Case study
Getting found for dinner in Canary Wharf
London-based wine and restaurant chain
The local Canary Wharf where-to-eat slice moved from an agreed baseline of 1.4% to 10.77%, up 9.37 percentage points and above the 3% target. The broader dining measure did not improve.
Measured visibility in AI answers
1.4% → 10.77%Local where-to-eat visibilityLocal visibility target: 3%Agreed starting point and seven-day report through a partial September 7, 2026 capture.The program answered practical guest decisions with 12 guides across the wider brand, then measured local Canary Wharf discovery separately from the broader London dining picture. That branch-level view revealed a strong local result alongside a missed overall target.
Measurement window
What exactly was measured?
- Agreed baseline
- 1.4% for the local Canary Wharf where-to-eat slice. The source record does not report the exact baseline window dates.
- Latest comparison
- 10.77% in the latest seven-day cut available on the partial September 7, 2026 review. The exact comparison-window dates are not reported.
Work
What changed during the program?
- 01
Guest questions were separated by branch and occasion, including dinner, after-work drinks, groups, and booking decisions in Canary Wharf.
- 02
Twelve practical guides were live across the wider brand by August 26, giving assistants clearer material for location-led recommendations.
- 03
Local progress was read beside the broader dining measure, so one strong slice could not be presented as improvement everywhere.
Inside the program
The challenge was a local dinner decision
A London restaurant group can be recognized as a brand and still be missed when a guest asks where to eat in one neighborhood. The practical goal was to become more useful in Canary Wharf decisions, where place, time, and group needs shape the answer.
The agreed target for the local where-to-eat slice was 3% from a baseline of 1.4%. This was a narrow local goal, not a promise that every dining question across London would rise with it.
Inside the program
Content followed the questions guests use
The work focused on decisions a guest can act on: dinner, after-work drinks, group plans, and whether to book. One guide brought those Canary Wharf considerations together without turning the page into a list of search phrases.
By August 26, 12 program guides were live across the brand. The wider library created useful context for different venues and occasions while the local guide supplied branch-specific evidence.
Inside the program
The local slice moved beyond its target
The latest reported local reading was 10.77%, up 9.37 percentage points from the agreed 1.4% baseline and above the 3% target. The chart shows only this local where-to-eat measure; it does not represent visitors, bookings, or revenue.
The program tracked 55 questions across London and Canary Wharf in ChatGPT, Google AI Overviews, and Gemini. The record does not establish that all 55 remained one fixed cohort, or that 55 questions belonged to this branch alone.
Inside the program
The broader dining result was mixed
The broader dining measure moved from 51% to 50.70% and missed its 57% target. That result matters because the local improvement cannot be used as shorthand for stronger visibility across the whole program.
Separating the branch slice exposed the useful finding: focused local content can coincide with a large movement in one neighborhood even when the broader category stays flat. The measurement needs both views to remain honest.
Inside the program
What another multi-location group can take from it
Measure the guest decision at the level where it happens. A group-wide average can hide a location gaining relevance, just as one strong branch can hide a weak wider picture.
Build guides around real choices and review each location against an agreed baseline and target. Use the result to decide where evidence is still thin, rather than treating one movement as proof of commercial impact.
Prompt set
Did the question set stay fixed?
The program covered 55 questions across London and Canary Wharf. The source does not establish one unchanged 55-question cohort across every comparison, and it does not assign all 55 questions to the Canary Wharf branch.
Engine coverage
Which engines were included?
The program monitored ChatGPT, Google AI Overviews, and Gemini. The local 10.77% figure is an AI-visibility reading, not a count of diners or website visits.
Definitions
How are the measures defined?
- Local where-to-eat visibility
- The reported AI-visibility reading for the Canary Wharf where-to-eat question slice.
- Broader dining visibility
- The separate reported measure for the wider dining question set.
- Agreed target
- A program benchmark used to judge movement, not a forecast of bookings or revenue.
Interpretation
What are the limits of this result?
- The exact baseline and comparison-window dates were not reported, so none are inferred here.
- The latest seven-day cut formed part of a partial September 7 review rather than a final program evaluation.
- The wider dining measure fell from 51% to 50.70% and missed its 57% target.
- The record supports visibility reporting only. It does not support claims about bookings, traffic, revenue, or causation.
Next step
Measure the local questions that lead to a visit
The free audit separates brand-wide visibility from the location and occasion questions guests use when deciding where to go.