agentWOW

Turning whispers into insight

Incorrect PIN

PubAgent Limited

PubAgent Limited

White Paper

This paper argues that AI visibility — the thing we've built a tool to measure — is not the thing that matters most in hospitality.

It argues that the GEO industry has a structural blind spot, that optimising for AI recommendation without checking whether the experience can deliver is actively dangerous, and that 80% of the people in a well-run pub are there because of something no algorithm will ever capture.

This is not a dashboard walkthrough.
It is an honest account of what we've learned,
what we can measure, and what we can't.

Proceed to the white paper
Return to dashboard
White paper · March 2026

The Pint & Pie Index
Why we won't send traffic to a pub that can't deliver.

On measuring what AI says about your pubs — and why that's not the whole story.

Joe Mullane · PubAgent Limited · agentwow.co.uk

The thesis

Right now, somewhere nearby, someone is deciding where to go tonight. They're not Googling. They're not checking a review site. They're remembering. The last time they walked through that door, the beer was kept well, the welcome was warm, the food arrived hot, and they left thinking they'd go back. That memory — formed by a real experience, delivered by real people, in a real room — is making the decision for them. No algorithm will ever capture it. No schema markup will trigger it. No GEO strategy will measure it.

And yet it is the single most powerful force in hospitality. The return visit that was never prompted. The recommendation to a friend that was never tracked. The quiet loyalty that no dashboard will ever surface.

The premise of this paper is simple: drive traffic to pubs and beer brands we know can deliver — because they're proving it through reputation. Not through optimisation. Not through schema tricks. Through the evidence of customers who went, experienced, and came back.

The only honest starting point for measuring AI visibility is to acknowledge what it cannot measure: whether the experience was good enough to bring someone back without being asked.

Here is the uncomfortable truth for every GEO platform, including this one: in a pub that is doing its job well, roughly 80% of the people in the room on any given night are there because they've been there before. They are repeat visitors. They came back because the last visit was good enough. AI didn't send them. Google didn't send them. A schema tag didn't send them. The experience sent them. The remaining 20% — the new faces, the curious, the people who asked an AI where to eat tonight — are the ones discovery tools can influence. That 20% matters. It's where growth comes from. But it only converts into the 80% if the pub delivers when they arrive.

I've spent six months in search of a single truth: how do brands get discovered when the customer never reaches a website? The answer changed everything — but not in the way the GEO industry would have you believe.

SEO was about persuasion — gaming an algorithm with keywords and links. The content didn't have to be honest. It had to be optimised. GEO is the opposite. Generative Engine Optimisation rewards honesty emitted in a structured way. Schema markup, JSON-LD, clean metadata — these aren't tricks. They're declarations of fact. You tell the machine what is true about your business. The machine decides whether to believe you. That's what I love about it. You can't game a conversation.

But here's what nobody in the GEO industry is saying: visibility without readiness is dangerous. When AI recommends your pub, it doesn't hand the customer a link. It hands them a promise. And if the experience doesn't match the promise, you haven't just failed to benefit from AI visibility — you've used AI to accelerate your own reputational damage.

I know this because I've lived it. I've run pubs. I've been on the floor every night. I've watched the same faces come back on a Friday because the previous Friday was good enough. I've seen what happens when the kitchen falls apart on a Saturday night and the regulars quietly stop coming. The 80/20 isn't a statistic I read somewhere — it's something you feel when you're behind the bar and you recognise most of the room.

Building agentWOW required both of those worlds to meet. The structured data, the schema, the prompt engineering, the AI measurement framework — and also the floor-level understanding of what actually makes a pub work. What keeps a cask ale well. Why a particular table by the window gets requested. How a kitchen holds together under pressure and what it looks like when it doesn't. The measurement system only makes sense if it's built by someone who has stood on both sides of the door — the digital signal and the physical experience. Without that, you're just counting whispers and hoping someone else delivers the truth.

The tool we've built is good at measuring whispers. But the honest truth — the boy-in-the-emperor's-new-clothes truth — is that the whispers aren't the thing. The thing is what happens in the room. The tool only matters if it sends people to rooms worth being in.

That's the starting point. Everything in this paper follows from it.

The amplification trap

A coastal pub in Kent scores 29 on the Pint & Pie Index — invisible to AI. The Blueprint engine identifies the gaps. Schema is updated. Content improved. Three months later it scores 68. Gemini is recommending it for coastal dining. Perplexity is citing it for Sunday lunch. ChatGPT is describing the sea views.

The kitchen is slow. The Google rating is 3.6 stars, with dozens of mentions of cold food and indifferent service. This was true before the optimisation began. It's still true now.

New customers arrive with elevated expectations — because that's what AI recommendations create. They encounter the slow kitchen. They leave disappointed. They write reviews. The rating doesn't improve. The AI learns from the new reviews. The recommendation hedges. The score falls. Within six months, the brand is back where it started — or worse.

A pub that scores 29 and stays at 29 has a passive problem — it's missing customers. A pub that scores 29, optimises to 68, and drives disappointed customers through its door has an active problem. It has spent money to produce new customers specifically in order to disappoint them. The GEO industry, in its current form, has no framework for this. It measures visibility. It does not measure readiness.

How the index works

In 1986, The Economist's Pam Woodall created the Big Mac Index — purchasing power parity made instantly understandable by anchoring it to a burger. Same product, available everywhere, easy to compare. The Pint & Pie Index does the same thing for AI discoverability.

AskReal customer queries sent to four AI platforms. "Best pub for Sunday lunch in Whitstable." Every response recorded.
ReadExtract which pubs and beers were mentioned, in what order, whether the facts were right, what competitors appeared.
ScoreEvery entity gets a 0–100 score. How visible, how accurate, how strongly recommended — across all platforms.
CompareEvery score measured against the index average. The number isn't the story. The gap is the story.
GateBefore amplifying, check: is this pub ready to receive the customers AI will send? Reputation readiness before visibility.

Three principles govern the index: measure what you can measure (share of conversation, not size of conversation), compare like with like (each pub against its own competitive context), and reject vanity metrics entirely.

What gets scored

Pubs — six dimensions

Mention rateDoes your pub come up?
AccuracyDoes AI get the facts right?
Recommendation strengthFirst pick or afterthought?
Attribute coverageDoes AI know what makes you special?
Competitive positionYou or your competitor?
Brand attributionDoes AI name the parent brand?

Beers — five dimensions

RecognitionDoes AI know it exists?
AccuracyStyle, ABV, description correct?
RecommendationDoes AI suggest it?
Brand connectionConnected to the parent brand?
Competitive positionThis beer or a rival?

Brand attribution matters because in hospitality, the product builds the brand from the bottom up. A pub recommended with its brand is a brand ambassador. A pub recommended without it is a missed opportunity. A pub recommended with its brand whose experience disappoints is a brand liability.

The four signals

AI visibility is one signal among four. The full picture requires all four layers read together.

1
AI narrativeWhat is AI saying about this pub? The Pint & Pie Index score.
2
Customer voiceWhat are customers actually experiencing? Reviews, ratings, recurring themes.
3
Customer behaviourWhat are customers doing? GA4 sessions, direction clicks, booking actions.
4
Structured truthWhat have you told the machine? Schema, JSON-LD, GBP, menu data.

The diagnostic power is in the gaps between layers. A gap between what AI says and what customers experience is the amplification trap in its earliest detectable form. A gap between strong reviews and falling GA4 isn't failure — it's the shift in action, customers arriving via AI instead of Google. Understanding which gap you're looking at determines the response entirely.

The four scenarios

When the four signal layers are read together, every pub falls into one of four diagnostic quadrants.

Earned visibilityHigh AI score, strong reviews, healthy traffic.Amplify with confidence. This pub deserves every customer AI sends.
Hidden gemLow AI score, strong reviews, healthy traffic.Visibility gap. The pub is ready — AI just doesn't know it yet.
The trapHigh AI score, weak reviews, fragile traffic.AI is making promises the experience can't keep. Fix operations first.
Invisible and unreadyLow AI score, weak reviews, weak traffic.Operational recovery before any AI work. Visibility would make this worse.
Most of the GEO industry operates in two of these quadrants — the green ones. That's not a criticism. It's a structural gap. Nobody is asking the harder question: what happens in the red quadrants, where visibility is actively doing damage?

Reputation readiness

Before any AI optimisation work begins, there is a prior question that the industry is not asking: is this pub ready to receive the customers that AI visibility will send? Readiness isn't perfection. It's alignment — ensuring that what AI says corresponds, within a reasonable margin, to what customers experience.

Tier 1
Fix firstReview rating below threshold, recurring negative themes. AI optimisation should be paused. Operational improvement comes first.
Tier 2
Fix and buildMixed signals. Selective AI work alongside reputation improvement. Watch the gap between what AI says and what reviews say.
Tier 3
AmplifyStrong reviews, consistent experience. AI optimisation can proceed with confidence. These are the pubs where new traffic has the highest probability of converting into returning trade.

For estate operators, this answers the question that's almost impossible to have on instinct alone: which pubs are ready for the spotlight, and which need the kitchen fixed first? Reviews — interrogated properly, with recurring themes surfaced by AI — are the closest thing we have to hearing the whispers we can't measure directly. They're not the full picture. But they're the nearest proxy for the experience that's actually being delivered.

Platform coverage

Gemini0.45750m direct + 2bn via Search + Siri. The bridge between old and new discovery.
ChatGPT0.25800m weekly users. Where people plan meals, trips, nights out.
Claude0.15Different training data. Genuinely contrasting perspective.
Perplexity0.15780m queries/month. AI-first search with citations.

Four platforms, four distinct models, ~90% of global AI search market. Gemini weighted highest because it powers Google AI Overviews (78% of hospitality searches) and Apple's Siri from 2026. Structured data is the single most important technical lever — Gemini is built on Google's index, and Google rewards structured truth.

The parent brand effect

In a multi-site estate, every pub's AI visibility reflects on the parent brand. A customer sent by AI to a pub that disappoints doesn't just form a negative view of that pub — they form a negative view of every pub carrying that name.

The risk isn't symmetrical. An AI-invisible pub causes no reputational damage — it simply misses customers. An AI-visible but operationally unready pub causes damage that extends across the entire estate. In the AI era, the weakest operational link in a branded estate is a brand risk, not merely a commercial one.

The question is no longer "which pubs are underperforming commercially?" It's "which pubs are AI-visible but reputation-unready, and therefore actively putting the brand at risk?"

The sequence

This is not a counsel of passivity. It's a counsel of sequence.

Operational excellenceThe experience a customer has when they walk through the door. The kept ale. The warm welcome. The food that arrives hot.
Reputation that reflects itReviews, ratings, word-of-mouth that evidence the reality customers actually encountered.
Structured data that communicates itSchema, JSON-LD, clean metadata — honesty emitted in a structured way. Declarations of fact, not persuasion.
AI visibility that amplifies itTraffic driven to pubs we know can deliver. Pints sold. Pies sold. People through the door who will come back.

Get the sequence wrong and the amplification works against you. Get it right and you're not gaming an algorithm — you're giving AI the information it needs to represent a genuinely good experience accurately. The recommendation follows naturally from the reality.

But there is a fifth thing in this sequence that no measurement tool — including this one — will ever capture. It's the customer who came back because the last visit was good. Not because AI told them to. Not because a schema tag was present. Because the experience was worth repeating. That decision, formed silently from memory, is the most commercially powerful signal in hospitality. It has no digital footprint. It leaves no trace in any analytics platform. And it is the only truly honest measure of whether a pub is doing its job.

Everything we measure operates in the 20% — new discovery, new faces through the door. But the point of the 20% is to feed the 80%. Every new customer sent to a reputation-ready pub is a customer who, if the pub does its job, becomes part of the returning majority that no algorithm needs to reach. That's the conversion that actually matters. Not click to visit. Visit to return.

We can measure the whispers. We can interrogate the reviews. We can read the structured data and score the visibility. But the thing that fills a pub on a Tuesday night in January — the quiet certainty of a regular who knows the beer will be good and the welcome will be warm — that was never going to show up in a dashboard. It was earned on the floor, one good night at a time.

The brands that win the AI era will not be the ones who shouted loudest at the algorithm. They will be the ones who earned the recommendation before they sought it — and delivered an experience good enough that the customer came back without being asked.
© 2026 PubAgent Limited. All rights reserved.
The Pint & Pie Index methodology and agentWOW platform are the proprietary intellectual property of PubAgent Limited.
agentwow.co.uk
Close white paper