AI Recommendations Are Now a Top Restaurant Trend - But Most UK Operators Still Haven't Protected the Brand Behind the Tech

The Algorithm at the Table

Published yesterday, restaurantonline.co.uk's new trends report, produced in partnership with SevenRooms, puts AI squarely at the centre of the UK dining story for 2026. Alongside the resurgence of food halls, the pull of Aegean cuisine, and the stubborn appeal of fried chicken, the uptake of AI for restaurant recommendations appears as one of the 11 trends shaping the sector this year.

That framing matters. Recommendation is no longer a side effect of a Google search. It's a primary channel. SevenRooms' own data shows that 22% of consumers now use AI tools to discover restaurants, typing queries rather than keywords, and that searches for "when to book a table" have climbed 140% in 2026. Meanwhile, 74% of diners say they are open to AI handling their reservations entirely. The front-of-house is being rewired, one API call at a time.

For operators, the commercial logic is blunt. An AI that fields every inbound call, logs the booking, and writes the reservation into a guest CRM without a bad shift is hard to argue with. Casper Hospitality put SevenRooms Voice AI on their phones and, in a single month, handled 3,800-plus calls and generated $28k in booked revenue. JKS Restaurants in the UK told SevenRooms that the most impactful use of AI for their group has been summarising qualitative guest feedback, removing human bias and surfacing service issues faster than any manual review process could.

Margin Is Where AI Actually Earns Its Keep

Beyond front-of-house, the shift is equally pronounced in the kitchen. Restaurantonline.co.uk's coverage of Crunchtime's operational AI makes the point plainly: even where sales remain stable, profitability is under pressure, and the difference between strong and underperforming sites often comes down to how well operations are executed shift by shift. AI connects inventory, sales, and execution data that currently sit in separate systems, each offering partial visibility but little actionable direction.

The results at scale are striking. Marriott UK, Ireland and Nordics reports a 50% reduction in food waste portfolio-wide using Winnow's kitchen AI. At the London Marriott Canary Wharf specifically, food waste dropped 67% in six months, saving 20,000 meals and 33 tonnes of CO₂ equivalent annually. For a sector where margins are eroded quietly, shift by shift, those numbers are not abstract.

Zonal's product director Glenn Tait describes his company's approach as "human in the middle": AI outputs presented as options or recommendations, backed by data, rather than final decisions. That positioning is becoming the industry standard. Operators want AI that reduces the volume of manual decisions, not one that removes human judgement entirely.

A New Incorporation Wave, Without the IP Groundwork

Here is where the picture gets more complicated. AIBD analysis of Companies House data shows 1,678 new SIC 56.10 (restaurants and mobile food service) companies registered in 2026 Q3, a sharp contraction of 82.9% against the prior period. Fewer new entrants, yes, but those that are launching are doing so into a market increasingly mediated by AI discovery, where brand name recognition matters more than ever.

Trademark behaviour tells a cautionary story. UK Intellectual Property Office data for the same period shows only 120 Class 33 filings (spirits and alcoholic beverages) from across the entire food and drink cohort, a drop of 82.2%, suggesting brand registration activity is cooling in step with incorporation rates. More telling: 98.8% of active SIC 56.10 companies currently hold no Class 33 trademark at all. (Source: AIBD analysis of Companies House and IPO data, as of July 2026.)

Class 33 is just one dimension of brand protection. Nice Class 43, which covers restaurant, café, and catering services, is the foundational filing for any food-service operator. But the pattern holds. Operators are building AI-assisted guest journeys, optimising menus with machine learning, and investing in voice reservation systems, while skipping the step that would prevent a competitor from trading on the brand equity all that technology is building.

In a world where AI recommendation engines surface restaurant names based on structured data, verified listings, and brand signals, an unregistered name is a liability. Google's Gemini now answers natural-language queries like "find boutique restaurants with great vegan menus near Shoreditch" using schema data and verified listings. If your brand isn't anchored by a registered mark, it becomes considerably harder to challenge anyone else who decides to occupy adjacent territory.

Dark Kitchens, Deeper Risk

The exposure is sharpest in the dark kitchen segment, where brand is the only customer-facing asset. There is no shopfront, no ambience, no service team to carry a name. The UK food delivery market is projected to reach £11.2 billion by 2026, and Europe's dark kitchen market is estimated at $7.4 billion this year alone. The operators building virtual brands on Deliveroo, Uber Eats, and Just Eat are entirely dependent on app visibility, which increasingly means AI visibility. A brand that is undifferentiated in trademark terms is a brand that can be copied, diluted, or outranked without legal recourse.

Deliverect, whose autonomous agents launched in the UK first in 2026, now optimises menus and protects digital revenue by catching sync errors across platforms in real time. The tech stack is maturing fast. The legal infrastructure protecting the brands that run on top of it is not keeping pace.

What Operators Should Take From This

The conversation around AI in UK hospitality has moved, decisively, from experimentation to practical application. That shift is real and the data supports it. But AI amplifies whatever brand equity already exists. It does not create it from nothing, and it cannot defend it.

New entrants registering under SIC 56.10 this quarter are stepping into a market where AI tools surface names, route bookings, manage inventory, and respond to reviews. The operators winning on margin are the most connected: every system talking to every other, less manual work between them. The operators winning on brand will be the ones who registered before the algorithm decided their name was worth copying.