Resources for media covering The Guinndex UK
The Guinndex UK is the most comprehensive pint price index ever compiled for the United Kingdom. Built by AI engineer Matt Cortland and AI researcher John Fleming, it uses an AI voice agent named Rachel and the Google Places API to index 46,237 pubs across England, Scotland, Wales and Northern Ireland, with 6,544 verified prices across 108 ceremonial counties and council areas.
The Guinndex UK is the third project in six weeks from the duo, following The Guinndex Ireland (a viral sensation across Ireland in March 2026) and the Gas Index USA (170,000+ US petrol stations indexed in early April 2026).
Coverage from our sister projects: The Guinndex Ireland (guinndex.ai) and The Gas Index USA (gasindex.ai). UK-specific coverage will be added as it publishes.
Lumina Intelligence's 2025 UK Pub and Bar Market Report estimates 41,691 trading pubs and bars in the UK, down from 45,791 in 2019, with around eight net closures per week. Google Places, by contrast, lists 46,243 UK venues as pubs. The Guinndex UK is the first project to ground-truth the Google count by actually phoning every listing. Based on what Rachel found (3,334 confirmed non-pubs so far, extrapolating to 7,000 to 11,000 misclassifications), the real UK pub count is likely closer to 31,000 to 37,000. Google overcounts by 25 to 30 percent.
Historical drink price data sourced from the Office for National Statistics (series CZMS, CZMT, CZMR, KEF4, CZCG). Market context from Lumina Intelligence UK Pub and Bar Market Report 2025.
Two AI engineers built a voice agent and research platform to call 29,603 pubs over Easter bank holiday weekend to find the price of a pint. The result is the UK's largest venue-level pint price index and a new standard for AI-driven consumer research.
46,618 pubs indexed. 6,554 verified prices. National average: £5.82. Only 4% of those who answered realised they were speaking to an AI.
(LONDON, UK, 22 April 2026)
“Where can I find the cheapest pint of Guinness near me?”
It's a question that enthusiastic punters have been asking each other since 1759, and it wasn't until 2026 that AI has finally provided the answer. AI research engineers (and husbands) John Fleming and Matt Cortland have developed The Guinndex, an editable pint price index that used an AI voice agent named Rachel to call almost 30,000 pubs across the UK over the Easter Bank Holiday weekend and ask a simple question: how much does a pint of Guinness cost?
29,603 pubs dialled. 8,221 landlords and barstaff answered. 6,552 verified prices. National average: £5.83. London average: £6.74. Cheapest pint: £2.94. Most expensive pint: £9.50.
Website visitors can log, confirm, and correct pint prices via the “Contribute” button on the Guinndex UK website.
The goal of the Guinndex is to stabilise the price of a pint across the UK and Ireland and provide pricing transparency so that customers can more purposefully choose where and how to spend their money. It aims to highlight excellent pubs, hidden gems, and where to find a good deal, and which pubs may be price gouging.
To achieve this feat, the couple developed an AI research platform to do the things that are impractical for humans but easy for AI. In this case: place a phone call to every pub in the UK, have a polite chat on the phone, and find out the cost of a Guinness. In doing so, the couple has stumbled upon a seemingly novel way of gathering hard-to-find data insights at-scale that has the economic community reeling. The project cost around £500.
“One engineer and one researcher who know what they are doing with AI can now stand up national-scale infrastructure in days,” said Cortland. “The pint price is just the first (albeit important) use case. The same methodology generalises to any domain where the data lives inside a phone call no one has ever had time to make. It's pretty groundbreaking, actually.”
The Guinndex UK is the third project in six weeks from Fleming and Cortland, after the Guinndex Ireland in March and the Gas Index USA in April. Fleming's PhD research is in computational modelling of neurological disorders with current work in deep brain stimulation and neural networks, applied now to AI safety and evaluation research. The Guinndex is also, deliberately, a demonstration of what modern AI tooling makes possible in 2026.
Across roughly 15,000 Easter calls, only 4% of UK pub staff seemingly realised they were speaking to an AI. When Rachel used her native Northern Irish accent in Northern Ireland specifically, that figure fell to 0.6%, and successful pint price extraction rose by a third. Eight times lower detection from the same voice, same script, same prompts, simply by matching the local accent.
That the detection rate is so low in the first place is a testament to ElevenLabs' voice technology. That accent-matching drives it eight times lower again is the kind of real-world finding laboratory studies of voice AI have rarely been able to produce.
“As an AI researcher, the model evaluation side of the Guinndex project is particularly interesting,” said Fleming. “We get to see how a voice agent performs in the wild, not in a lab. You can't manufacture 13,000 real conversations with British and Irish pub staff on a Friday afternoon. This is a kind of deployment data that can truly enrich the field of Voice AI research.”
Bucketing Rachel's prices by latitude reveals a sharp north-south break. North of a line running roughly through Oxford, Cambridge and Norwich, the average pint is under £5.70. South of it, the average jumps past £6. The crossing point where bucket averages exceed the national average (£5.82) sits between 51.5 and 52.0 degrees north. Crossing south of the 52nd parallel adds roughly 60p to the average pint.
Most pubs picked up the phone, answered the question, and got on with their shift. A selection, verbatim:
At The Fontmell, Dorset, on a first-name basis within five seconds: “Hang on a minute, Rach.” (£6.50)
At Missions Arms, a barman clocked the project immediately and gave the price anyway: “Is this an AI voice bot? Yeah, will do. There was someone doing that in Northern Ireland, wasn't there, going around an index. Five pounds seventy-five for a Guinness.”
At The Duke's Umbrella, Glasgow: “I'm just gonna double-check that before I tell you some lies.” (£7.00)
At The Earl of Normanton, Wiltshire: “You think I ought to know that off the top of my head. That's the problem with, uh, technology.” (£5.60)
At The River View Pub, Wirral: “20p, babes. Thank you. It's 4.70 on my receipt.”
At Broad Street Townhouse, Bath: “Let me check again because sometimes the prices go up, as usual.” (£7.45)
At The Volunteer, London: “Oh my God, is this AI?” Still gave the price. (£5.50)
At The Spring Hill, Wolverhampton: “This has been cool.” Genuinely enjoyed the call. (£5.70)
At The Tywarnhayle, Cornwall: “A pint of Guinness, 5.80, my love, there.” Peak Cornwall.
At The Nags Head, Leicestershire, on Grand National day: “I'm very busy at the moment. I'm watching the Grand National. So you can be quick.” (£5.50)
At The George and Monkey, Islington: “Let me check my hat.” Came back with £6.80.
At The Fountain Inn: “Five hundred and ninety-three pound.” Rachel thanked them politely. The price was flagged and removed.
The Singing Barman (Ben, at an unnamed pub): looked up the price while singing to himself. “It's six pound thirty.”
And at Kevill Arms, Great Yarmouth: “Oh, piss off.”
A fuller selection of written transcripts is available to editors on request.
Of 46,237 listings:
The Guinndex is a living index, not a one-off snapshot. Check your local at guinndex.co.uk. If the price is wrong, contribute a correction right on the site. Every update is logged publicly with source, contributor and timestamp, a Wikipedia for pint prices.
Rollup to the 33 London boroughs via postcode-to-borough lookup. Greater London overall average: £6.72 across 783 verified prices.
A pint in Kensington and Chelsea costs £2.62 more than in Harrow. Both are London boroughs. The gap is the postcode.
Explore the full dataset: guinndex.co.uk
Sister sites: guinndex.ai (Ireland) and gasindex.ai (USA)
Instagram: @guinndex.ai
Media contact: press@guinndex.co.uk
Matt Cortland and John Fleming are available for interview. Press assets at guinndex.co.uk/press, all graphics downloadable. Select call transcripts are available upon request.
Methodology, funnel view. Of 46,237 indexed venues: 35,659 were dialled; 23,027 transcripts were logged; 12,948 reached a human; 6,544 verified prices stored.
Calling timeline. The bulk was done over the Easter bank holiday weekend 2026 (~15,000 calls). Further runs before and after brought the total to 35,659 pubs dialled.
Detection figures. The 4% figure is the Easter weekend run rate. The 0.6% figure is specific to Northern Ireland calls using Rachel's native accent.
Verified prices exclude placeholders and outliers (below £2.50 or above £10.00). Every price is logged publicly with source, contributor and timestamp at guinndex.co.uk/changelog.
Historical context. The Office for National Statistics has tracked a pint of draught lager every year since 1987 (93p then, £4.77 in 2024). It has never tracked Guinness, stout, cider, craft beer, or any drink at any individual venue. The Guinndex is the largest venue-level pint price dataset the UK has ever had.
AI transparency policy. Rachel identifies herself as an AI agent and never denies or obfuscates. Research is being led by Dr. John Fleming as part of a forthcoming paper on AI voice agent field performance.
Sister projects. Guinndex Ireland (guinndex.ai, launched 16 March 2026, 1,818 verified prices, 995 community contributions since). Gas Index USA (gasindex.ai, launched 6 April 2026, 170,000+ stations indexed, 93,650+ verified prices).
Site features. Interactive UK map, Find-Your-Pub search, Near Me geolocation, live changelog, Snap the Taps photo upload, three-tap price contribution, My Local saved pubs, pub owner claim, county and London borough pages, 37 years of ONS data, and cultural sidebars covering pub names by nation, Ye Olde, Royal Names, Questionable Names and more.
Research from Anthropic earlier this year found bartenders, cooks and dishwashers among the 30% of occupations with effectively zero exposure to AI automation. Software engineers, by contrast, are the most exposed.
“AI is not coming for the person behind the bar,” said Cortland. “As a former bar owner, I know this. AI can't pour a pint, it can't read a room, it can't tell when someone has had enough. But it can call 35,000 pubs in a fortnight and tell you where to find a decent one for under a fiver. The physical work is safe. The information layer on top of it is where AI lives, and where we should all be paying attention.”
Of 46,237 Google Places listings the UK returns as “pubs” or “bars”, at least 3,334 are something else. 913 are hotels whose lobby bars have been tagged as pubs. Several hundred are restaurant chains, others are coffee shops and cocktail bars. Our ground-truthed estimate of trading UK pubs is 31,000 to 37,000, below Lumina Intelligence's 2025 figure of 41,691. The Guinndex UK is the first project to audit the Google Places pub universe by phoning listings.
Matt Cortland is an American AI engineer from New Jersey, based in London. A former US-Ireland Alliance Scholar (the George Mitchell Scholarship) and Henry Luce Scholar, he holds a Master's in Creative Digital Media from TU Dublin and previously owned a global pub and entertainment company that operated across Ireland, the UK, and the US before moving into AI engineering and consulting. Together with his husband John, he built the Guinndex Ireland, a viral sensation across Ireland in March 2026, and the Gas Index USA, launched in April 2026. The Guinndex Ireland was for John's country. The Gas Index was for Matt's. The Guinndex UK is where they live.
John Fleming is an Irish AI researcher. His PhD research used neural networks to model neurological disorders and their treatment through closed-loop deep brain stimulation. He is now applying what he knows about how neural networks work to AI safety and evaluation research, using real-world deployments like the Guinndex UK as a testbed. He sees projects like this one as the kind of honest field data the voice AI research literature has largely been missing.
Use the Contribute button on guinndex.co.uk