The UK restaurant lives between aggregator commission and Time Out citation.

The UK hospitality estate is the third-largest private-sector employer in the country. UKHospitality's 2024 Workforce Strategy puts the sector at ~143,000 licensed venues, 3.5M jobs and £93bn GVA, behind only retail and health/social care.

Restaurants account for ~50,000 of those venues; pubs and bars another ~45,000; hotels and other accommodation ~10,000 plus a long B&B tail; cafes, quick-service and contract catering carry the rest (ONS UK Business Counts 2024). Lumina Intelligence's Restaurant Operator 2024 report tracks 22,247 outlets across the top 200 restaurant groups, with branded chain sites contracting -2.1% YoY as small groups and quality-led independents absorb share.

Operator stratification is sharp and the Areza wedge is the layer below the largest groups. Mitchells & Butlers (1,600+ sites including All Bar One, Miller & Carter, Toby Carvery), Greene King (2,700+ pubs), JD Wetherspoon (~800 pubs) and Whitbread / Premier Inn sit outside the wedge: internal data teams, in-house dev, agency-of-record contracts, procurement cycles measured in quarters.

The wedge is small groups (3-15 sites) such as Dishoom, Hoppers, Honest Burger, Flat Iron, Padella, Pizza Pilgrims, Pasta Evangelists' physical estate, Black Sheep Coffee, Mac & Wild and Hawksmoor's UK estate, where the buyer is owner plus ops director plus GM rather than a procurement committee, and the contract signs on a Wednesday. Independents (1-2 sites) form the long tail of ~40,000 of the 50,000 restaurant base.

Geographic concentration is heavy and editorial weight is uneven. London alone holds ~18,000 restaurants and bars across 33 boroughs, with Soho, Shoreditch, Borough, Marylebone, Fitzrovia and Hackney carrying the editorial heft (GLA London Business Data 2024).

Manchester's Northern Quarter, Ancoats and Spinningfields; Edinburgh's Old Town and Leith; Bristol's Stokes Croft, Clifton and Wapping Wharf; Birmingham's Jewellery Quarter and Digbeth; plus Leeds, Liverpool, Brighton, Bath, Oxford, Cambridge and Glasgow Finnieston anchor the regional tier.

The MICHELIN Guide Great Britain & Ireland 2025 awarded 8 three-star, 25 two-star and ~187 one-star restaurants across the UK, plus a deep Bib Gourmand and Green Star layer. The National Restaurant Awards 2024 put Mountain (Tomos Parry, Soho) at #1, followed by The Ledbury and Brat: a heavy small-group plus independent skew across the top 100.

Boutique hotels and event venues sit on the same software stack. VisitBritain accommodation supply data puts the boutique hotel and B&B tier at ~10,000 establishments, anchored by groups like Hotel du Vin, Pig Hotels, Soho House (UK estate) and Firmdale Hotels, with a long tail of country-house, coastal and Cotswold operators.

Event venues (wedding barns, dedicated function spaces, gastropub function rooms) overlap the restaurant tier and run identical reservation and POS stacks. Foundation and Workflow Ops translate cleanly across the boundary.

Foundation. A UK-restaurant site signalling what diners screen on: menu with prices and the 14 named allergens, head chef bio plus culinary direction, dining-room photography, integrated reservation widget (OpenTable / Resy / ResDiary / SevenRooms / Tock), order-direct deep links to Deliveroo, Just Eat and Uber Eats plus own-channel ordering, FHRS rating badge, Google Maps embed, opening hours that handle bank holidays cleanly, private dining and event enquiry form, accessibility statement.

Most independent UK restaurant sites today carry no Restaurant or FoodEstablishment Schema.org markup, no MenuItem markup with allergen properties, no Reservation Schema, no FAQPage Schema for dietary and allergen Q&A, no Review aggregateRating Schema. Every one of those is a sourced anchor for AI Search citation.

AI Search and citation. The wedge is getting cited for cuisine × borough intent ('best Indian in Shoreditch', 'best ramen in Soho', 'best tapas in Manchester Northern Quarter', 'best Sunday roast near Borough Market'), occasion intent ('best date-night restaurant in Edinburgh', 'business lunch in Mayfair', 'anniversary restaurant Bristol Harbourside'), dietary intent ('best vegan restaurant in Brighton', 'gluten-free bakery Marylebone', 'halal restaurant Birmingham'), and long-tail Q ('does Hoppers take walk-ins on Sunday', 'what time does Padella stop seating for dinner', 'is Honest Burger dog-friendly').

ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews answer those prompts today from Time Out, SquareMeal, The Infatuation, Reddit, Tripadvisor and the Michelin Guide.

The restaurant's own site is typically invisible because it carries no structured data the model can extract. Sourced pages on dish catalogue, allergen tagging, head-chef CV, sourcing narrative (Cornish day boats, Lake District lamb, Norfolk asparagus, Cumbrian beef, Yorkshire game), private-dining capacity, accessibility and a clean FAQ block walk into the citation surface and stay there.

Voice Agent. RP-neutral UK English voice agent with regional optionality, handling reservation booking, group enquiry, dietary accommodation, modification and cancellation, walk-in availability and after-hours bookings.

Multilingual fallback for the relevant clientele (Mandarin and Cantonese at Chinatown sites, Polish and Romanian at the migrant-banking-adjacent estate, French, German and Italian at central-London tourist sites). The agent qualifies on name, party size, date and time, dietary requirements, occasion and deposit handling, then checks live availability against the OpenTable / Resy / ResDiary / SevenRooms API and confirms by SMS.

After-hours coverage (22:00-08:00, all-day Monday for restaurants closed Mondays) is high-ROI without staffing cost. The leak being fixed is precise: 30-40% of inbound peak-service calls converting to competitor bookings, not UX polish.

Workflow Ops. Three highest-value flows. (a) Aggregator menu synchronisation. Deliveroo, Just Eat, Uber Eats, own-channel and dine-in menu kept on a single source-of-truth with prices, 14-allergen tagging, availability and 86-list pushed bidirectionally. Mismatch between dine-in and delivery menu is a daily ops failure that costs covers and 1-star reviews.

When the head chef flags '86 lamb shoulder' at 18:00, the flag reaches all three aggregators and the website inside a minute. (b) Reservation and reviews loop. Pre-visit confirmation with allergen pre-check, deposit collection for 6+ groups and tasting menus, no-show prediction, post-visit thank-you with review prompt (Google, Tripadvisor, OpenTable, Resy), VIP recognition for returning guests via SevenRooms-style guest-history surfacing.

(c) Stock and rota. Demand forecast over POS history plus reservation book plus weather plus local events (football fixtures, concerts, theatre, exhibitions), translated into ordering lists for the head chef and FOH/BOH rota suggestions for the GM, with a human-approval gate before any pay-affecting decision.

Knowledge Bot. Trained on the restaurant's menu, the 14-allergen matrix, dietary policies (vegan, vegetarian, halal, kosher, gluten-free, dairy-free, nut-free), head chef bio, wine and cocktail list, reservation rules, private dining capacity, accessibility and opening hours including bank holidays.

Lives on the site, Instagram DM, WhatsApp Business and the Google Business Profile chat surface; falls back to FOH on any allergen-sensitive query. Handles 60-70% of pre-booking enquiries (parking, dress code, kids policy, dog policy, set menu pricing, corkage).

Allergen handling stays human-in-the-loop under Natasha's Law: the bot surfaces the published matrix as structured fact, never substitutes a dish on the basis of an allergy claim, never confirms safety in free text, and escalates with the diner's explicit consent.

Growth Stack. Local SEO foundation tied to Google Business Profile (Posts, Q&A, Reviews response, photo refresh, attribute completeness), Bing Places for the rising Copilot tail, Apple Business Connect for Apple Maps and Spotlight, Tripadvisor optimisation, OpenTable and Resy listing optimisation.

Review management across Google, Tripadvisor, OpenTable, Resy, Deliveroo, Just Eat and Uber Eats. Schema markup AI engines extract cleanly: Restaurant, MenuItem with suitableForDiet and allergen properties, Reservation, FAQPage, Review aggregateRating, OpeningHoursSpecification with bank-holiday coverage, AcceptedPaymentMethod and hasMenu.

The compounding lift is shifting share of demand from rented marketplace surfaces (Time Out, SquareMeal, Tripadvisor) into the restaurant's own organic AI-citation footprint.

Type 'best Indian in Shoreditch', 'best ramen in Soho', 'best tapas Manchester Northern Quarter' or 'best Sunday roast near Borough Market' into Google today and the first page is Time Out, SquareMeal, The Infatuation, Tripadvisor, the Michelin Guide and the Google Maps Local Pack.

The independent restaurant surfaces only where it has invested in SEO and Schema. ChatGPT, Perplexity and Google AI Overviews answer the same prompts from the same editorial and marketplace sources plus Reddit and Wikipedia.

The restaurant's own site is invisible because it carries no Restaurant or FoodEstablishment Schema, no MenuItem markup with allergen properties, no FAQPage Schema, no Review aggregateRating, no OpeningHoursSpecification with bank-holiday handling. The editorial monopoly breaks the moment the restaurant's own site carries cleaner structured data than the editorial alternative.

The aggregator economics speak directly to the citation gap. Deliveroo, Just Eat and Uber Eats take 25-35% commission. A London restaurant generating 30% of revenue through aggregators surrenders 7-10% of total revenue to commission alone, and rented marketplace demand has no compounding gain; next month's number resets to zero.

The same restaurant, capturing two organic AI Search reservations a week against 'best vegan in Hackney' or 'business lunch Mayfair', pays back a Foundation plus AI Search retainer inside the first quarter and compounds month over month.

Aggregator dependence does not collapse, because Deliveroo and Just Eat are structural delivery rails. The centre of gravity shifts from rented marketplace surface to owned organic AI visibility, and that is the only lever that breaks the commission wall.

The 2024-2025 sourced-content cohort shows the lift pattern. A representative pattern from the Padella, Brat and Mountain orbit plus Hoppers, Manteca, Akoko and Lita: restaurants that publish sourced pages on head-chef provenance, ingredient sourcing (Cornish day boats, Lake District lamb, Norfolk asparagus, Cumbrian beef), dish-level descriptions with allergens and clean FAQ blocks get cited in ChatGPT and Perplexity answers for 'best {cuisine} in {borough}' prompts within weeks of publication.

Areza's wedge is industrialising that posture across the 1-15-site indie and small-group tier where the buyer is owner plus ops director plus GM and the contract signs on a Wednesday.