however, adapt and deploy them for a business as part of an engagement.
BlotMKT: the lab behind a GEO architecture
BlotMKT is the lab where I designed and tested a complete GEO architecture. Its purpose was to answer a precise question: can you automate the production of French-language content that AI systems will cite, without losing what makes it citable?
The pipeline runs from content generation to distribution, by way of structuring. What sets it apart is that it is designed backward, starting from the citation. Each step asks what will lead a generative engine to reuse a passage, and the shape of the content follows from that.
On the technical side, the lab runs on Python and FastAPI, deployed on Google Cloud, with ChromaDB and LlamaIndex to organize the knowledge the pipeline draws on. The pipeline details remain private. I’ll present them on expertise.antoine-blot.com, the space I’m preparing for this kind of work.
BlotMKT isn’t meant to become a product, and its value isn’t measured in article counts. This test bed proved one thing: an automated GEO pipeline can be built quickly and works, as long as you navigate between the algorithmic raindrops, that is, anticipate what engines reward without falling into what they penalize.
AI citation tracking: a blind spot turned into a metric
Most businesses don’t know whether they appear in generated answers. Their dashboards track rankings and traffic, but nothing of what is said about them in ChatGPT or Google’s AI Overviews. That’s a blind spot, even though a growing share of their clients start right there.
I designed a measurement method and the tool that applies it. The tool records a brand’s citations and its share of voice against competitors in AI Overviews, Google’s AI Mode and ChatGPT. It also estimates the share of traffic coming from these engines. These metrics go into the monthly reports, alongside standard search data.
This project complements the first. Producing citable content only makes sense if you can verify that it gets cited, and measurement turns AI visibility into a metric leadership can track and act on. I explain why this measurement matters on my page about GEO.
This website: the method applied to myself
The page you’re reading runs on a system I designed to be read and cited by AI. I didn’t adapt an existing CMS; I built my own, drawing on AI research I’ve been following since June 2025.
Every architectural choice targets machine readability. Pages are served as static HTML, which bots read without having to execute code. Every piece of content also exists in a clean text version, and an llms.txt file tells language models what the site contains. Structured data describes the same entity from page to page, the author and his services, while an entity map links each topic to the passages that document it.
That’s the argument I like best: I don’t just recommend a method, I live it. What I advise my clients to do to get cited, this site does first.
In service of SEO and marketing
These systems aren’t a separate line of business. They extend my work as an SEO consultant, where AI citation is one dimension of search, and my role as a marketing director, where visibility must be measured to inform decisions. AI architecture is where those two fields meet.
Frequently asked questions
Do you build chatbots or AI assistants?
No. I don't build conversational assistants. I build the infrastructure that makes your brand visible in the answers of existing AI systems, along with the tools that measure that visibility.
Can I use your tools for my business?
They are proven systems, not off-the-shelf products. I adapt and deploy them for a business as part of an engagement, based on its market and goals.
Why a custom CMS rather than WordPress?
A general-purpose CMS wasn't designed to be read by language models. By building my own, I could decide every detail that matters for citation, instead of stacking plugins to get close.
Let’s talk
Facing an AI visibility challenge that calls for a custom system? Let’s talk.