If LLM Visibility Is the Answer, What Is the Question?

That was the prompt on the invite for our breakfast at the Union Club last week, and it turned out to be the right one. 

I’ve sat through enough GEO sessions in the last six months to recognise a familiar rhythm – list the platforms, name the tools, count the prompts, leave with a checklist. But between Jess Larsen’s keynote and a panel of B2B marketers and strategists who’ve been doing the work for real, the takeaway was something more uncomfortable and more useful: the question behind LLM visibility isn’t “how do we game the algorithm?” It’s “do we actually know what we stand for” and “is that what’s being said about us?”

Here are our five takeaways from the event below: 

1. AI is an amplifier, not a strategy

AI is not neutral. It’s an amplifier. If you’re dysregulated, overwhelmed and operating performatively, AI will accelerate that – more noise, more volume, more content that looks fine on the surface and means nothing once you read into it. That’s AI slop in action. 

But if you can hold onto clarity, original voice and genuine thinking, AI amplifies that too. You get more reach, more leverage, more impact.

2. Start with the audience, not the algorithm

The panel – Danielle Le Toullec (CMO at Erevena), Alex Deakin-McKay (Marketing Director Xeinadin), Kate Sikora (Managing Partner, Noble Performance) and our own Harry Webster – kept arriving at the same answer from different routes: stop with the tools and start with the people.

Danielle made the case that personas need to evolve. Add the question “which LLMs are you using and how are you using them?” to your next persona exercise. Different audiences live in different LLMs at different stages of the buyer journey, and assuming ChatGPT covers it is already out of date – the audience she’s been working with overwhelmingly uses Claude. The strategy follows the audience, not the other way round.

The thread underneath all of it? GEO doesn’t start with the LLM. It starts with knowing who you’re talking to, what they’re asking, and where they’re asking it.

3. What AI Trusts, and Why

Roughly 95% of AI citations come from earned media. That stat got quoted around the room more than any other, and the implication is straightforward – the path to LLM visibility runs through PR and content strategy, not ad spend or back-end optimisation. Brands that have invested in genuine thought leadership and earned media for years have a head start they may not even realise they have.

What surprised people more was where else the trust signals are coming from. Alex talked about Xeinadin showing up in LLM citations because of Trustpilot reviews, Google Reviews and trade press articles in places like Accountancy Age – not because of anything Xeinadin had said about itself. AI models trust what others say about you more than what you say about yourself.

The compounding effect: editorial standards are going up, not down. One UK publishing group has reportedly started operating an approved list of PR agencies it will take stories from, in response to the volume of AI-generated pitches landing in journalists’ inboxes. Real journalist relationships and original thinking just got more valuable, not less.

4. The traps to avoid

The “what to avoid” thread was, judging by the LinkedIn posts since, the part most people found most valuable. Two patterns stood out.

The first is the overnight expert. Alex was sharp on this: a lot of people have suddenly become AI and GEO experts in the last twelve months, and not many of them have actually done the work. His test, when he’s hiring suppliers, is to ask what their team has looked like over the last twelve to twenty-four months and what upskilling they’ve done. If the answer is vague, he moves on.

The second is treating tooling as the strategy. The measurement platforms – Peak, Profound, Semrush, Muck Rack and the rest – are useful, but Harry’s point was that none of them are useful without human interpretation. Visibility scores, share of voice, sentiment scores and accuracy scores all need someone in the room asking whether the underlying data actually makes sense. (We’ve written separately on the GEO measurement tools we tested if you want the detail.)

The encouraging news, sitting underneath both warnings, is that the conversion data rewards getting it right. Stats discussed on the day put referral traffic from Google down by around a third in mid-2025 (probably closer to 50% now), but conversion rates from LLM traffic running about 4.4x higher than from traditional search. The buyer has done more research before they arrive – intent is higher when they hit your site, but the window is shorter. The homepage that asks “who are you?” is already losing.

5. Where this leaves us

If you’re early in the GEO conversation, we’ve already published the practical pieces – a GEO explainer, a glossary of AI terminology, six practical steps to improve GEO visibility and a piece on why press releases work for LLMs. Take whichever is most useful and run.

But the strategic argument sitting above all of those tactical pieces became clearer in the room than it has been anywhere else. GEO is not a technical discipline being sold by people who learned it last quarter. It’s a communications discipline that rewards clarity, consistency, earned credibility and human judgement – exactly the things AI can’t replicate, and exactly the things it amplifies when you bring them.

So the question I’d leave you with is the one I think every B2B marketer should be sitting with right now. If a buyer asked an LLM about your business tomorrow, would the answer be the one you’d want?

If you don’t know, that’s the place to start.

Sean Hodges, Account Director