The question we study
We study the gap between being online and being genuinely bookable by AI
Discovery moved to AI assistants faster than local businesses moved with it. Having a website, a Google profile and a booking widget no longer means an assistant can actually complete a booking. G-Lab Research measures that distance, market by market, and publishes what it finds along with how it was measured.
How the measurement is named
One instrument, three scopes. Keeping them distinct is deliberate: it is what lets a single business result and a whole-market result live side by side without either losing its meaning.
One business
Agent-Ready Booking Score
A 0–100 result for a single business: can it be found, understood and booked by an AI assistant? This is what the free diagnostic produces.
Methodology →A market
Agent-Ready Booking Index
The same measurement applied to a whole city, country or vertical, so a market can be compared with itself over time and with other markets.
About the Index →The publisher
G-Lab Research
The studies, datasets and press materials. Findings are published with the sample, the method version and the limitations attached.
Press materials →Studies
The Amsterdam Restaurant Agent-Ready Booking Index 2026
In progressCan an AI assistant actually complete a restaurant booking in Amsterdam? We are running the first city index now: a fixed sample of independent restaurants, one identical set of checks, an agent walking each public booking path to the final confirmation step and stopping there.
The sample rules, the date of collection and the method version will be published with the results. No figures appear on this page until the run is complete and a share of the rows has been re-checked by hand.
Next after Amsterdam: a second European city, then the first comparative report. If you are a journalist or researcher and want to be told when a study publishes, or you run a business you would rather we excluded, write to hello@g-lab.studio.
How we work
Method before findings
The methodology is public and versioned before a study runs, so nobody has to take a headline on trust. If the method changes, the version changes.
No phantom bookings
Our agent walks a booking path to the final confirmation step and stops there. We never submit a reservation a business did not agree to. Measuring readiness must not cost a restaurant a table.
Aggregates in public, detail in private
Studies publish distributions, not league tables of named low scorers. A business can always request its own full result privately, and can ask to be excluded.
The headline follows the data
We write the finding after the run, never a target before it. A number we cannot source or reproduce does not ship.