Performance Communications | أغسطس 11, 2026
Andy Francis explains how LLMs are the ultimate dinner party hosts when it comes to understanding the scrap for AI recommendations
Measuring success on how many times your brand website is cited is NOT the goal in the age of AI-assisted search.
Let me explain.
Picture a dinner party. The LLM is taking your place as the host at the head of the table. Every journalist, reviewer, owner, forum lurker and YouTuber is a guest, and every one of them has a fact or an opinion about your brand or your car.
The host’s job is to listen to the room, work out where the truth sits, and then report back to you with a straight answer to your question… whether you’ve asked for the best seven-seater for under £30k, the most reliable EV for a 200-mile commute, or the ideal family car for a Labrador and two child seats.
The host’s job is to find consistency. To find a consensus. To find the truth from the brand buzz. And ultimately, answer the specific question that has been asked.
That’s it.
Which is why the first thing to fix isn’t your content. It’s your measurement.
Monitor and measure with honesty
Measure the prompts that reflect the actual consumer journey. The questions real buyers ask, in the order they ask them, in the language they use. Not the ones where you already know you’ll look good.
And while we’re here: how many times your website gets cited is arguably irrelevant.
Three things matter. Is the brand or product mentioned? Is it mentioned positively? And is it recommended? That last one is the whole game. An LLM can quote your website at length and still tell the buyer to go and look at someone else’s car.
It’s worth stating the facts here because it comes up in every conversation I have: the LLMs will always come to your site – be that at the start of the journey, the middle of it, or in the transactional phase.
Making sure the content you have is structured, readable and answers the intent alone, does not guarantee you are Mentioned, let alone Recommended. Don’t get me wrong, it is important but it’s not the measure that will ultimately impact the action of the consumer reading the answer provided.
It’s not about being the best. It’s about being the right answer.
This is where a lot of the AI visibility conversations go wrong. It gets framed as a quality contest, as if the LLMs are running a giant group test and the winner takes the recommendation.
They’re not. They’re matching a question to the most credible available answer.
Think about the budget airlines. Nobody is claiming they’re the best airline in the world, least of all them. Their proposition is cheap, and they’ve never once wobbled on it. Ask an AI for the cheapest way to get to Lanzarote in October and they’ll be right there, because the claim, the coverage, the reviews and the customer experience all point at the same thing.
The automotive equivalent is sitting in plain sight. Dacia doesn’t pretend to be premium. It says it’s honest, affordable transport, the media review it as honest, affordable transport, owners recommend it as honest, affordable transport, and the registration figures show people buying it in numbers. Every signal agrees. Ask for the best value new car and there’s no ambiguity for the host to resolve.
That’s the real lesson. You don’t need to be the best car on sale. You need to be unmistakably the right car for a defined set of buyers, and you need every source the AI can reach to say the same thing about who that buyer is.
You can’t blag it.
Where it falls apart is when the claim and the reality drift.
You can say you’ve built best-in-class refinement and segment-leading practicality. But if the media say otherwise, if the forums are quiet, if nobody’s posting about their delivery day, and if the SMMT figures say the country isn’t buying it, you’ll be found out.
In short… you can’t trick an LLM. It’s got the whole room to ask.
Consistency is the currency.
Brand, earned, owned, CRM. It all has to say the same thing.
Marketing language is built for awareness, and it does a critical job in familiarity and desirability. Nobody’s arguing for stripping the romance out of a brand campaign. But if the way you describe the car is a Count Binface Galaxy away from how the media describe it, the host doesn’t hear consistency. It hears a discrepancy. And discrepancies don’t get recommended.
The test is simple enough. Take your positioning line, take the last 10 reviews, take the last CRM email a customer got. Do they read like the same car?
It’s a together job.
I’ve been saying this from the beginning. This is the biggest risk to long-term success in GEO and AEO. AI visibility isn’t a PR project or a marketing project, and it isn’t solved by either function applying its old theory to the other’s work. It is a TOGETHER job.
Consistency and the truth is the strategy now. Not the truth about how good you are, but the truth about who you’re for. Which is either terrifying or liberating, depending on how clear you’ve been.
