Home / Tasks / Task 1

Can an AI agent book a restaurant in Australia?

We ask AI agents to book a table for 2 at 7pm on Saturday on the Gold Coast using OpenTable or the restaurant's own booking page. Here's the exact prompt and how we score it.

Round 1 results

Testing in progress. Round 1 results drop mid-November 2026.

AgentResultNotes
ChatGPT WorkTestingResults pending
ManusTestingResults pending
OpenClawTestingResults pending
Gemini in ChromeNot available in AUAgent features are US-only

The exact prompt

Every agent gets this word for word, in a fresh session, logged in to the same test accounts, from a Gold Coast location.

Prompt · Task 1Book a table for 2 people at 7:00pm this coming Saturday at a well-reviewed Italian restaurant in Broadbeach or Surfers Paradise on the Gold Coast, Queensland. Use OpenTable or the restaurant's own online booking system. Use the name and contact details from the logged-in account. Do NOT confirm the booking: stop on the final confirmation screen and tell me the restaurant, date, time and party size you would submit.

Sites involved: OpenTable, Restaurant booking pages. Where we stop: Final 'Confirm booking' button. Never pressed.

How we score it

Pass

Reaches the final confirm screen for a real Gold Coast Italian restaurant in Broadbeach or Surfers Paradise with 2 people, 7:00pm (or tells us 7pm is unavailable and offers the nearest time), on the correct Saturday.

Partial

Gets there with 1-2 nudges, or makes one minor error (e.g. wrong suburb nearby, 7:15pm without asking).

Fail

Gets stuck, books the wrong date/party size, says it can't do bookings, or claims the table is booked when it isn't.

Logging in and 2FA codes don't count as help. Full rules on the methodology page.

Fill a grocery cart →