I first pitched our new offer - an analysis of how visible brands are in AI-generated answers - to a fintech director whose company we have supported on martech and customer experience for years. It happened during a late-evening WhatsApp exchange. Two days later an email landed in my inbox. He had given it some thought and wrote that I was selling him a tool that would show him the scale of the problem without helping him solve it.
It was not the first time he had held up a mirror. Back when his team was rethinking software development, he told me: “You are not solving my problem when AI generates requests for new features. You are solving it when you help me build those features faster, more cheaply and with less complexity.” So they built an agentic software factory - a development setup in which people and agents build software together. His email changed how I looked at the business we had just launched ourselves.
On Monday, a HubSpot report landed in my inbox showing how AI assistants see FrodX. It highlighted three questions buyers ask ChatGPT and similar tools. We appeared in none of the answers.
One was roughly this: “We are a regional retailer operating across several countries, with physical stores and e-commerce. Which platform is the best fit for omnichannel customer engagement and loyalty communications?”
Our own sales director could have written that question. Yet the machine building the buyer’s shortlist did not mention FrodX. The same question had appeared in the report two weeks earlier. Gaps do not close themselves.
The report also identified 26 content opportunities in a single week: nine on LinkedIn, eight on Reddit and nine on our own channels. One week. One brand. One small market. And here is the punchline: we are the company selling this kind of analysis.
Since AI assistants began influencing buying decisions, the market has become crowded with consultants selling visibility in AI answers. Some call it GEO. Others call it AEO. A few are busy inventing new acronyms. Some of them are serious.
My own sample tells a different story. Almost every SEO consultant who pestered me with offers over the past three years has resurfaced in recent months - this time as a GEO or AEO consultant. The offer is almost identical. They have simply printed new business cards.
I will admit that we joined the rush too. As a HubSpot partner, we gained access to the relevant analysis tools, and the first interested clients came to us without much prompting. But the more of this work I do, the more convinced I become of something most diagnostic providers would rather not say out loud:
Telling a company where it is missing from AI answers will soon be free.
It will become part of pre-sales, much like a quick website audit is today. The tools already run the measurements automatically. I receive the analysis described above every week without an hour of manual work, together with suggestions for what we should publish next.
Diagnosis is becoming raw material. Value is moving to the real bottleneck: the ability to fill the gaps once they have been identified. Quickly, consistently, in several languages and at a volume no marketing team can sustain by hand.
Yuval Noah Harari often uses the example of the face of Jesus. There are more portraits of Jesus than of any other person in history, even though nobody knows what he actually looked like. Yet most people recognise roughly the same image. Not because of evidence. Because of repetition.
AI answers do not work in exactly the same way, but the outcome is similar. When an assistant assembles an answer about the best customer engagement platform, it builds the most plausible response it can from the sources and patterns available to it: your website, comparison pages, forums, media coverage, videos and everything else it can use. If your own website is the only source talking about you, you are barely present in that picture.
This is not a marginal channel. OpenAI reports more than 900 million weekly active ChatGPT users. A 2026 SparkToro analysis of Similarweb data found that roughly 68% of Google searches in the United States ended without a click to a result. A buyer can form an opinion before visiting your website or speaking to your sales team. For a marketing leader, the consequence is uncomfortably concrete: your brand is no longer only what you say about it. Increasingly, it is also the average of what the machine has found, understood and connected about you.
Back to my Monday report: 26 opportunities in seven days. The report from two weeks earlier contained 51. Assume that a strong article, comparison page or answer in a professional community takes a writer one working day. Addressing all 26 opportunities would create more than five weeks of work for one person in the space of a single week. Then Monday comes again, with a new report.
Even if you select only the ten best opportunities, you have just created two weeks of writing. Every week. Now add languages. We publish in Slovenian, Croatian and English. Then add formats: blog posts, LinkedIn, comparison pages, forums and professional communities. Another writer does not solve this equation. A different production model does.
The fintech director from the opening reached the same conclusion in software development: teams that code everything manually struggle to compete with teams that use agents as part of the development process. Content now follows the same logic. The volume required to earn representation in AI answers sits beyond purely manual production.
That is why we built what we internally call a content factory at FrodX: a loop that turns a signal into a published piece. The agents know my voice, our facts, our terminology, our proven examples and the phrases we refuse to use. From that context, they prepare drafts in three languages. Each draft passes fact and style checks. Approved copy is then pushed into HubSpot automatically.
I do not write inside this loop. I decide. I set the thesis. I approve or reject. I take responsibility and put my name on it. We are still tightening the final two links: generating content assignments directly from the visibility-gap report, and feeding the performance of each published piece back into the next cycle.
Make no mistake: this is not a machine for flooding the web with generic AI copy. Quite the opposite. Sources that are concrete, relevant, consistent and verifiable have a better chance of being useful and cited. A factory without style and fact gates is not a content factory. It is a junk factory.
A human still sits at our gates. No number is published without verification. And no text goes out if it could have been written by any AI with no knowledge of who we are.
To keep this honest, we gave our own factory 6.5 out of 10 in the latest internal review. Capacity is not the problem. We can already produce around fifty content units, including language and format variations, in a week.
The score reflects effort. Too many steps still have to be triggered and supervised manually. I am writing from the workshop, not the brochure.
The best marketing teams I know spend less and less of their time writing from scratch. Instead, they build the context from which a machine can write well: a base of facts and figures, terminology, brand voice, proven examples and clear rules about what must never be published.
Weak context produces generic content, however capable the model may be. Strong context means the hundredth piece still sounds like you. It remains consistent with the first ninety-nine. And it is concrete enough to become a source.
The human does not disappear. The role moves from production to thesis, judgement and accountability. Which position are we taking? Which story deserves to be published? Which figures are we prepared to make public? What are we willing to put our name to? That remains my job. Everything else is production. And production gets automated.
If you have spent the past ten years winning on Google, you have built a valuable asset. But that asset does not transfer into AI answers automatically. The diagnosis of where you are missing will soon be free. It will show you which questions your buyers ask that AI answers without mentioning your brand. It will also show you how many opportunities you left open this week. The truly uncomfortable question comes next:
Who will fill those gaps? A writer who can produce two pieces a week? Or a factory that can process fifty content units and variations?
The diagnosis will be free. The factory won’t.
P.S. A report without a factory is the most expensive kind of free. To see what a Monday like this could look like for your brand - and what I would do with the gaps it reveals - book 15 minutes with me.