SEO Pipeline
AI-generated content is everywhere, and it all starts to sound the same. Worse, it sometimes makes things up about your business, like a warranty that doesn’t exist or a price you never published. So I built an AI SEO pipeline that tackles the problem from the other end. It starts from what your customers actually ask, it keeps a source for every claim, and it measures the effect of the content every month.
Behind this project are more than 15 years of web development and more than 20 years in sales and marketing, including many years in call centers training hundreds of agents and writing sales scripts. That work left me with one conviction: before you speak, you listen.
The first problem is well known: generic content. When everyone starts from the same keywords and asks the same AI to write “a complete article,” everyone gets the same text. It answers a query, but not the person typing it.
The second problem is more serious, because it touches your reputation: invented claims. In a real FAQ generation test, almost one answer in five contained an invented fact about the company. It wasn’t a style issue. It was a promise the company had never made.
The third problem changes the rules. Google, ChatGPT and other assistants now answer questions directly. The question is no longer only “am I on page one?” but also “am I the source the AI cites when it answers my customers?” To be cited, content has to answer real questions, directly, with verifiable facts.
1. Understand real demand. Before choosing a keyword, the pipeline collects what people actually say: the questions they ask Google, the answers Google shows them, the conversations they have with each other on forums. It keeps their exact words, and it takes the city into account, because the same question doesn’t carry the same weight from one market to the next. That’s the foundation for understanding what your customers really want.
2. Gather the evidence. Everything collected goes into an evidence file where each item keeps its original text and its source. Nothing is rewritten. That’s what makes it possible to produce content where every claim has a source, and to bring invented claims down from almost one answer in five to fewer than one in a hundred.
3. Turn demand into personas. From that evidence, the pipeline builds portraits of typical customers. Every persona trait points back to the real words that justify it. These are synthetic personas grounded in real verbatims, not archetypes invented in a meeting room.
4. Produce, verify, measure. FAQs, article ideas and articles grow out of this material. Before publication, an automated check blocks any forbidden claim. After publication, a monthly report built from Search Console and Google Analytics 4 shows what moved. That’s the loop to produce and measure content every month.
Three principles guide every step:
It doesn’t promise page one. Nobody can honestly guarantee a ranking, and Google’s rules change without notice: I’ve seen its results change shape twice in a few weeks. What the pipeline guarantees is a method, evidence and measurement.
Nor does it replace your knowledge of your own business. It shows you what your customers ask and what your website actually promises; you decide what you want to say. And nothing is published without your approval.
The pipeline was designed for businesses with several locations: franchise networks, clinics, branches. They need content that speaks to each local market without saying just anything, and reports their leadership can read. It was proven on a residential renovation franchise network in North America, then in a completely different industry, in French.
It’s also for SEO agencies and consultants who want large-scale research capacity without sacrificing rigor, for example to stand out from the generic content their clients see everywhere.
The starting point is a demand study of your market:
Optionally, I can then support content production and deliver the monthly report that measures its effect. Every engagement is different: pricing is set by quote, based on the number of markets and the scope of the study.
In a little over five months, working alone with an AI coding agent, I wrote several hundred scripts and made more than 750 commits. It taught me one thing: AI is a brilliant executor and a lazy strategist. Give it the frame, and it writes and scales remarkably well. Let it find the frame, and it picks the most ordinary answer. So the pipeline is first and foremost a frame. The mistakes, incidents and lessons of building it are told in the behind-the-scenes story of the pipeline.
In the call center, I learned that a good script doesn’t start with the pitch, but with listening. The pipeline works the same way: we start by listening to what your customers are already saying, before writing a single line. Write to me to talk about your market, and let’s see what your customers really ask.