My approach
Nothing's set in stone. Every piece of work starts with what's actually going on for your brand, and the work fits that, not the other way round. I'll always tell you exactly what I'm doing and how.
I spend a lot of time telling people to ask GEO agencies how they actually work. So here's mine, in full.
Most work sits somewhere on this range. Plenty of clients use more than one.
Briefings, workshops and talks for PR, comms and leadership teams. How AI search works, what it means for reputation and coverage, what the evidence says and where the hype runs ahead of it. Good for teams who keep being asked about AI and want a straight answer.
A review of the things that make your brand easier for AI to find, trust and cite: your own site, your coverage, your spokespeople, and the third-party places AI reads in your sector. You get clear, prioritised recommendations you can act on without buying anything.
A proper look at how your brand actually shows up in AI answers, who's doing better and why, and what to do about it. Built around your categories, competitors and markets. More on what that involves below.
It's the first thing everyone does, and it tells you almost nothing. AI doesn't have one view of your brand. It gives different answers, drawn from different sources, depending on who's asking, what they're asking, where they are in their decision and which tool they're using.
So the questions decide the answer. Generic questions get you generic results that have nothing to do with the story you're trying to be part of. Questions built around your brand, your audiences and your objective get you something you can act on.
Where they are in the journey
84% vs 88%
Across all kinds of questions, 84% of AI citations point to earned media.[1] But when B2B buyers are evaluating options, 88% of citations point to content brands control: product pages, comparisons and guides.[2] Discovery, consideration and decision draw on different sources.
Who's asking
1 in 4
Give the same question a different persona and the answers share only about one in four recommended brands, down from two in five for the same persona. Shared sources fall from 34% to 20%.[3]
How they ask
+25%
Phrasing a question as a ranking or a set of keywords lifted brand visibility by up to 25%. Questions that drift from the core intent can halve it, and consideration-stage questions are the most sensitive.[4]
Which AI tool
1.7%
ChatGPT, Claude, Gemini and Perplexity all cited the same website for the same question just 30 times out of 1,792. Each tool reads its own corner of the web.[5]
When they ask
<1 in 100
Ask the same question twice and the chance of the same list of brands is under 1 in 100. One answer is a sample, not a verdict.[6]
Where, and in what language
Varies
The same question can get a differently slanted answer in a different language, shaped by what dominates that country's information environment.[7] An English-only check can miss what other markets are being told.
What that means in practice. I design the research around you: your categories, competitors and audiences, the stages of the decision that matter, the markets and languages, and the narrative you want to be part of. Small changes in the set-up change what comes back, which is exactly why it's worth the thought. The closer the research fits your objective, the more useful the answer.
What tailored research looks like
The questions
Built around realistic personas, the factors that matter in each category and the stages of the decision: discovery, consideration and choosing. Mostly unbranded, with branded, head-to-head or reputation questions if they're useful.
The run
Run across the AI tools your audience uses, typically ChatGPT, Google's AI Overviews and Perplexity, over several days rather than once. In every market that matters, in the local language.
What you get
Your share of voice against named competitors. The sources, outlets and articles AI is citing, and why. The types of content that are shaping the answers. Usually a highlights deck and walkthrough first, then a fuller report.
No score at the end. A clear picture of who owns the conversation in your space, and what you'd change because of it.
I'm not tied to any platform. That doesn't mean I avoid tools. It means I pick them for the job.
What I won't do is tell you to sign up to anything that promises to solve AI search for you. Even if it did this week, it wouldn't next week. AI tools change too often for any one product to be the answer.
It depends on the brief. A team briefing and an eight-country research study are very different things. Once I understand the problem, I'll send you a short written scope with a fixed price, so you know exactly what you're getting before anything starts.
Each study measures something slightly different, and several come from companies that sell AI search tools. Read them as evidence of how much the set-up matters, not as benchmarks for your brand.
Get in touch
A short email is fine: what's going on, what you've been asked, and when you need it. I'll give you an honest answer, including if I'm not the right person.
hello@knownandcited.com