In AI search, the most-cited brand wins. Here's how we became it.

Rankings measure where you show up. Citations measure whether you're the answer. That difference decided our category.

In early 2025, Google's AI Overviews started answering questions before anyone scrolled to the results. My spidey sense was tingling. If people were going to get the answer before they saw a single link, ranking wasn't the goal anymore. Being the answer was.

My first move was the classic marketer move. I asked our Google rep when we'd be able to buy ads in AI Overviews, since every promoted placement sat below the magic box. We had to be able to buy our way in, right? He didn't know when, or if, that was coming. If we wanted into the answer, we needed to be the chosen one.

A ranking is a chance to be seen. A citation is being chosen. In a category where the buyer is often a caregiver, tired and asking a very specific question at 11pm, being the answer is worth more than being on the list.

1,291citations, against 986 for the nearest competitor. As of June 2026, the brand is the most-cited domain in its category (source: Scrunch).

Being first has its perks. And its problems.

The first problem: nobody could measure it. Rankings went into the monthly report, but citations, the moments when an assistant actually names your brand, weren't tracked at all. The SEO tools we already used didn't track AI answers yet. We found a tool called Scrunch, layered it into our reporting and put AI citations right next to the search report.

That changed the conversation. The team stopped asking "where do we rank?" and started asking "are we the answer? If not, who is, and why?"

The second problem: the bots made a mess of GA4. The same LLM crawlers we were racing to attract flooded our sessions data. So while we worked to get assistants to crawl and cite us, we also had to figure out how to separate their traffic from real visitors.

What we actually did

Inside the team, AI search became almost a mythical creature. We called it the AI monster, and the strategy was simple: feed the beast. Here's what we fed it.

• Retrofitted hundreds of our 400+ blog posts. Each one got a key takeaways section up top (a TL;DR for the LLMs) and an FAQ section, because LLMs love a Q&A format.
• Built authority hubs. Long-form pages designed to be the definitive resource on the topics that matter most to our customers. The kind of page an assistant can reference with confidence.
• Added schema and structured content. So machines can tell what each page is, what it answers and how the pieces connect.
• Doubled down on video. We hired a full-time, in-house video producer. The bet: Google loves Google, so YouTube is likely to show up in AI Overviews.

Here's the part people find hard to believe: assistants reward the same clarity people do. If a person can skim your page and get the answer in ten seconds, a model can too. If a person has to wade through preamble, so does the model, and it'll quote someone else.

The three things that made the difference

Moving early. We started before there were tools built for it. Complicated? Yes. Worth it? Also yes.

Structure. Every hub has a clear question at the top, a direct answer beneath it, then the depth. Headings that say what the section is. Tables where a table is honest.

Measurement. We started out chasing AI click-through: traffic from assistants to our site. Then we learned LLM traffic is more of a research step, so clicks aren't the most telling number. Citations are. Content programs die in the gap between "we published it" and "it worked." The right number closes that gap.

Rankings tell you where you stand. Citations tell you whether anyone listened.

What to do this quarter

Pick the ten questions your buyers ask most. Not the ten keywords with the highest volume. The ten questions a real person would ask a knowledgeable friend. Answer each one better than anyone in your category, in plain language, on a page you own.

Then go back through what you've already published. Add a TL;DR and an FAQ to your top posts. It's the fastest win on the list.

Then start counting citations. Ask the major assistants your ten questions once a month and record who gets named. Don't judge it by clicks. It isn't perfect measurement. It's good enough to see a trend, and a trend is what you need to keep the program funded.

And before you trust your sessions data, check how much of it is bots.

If nobody in your company is counting citations yet, that's the opening. It was ours.

Up next

We're prepping for agentic shopping. Because if Claude can plan my Christmas trip and map out how to max my credit card reward points, it can surely buy adult diapers.