Practitioners whose job it is to manage brand visibility have a problem the market hasn’t fully solved yet. AI platforms are forming opinions about your brand right now, answering questions your potential customers are asking, citing some sources and ignoring others, framing some brands favorably and others less so. The data behind those responses exists. It lives inside these systems. Getting reliable, actionable access to it remains the open problem.
I’ve watched this exact dynamic before - from inside. I was part of the team that built Bing Webmaster Tools, at a moment when businesses had the same relationship with search engine data that brands have with LLM data today. The model that worked then is the one I’m applying now.
Today I’m launching CitationIQ, and it would mean the world to me if you’d help me spread the word. It’s a platform that shows you how AI sees your brand, and not in the surface way you’re expecting. It answers questions about your AI visibility that I genuinely haven’t seen anyone else ask, let alone answer, and it does it whether you’re in-house on your own site or an agency managing a roster of clients. More on exactly what that means in a minute.
Here’s why I’m the person who built it.
Learning to operate without a rulebook
I started in SEO during a period when the industry was still making up its own rules, and nowhere was that more true than online gambling. Real stakes, fast feedback loops, no patience for slow results - and no guidance from the search engines themselves. You compared notes with others who were figuring it out alongside you, or you stayed blind. Those early years burned in a simple truth: without data, you’re guessing. And guessing is expensive.
That period built something in me I’ve never fully been able to explain to people who came up a different way. It wasn’t recklessness. It was a willingness to treat “we don’t know yet” as a working condition rather than a reason to stop.
The room where the data lived
When Microsoft called, I moved to running SEO at MSN, then went to Bing when MSN Search became Bing. What I found there was a familiar problem at a different scale: businesses that should have been engaging with Bing had no reliable way to understand what the engine was doing with their sites. The data existed. It lived in our systems. There just wasn’t an organized way to get it all to the people who needed it.
What we built to solve that was the 2nd gen Bing Webmaster Tools. It became the single largest driver of new Bing Advertising accounts at the time and contributed to moving Bing from operating in the red, to the black. The model was simple: open the data, make it readable, give businesses what they need to make decisions. During the same period I helped launch Schema.org - in hindsight, we were building the vocabulary the internet would use to describe itself to machines we hadn’t even begun to conceive. H/T to the many brilliant minds involved in that project around the world!
After Microsoft, I ran operations at Bruce Clay Inc., then joined Yext just ahead of their IPO - a VP role that put me inside operations at a scale most practitioners only dream about. Then I was part of a reduction in force. New ideas came in. They didn’t include me.
What you do when the industry doesn’t call back
The industry I spent 20 years helping build went quiet so I stopped waiting for a callback and started building instead. What I did with that silence was simply to stop waiting for permission. (The podcast for this article has more details.)
And that’s when something clicked. The LLM landscape felt very familiar. Brands without clear visibility into how they’re being cited, data living inside systems that businesses can’t reach, no officially organized way to get it to the people who need it.
We have been here before.
One thing to be clear about, because practitioners will ask: there’s no inside access involved in what I provide. No one has that. This is all data that’s openly available if you know where it is, what it means, and how to turn it into something actionable. That’s the work, the hard part. More than nine months of it, twelve hours a day.
What I built
CitationIQ is a platform. Not a tool, a platform, and by the end of this paragraph I think you’ll see why I keep insisting on the difference.
Most things in this space tell you whether an AI mentioned you. CitationIQ tells you things I haven’t seen anywhere else. It can tell you whether ChatGPT, Claude, Gemini, or Perplexity already knows your brand from its training versus pulling you in live from the open web, which sounds academic until you realize the two require completely different strategies to influence. It can take a piece of your content and test whether it would earn a citation at all, often before any platform has even discovered it, which separates a content quality problem from a discoverability problem, two things practitioners usually can’t tell apart. It puts your content beside your competitors’ on the exact same LLM queries and shows you where they’re pulling ahead and which gaps are actually worth closing first. And it does all of this across languages, not just English, because the brands worth tracking don’t all operate in one market.
It comes in two forms. If you’re in-house, you point it at your own site and your competitors and go. If you’re a consultant or an agency, there’s an entirely separate side built for managing a portfolio of clients, because doing this work for ten brands is a different job than doing it for one, and most platforms pretend it isn’t.
And yes, underneath all of that, I also had to build the ordinary things. Query alignment, citation tracking, technical health, the query fan-out that turns a handful of seed questions into a real test library, the reporting layer that ties it together. The expected pieces are all there. I just don’t think they’re the reason you’d want to look.
Here’s what the data actually looks like. In one anonymized beta run, CitationIQ analyzed 100 AI responses and found 57 paraphrased matches to the tested site’s content - meaning the AI’s answer sounds like their content, may be built from it, structured around it. Linked citations: zero. Direct mentions: zero. That is a gap most teams cannot see.
The platform also classified every response by source layer. Forty percent came from model memory alone - the AI answering from training data, completely independent of what’s currently on the site. Twenty-five percent came from live retrieval only. Thirty-three percent came from neither layer, meaning the AI answered the query but the brand’s content played no role at any point. Just two percent drew from both layers simultaneously. Each of those states requires different approaches to optimize for.
The same run flagged three queries as competitive risk, including one where a rival was outscoring the tested site on the most foundational query in their entire category by a margin that requires real work to close, if they even can.
A paraphrase problem is not the same as a citation problem. A memory problem is not the same as a retrieval problem. A competitive gap is not the same as a content quality issue. This is why “AI visibility” cannot be reduced to a mention tracker. The useful question is not just whether AI mentioned you. It is whether the system knows you, retrieves you, cites you, paraphrases you, ignores you, or gives the opportunity to someone else.
You still need to do the work. I’m giving you a flashlight so you’re not doing it in the dark. The model is the same one that worked at Bing: open the data, make it readable, give businesses what they need to make decisions. The only thing that’s changed is which machine is keeping the data.
I am not a typical founder. No CS degree, no recent employer whose logo opens doors. What I have is 20-plus years of watching data accessibility problems emerge before most of the industry recognized them, and a track record of building infrastructure that closed those gaps. The Machine Layer covers the broader landscape for practitioners who want to understand where this is all heading. CitationIQ is the data platform for those who are ready to act on it.
Data for decisions. Evidence for your conversations.
If that sounds relevant to work you’re doing, the platform is at www.citationiq.com.
This is bootstrapped - my own money, my own time, my own work. Trademark in process. Patents if I want them. I’ve been busy. Of course, it’s a beta launch, so you should expect some rough edges, and things you might have done differently. Languages, for example - the work and outputs can be managed across 25+ languages, but the interface inside the website remains in English for the time being. Expect lots of things you simply don’t understand, too. For those moments, everything has a handy “Understand This Data” button with dropdown explanations nearby.
I know what this data can do for practitioners who are ready for it. The platform is open. Come find out. And as I said earlier, it would mean the world to me if you can help me spread the word about this launch.
