🌟 In today’s Issue
This weekly dispatch is specifically designed for restaurant and small business owners who are trapped in the daily grind and ready to move from "struggling artist" to strategic operator.
I got something wrong in front of seventeen operators a fortnight ago. This issue is the correction, and it turned out to be bigger than the mistake
Bookable By a Machine... Google's assistant has been booking Australian restaurant tables since April. No announcement you would have seen. No opt-in you could have ticked.
Strategic Marketing:
Since 10 April... Agentic restaurant booking in Google's AI Mode expanded to eight markets including Australia and New Zealand [1]. Five months ago. Nobody in the sector is talking about it…
The List Everybody Is Repeating Is the Wrong List... The eight booking partners named in the coverage come from Google's United Kingdom announcement [2]. One of them shut its Australian operation in 2024. Another is a British card-payments company…
You Cannot Opt In... Google's own words: Reserve with Google works only with businesses using a supported scheduling provider [4]. The provider does the integration, not you. A venue on a phone and a notebook has no route in at all…
Practical AI Implementation:
The Booking Rail Audit... Whether your availability is actually visible to a machine, and who would know…
The Availability Fact Sheet... Your hours, sittings, walk-in policy and deposit rules, written so a machine can quote them…
The Paper Venue Escape... For the venue that takes bookings in a book. The smallest honest step, without buying software you do not need…
Actionable Growth Tactic:
The Ten-Minute Test... Someone who is not you, a phone, and one question. Then the two venues nearest you…
The Savvy Operator Mindset:
From Taking Bookings to Being Bookable... A booking you can only take when the phone rings is not availability. It is a queue for your attention…
The Booking You Never Saw.
It is Saturday. Two covers at 7:15, a name you do not recognise, and the table turns up on time.
You check later, because you always check. Nothing in your analytics. No click from Google. No enquiry in the inbox. No missed call.
They did not go to your website. They asked an assistant where to eat, it found a table, and it booked it. The whole conversation happened somewhere you cannot see, about you, without you.
That is either very good news or very bad news, and which one depends entirely on something you have probably never checked.
First, The Correction.
A fortnight ago I told a room full of operators that agentic booking was not live in Australia.
That was wrong, and it had been wrong for five months.
On 10 April 2026 Google expanded agentic restaurant booking in AI Mode from the United States to eight new markets. Australia and New Zealand were on the list, along with Canada, Hong Kong, India, Singapore, South Africa and the UK. No Labs opt-in. No subscription [1].
I want to be precise about how solid that is, because being precise about it is the entire point of this newsletter.
Google published a dedicated announcement for the UK. It published one for Canada and one for South Africa. It published nothing specific to Australia.
The source naming our market is a post by Google's vice president of product management, on the day. Plus the technology press reporting the full list of eight [1].
So: firm enough to act on, and I am telling you exactly how firm rather than rounding it up.
Here is why I am opening with my own mistake. I sell measurement. A business that sells measurement and quietly edits its errors is selling something it does not have. Fifty-four issues in, this is the first correction I have printed. It will not be the last, and the day it becomes the last is the day you should stop reading.
Then I checked the rest of it, and found a second thing wrong.
Every piece of coverage repeats the same eight booking partners. TheFork, SevenRooms, ResDiary, Mozrest, Foodhub, Dojo, DesignMyNight and OpenTable.
That list is real. It comes from Google's own announcement [2].
Google's own United Kingdom announcement.
Nobody has published an Australian partner list, and when you check the eight against this country, the list stops making sense.
TheFork closed its Australian operations in March 2024 [3]. It is not a booking option here and has not been for two and a half years.
Dojo is a British card-payments company [3]. It does not take restaurant reservations anywhere, including Britain.
Foodhub is takeaway ordering, not table reservations [3].
Mozrest is UK and European reservation middleware, with no Australian presence I could verify [3].
Of the eight, the ones that clearly operate here are OpenTable, SevenRooms, ResDiary and DesignMyNight.
I am not telling you this to be clever about somebody else's article.
I am telling you because of what happens next. You ring your booking provider, you ask "are we on Google's partner list", and they say "what list". The list you are holding may simply not be about your country.
Ask the better question instead. It is three sections down.
Strategic Marketing: Three Rungs, and Most Venues Are Stuck on the First.
The last three issues have been climbing one ladder, and it is worth naming now that we are at the top of it.
Rung one. Does the machine know you exist? That was Issue #51. Three questions, one assistant, twenty minutes, an honest number.
Rung two. Does it know your facts? That was Issue #52. Your hours, your prices, your quiet hour, written down as plain sentences a machine can lift.
Rung three. Can it actually seat someone? That is this issue. And it works completely differently from the first two.
Rungs one and two you can earn. Write better answers, publish real prices, collect evidence, and over months the machines have more to work with. Slow, compounding, in your control.
Rung three you cannot earn. There is no amount of good copy that puts a bookable table inside a Google answer.
Here is Google's own wording, and it is unusually plain. Reserve with Google is available only in certain countries or regions. And only for businesses that work with a supported scheduling provider [4].
Read that twice. The requirement is not quality. It is plumbing.
Google's documentation for the providers themselves makes it plainer again. The booking platform must hold a direct contractual relationship with every venue it submits. It must match each venue's data to its Google Maps location. And it must meet Google's service definition [5].
The provider does the work. You are a row in their feed.
Which means the venues that got included in April did not win anything. Nobody wrote a word to earn it. Their availability already existed as a machine-readable record inside a partner system, so it was there when the assistant went looking.
That is the same sentence as Issue #52, arrived at from the other end. A machine cannot repeat a fact you have not written down. And it cannot book a table that exists only in your head.
And if your bookings live in a book?
This is the part I expected to have a workaround for, and I do not.
Google's advice to a venue with no booking software is to go and get some. Its own words:
If you do not work with a provider to manage your bookings, you can find providers available in your country through Reserve with Google. [6]
There is a lighter option, and it is worth knowing what it is and is not. You can put a booking link on your Google Business Profile that points at a page on your own site. That is a link, not a bookable action. The assistant cannot read live availability from it, cannot hold a table, and cannot complete anything. It sends a human to a page and hopes.
A venue taking bookings by phone and a notebook has no native route into an agentic booking at all. Not a bad one. None.
I am not going to turn that into a recommendation to buy something, because I do not know your covers, your margin or your February numbers. What I will say is that this has stopped being a software preference and started being a question about which conversations you can appear in.
The size of the room, and it is smaller than the headlines suggest.
Before anyone panics, some proportion.
SOCi published a restaurant-specific breakout in August. The average restaurant brand reaches Google's local 3-Pack 24.3% of the time, is recommended by ChatGPT 5.3%, Gemini 7.1% and Perplexity 7.6% [7].
Note those are the restaurant figures. The numbers you will see quoted in most articles are cross-industry, and they flatter the map pack and understate ChatGPT for a venue.
So this is not most of your bookings and it is not going to be next month. It is a door that opened in April. You cannot walk through it by trying harder. And a competitor two streets away may already be standing in it without knowing.
The two that are not here yet, and why they matter anyway.
Yelp put reservations and waitlists inside ChatGPT on 10 August. United States and Canada [8].
Google's Ask Maps began taking food orders through Square and Toast on 6 August. United States only [9]. The conversational part of Ask Maps is live here; the ordering is not.
Three transactional surfaces in five months. Two of them not in Australia. One of them here since April and unreported.
The pattern is what matters. Every one of them worked the same way — the venues included were the ones whose inventory already sat in a connected system. Not the best venues. The readable ones.
One more, because it explains why this is arriving faster than it feels. AlsoAsked measured 19.2 million English-language queries this year. In the first week of September, AI Overviews answered about 97% of People Also Ask results [10]. In August it was 86%. Fourteen months ago it was 12%.
The last plain slot on a search page closed over during winter. The answer is now the page.
The Brutal Math, and the honest problem with it.
I cannot tell you what this costs you, and that is the finding rather than a dodge.
A booking that never happened leaves no trace. No impression, no click, no abandoned form, no missed call. It is the only marketing loss in your business that is completely invisible, which is precisely why it has gone unexamined for five months.
So here is a floor, clearly labelled as illustration.
Tyro puts average spend per visit in Sydney at $32.28 [11]. Say the rail costs you one two-top a week. That is about $65 a week, or $3,400 a year.
Two a week is $6,800. Four is $13,600.
Pick your own number. The point is not the figure. The point is that you currently have no way of knowing which line of that table you are on, and ten minutes fixes that.
Practical AI Implementation: The Booking Rail Desk
Same working folder as the last three issues. If you built the Visibility Desk or the Golden Hour Desk, this sits beside them.
Create the project. Open Claude. Call it The Booking Rail Desk.
Feed it five things. The name of your booking system, or a note saying you have none. A screenshot of your Google Business Profile as a customer sees it. Your opening hours, including public holidays and any closures. Your table sizes and how many sittings you run. Your walk-in, deposit and cancellation policies, in whatever rough form they exist.
That last one is usually the hardest, because most venues have never written it down. That is the issue in miniature.
🤖 AI PROMPT #1: The Booking Rail Audit
Whether a machine can actually book you, and who would know.
[TASK TITLE/GOAL] Audit whether my venue is bookable by an assistant.
1. Role & Expertise (Function): You are a restaurant technology analyst who understands how reservation platforms connect to search and assistant surfaces. You explain plumbing in plain English to owners who are not technical. You never recommend a specific platform to buy.
2. Context & Background (Pre-loaded / Specific to Task): My Business: [Insert name, suburb, seats, cuisine]… My Booking System: [Insert the platform name, or write "none — phone and a book"]… My Google Business Profile: a screenshot is in this project… My Website Booking Route: [Insert what happens when someone clicks book on my site]… My Volumes: [Insert roughly how many bookings a week and how they arrive]…
3. Task Description & Output Requirements (Function & Modifiers): Give me one markdown table with a row for each booking route a customer could take — my website, my Google profile, my booking platform's own page, a phone call, and walk-in. For each row show what a person sees, what a machine could read, and whether live availability is exposed. Then list the exact questions to put to my booking provider in writing, each with a yes-or-no answer I should expect. Under 500 words. End by naming the single weakest link in my booking chain and why.
4. Work Order: First, trace each route end to end from a stranger's point of view… Then, mark which routes expose real-time availability rather than a contact form… Next, write the provider questions so each one can only be answered yes or no… Finally, rank the routes by how much of my booking volume they carry today…
5. Examples: Route: Google Business Profile. What a person sees: a Book button. What a machine could read: depends whether my provider submits a feed. Live availability exposed: unknown — this is question one for my provider…
6. Warnings/What to Avoid (Modifiers): Do not tell me which booking platform to buy. Do not guess whether my provider is integrated — tell me to ask and give me the words. Do not invent a figure about my booking volumes. Do not describe a link to my website as a bookable action, because it is not one.
🤖 AI PROMPT #2: The Availability Fact Sheet
The facts about getting a table, written so a machine can quote them.
[TASK TITLE/GOAL] Write my availability facts as machine-readable text.
1. Role & Expertise (Function): You are a plain-English writer who turns operational policy into short factual sentences that a search engine or assistant can lift without changing the meaning. You write the way a good host answers a question at the door.
2. Context & Background (Pre-loaded / Specific to Task): My Business: [Insert name, suburb]… My Hours: [Insert, including public holidays and seasonal closures]… My Sittings: [Insert how many and at what times]… My Tables: [Insert sizes and how many of each]… My Walk-In Policy: [Insert the truth, including how long people usually wait]… My Deposits and Cancellations: [Insert, including group thresholds]… My Booking Route: [Insert how someone actually books]…
3. Task Description & Output Requirements (Function & Modifiers): Write twelve answer blocks of 40 to 60 words each, on getting a table specifically — walk-ins, group bookings, deposits, cancellations, last sitting, public holidays, dietary notice, high chairs, accessibility, large groups, private hire, and what happens when you are full. Each must open with a complete sentence that answers the question without the heading above it, and carry one checkable fact: a time, a number of seats, a dollar amount, a notice period. One markdown table with a suggested page for each. Under 800 words. End by naming the block I am most likely to have got wrong and why.
4. Work Order: First, rewrite each topic as the question a customer actually asks… Then, answer it in one standalone sentence… Next, add the detail that makes it usable, with a real number in it… Finally, flag any block where I have not given you a fact and you would have to invent one…
5. Examples: Question: Do you hold tables for walk-ins on a Saturday? Block: Yes. [Venue] keeps 8 of its 60 seats for walk-ins from 5:30pm on Friday and Saturday. The bar seats another 10 without a booking. Groups of 6 or more should book, as the courtyard is held for reservations after 7pm.…
6. Warnings/What to Avoid (Modifiers): Do not write marketing copy. Do not use the words welcoming, vibrant or unforgettable. Do not invent a policy, a price or a notice period — flag it and ask me. Do not write a block that only makes sense underneath its heading. Do not soften a policy to sound friendlier than it is, because the machine will quote it and a guest will hold me to it.
🤖 AI PROMPT #3: The Paper Venue Escape
For the venue that takes bookings in a book, and may not need software at all.
[TASK TITLE/GOAL] Work out my smallest honest step towards being bookable.
1. Role & Expertise (Function): You are a hospitality operations adviser who is sceptical of software a venue does not need. You have seen small venues ruined by systems that cost more in attention than they returned. You never recommend a specific product by name.
2. Context & Background (Pre-loaded / Specific to Task): My Business: [Insert name, suburb, seats]… How Bookings Arrive Now: [Insert — phone, walk-in, email, social messages, and roughly what share each]… Who Answers The Phone: [Insert, and when]… What I Have Tried Before: [Insert any booking system you have used and why you stopped]… My Constraint: [Insert what you will not do — no monthly fee, no tablet on the pass, no staff training]…
3. Task Description & Output Requirements (Function & Modifiers): Give me three options in one markdown table, ordered from doing nothing to a full booking system. For each, state what it changes about who can book me, what it costs in money and in attention per week, what breaks if it goes wrong, and who at my venue would have to own it. Be honest that option one, changing nothing, is a legitimate choice for some venues and say for which. Under 600 words. End by naming the one option you would pick for a venue of my size, and the single thing that would change your answer.
4. Work Order: First, work out what share of my bookings currently depends on a human being free to answer… Then, describe what each option changes for a stranger who never rings… Next, price each in weekly attention, not just dollars… Finally, name what would have to be true for the cheapest option to be the right one…
5. Examples: Option: change nothing. What it changes: nothing — I stay invisible to assistant booking, and keep every booking on the phone. Cost: nil. Right for: a venue that is full most nights on regulars and does not want more strangers…
6. Warnings/What to Avoid (Modifiers): Do not name a product or a brand. Do not assume more technology is better. Do not tell me a monthly fee is small, because you do not know my margin. Do not present being bookable by a machine as the only goal — a full room is the goal.
Pro Tip: Take this AI-generated content and add your personal touch. Change a word here, add a local reference there, include a quick story about a regular customer. The AI does the heavy lifting; you add the soul.
Actionable Growth Tactic: The Ten-Minute Test
This is the cheapest thing in the issue and the one most likely to actually change something.
Here's how it works:
Step 1: The Activation.
Action. Find someone who is not you. A partner, a regular, your chef's teenager. Someone whose phone is not logged into your accounts.
Ask them to open Google, go to AI Mode, and type the sort of thing a stranger types. Where can I book a table for two in [your suburb] this Saturday around seven.
Then watch what happens over their shoulder and say nothing.
Does your name come up? If it does, is there a way to book, or just a link? If there is a way to book, follow it until the moment before it would confirm, and stop.
Screenshot every step with the date visible.
Then do exactly the same for the two venues nearest you.
Ten minutes, and most operators have never once seen their own booking flow from the outside.
Step 2: The Collaboration.
Action. One email to your booking provider, today, with four questions in it.
Are we submitted to Google's Reserve with Google as a bookable venue? Is our live availability exposed, or only a link to your site? Which markets does that cover — is Australia included? And is there anything at our end you need in order for it to work?
Write down what they say and the date they said it. If the answer is vague, that is an answer.
If you have no provider, ask yourself one question instead. What share of my bookings would I lose if the phone went unanswered for a week? That number is the real size of this decision.
Step 3: The Transformation.
Action. Publish the twelve answer blocks from Prompt #2 on your website and your Google profile, in plain text.
That is not a workaround for the booking rail and I am not pretending it is. It is rung two, done properly, which is the only rung entirely within your control.
Then log it. One line: the date, what you found in the ten-minute test, what your provider said. Put a reminder in for the first Saturday of December and run the same test again.
Now the honest part. Nothing in this issue fills your room this month.
It is a door most venues have not looked at. The surface behind it carries a small share of bookings today and a larger one later. The reason to look now is that it costs ten minutes now, and a competitor's head start later.
Savvy Operator Mindset: From Taking Bookings to Being Bookable
The struggling artist takes bookings. The phone rings, someone writes a name in a book, and the system works exactly as well as the person nearest the phone. When it is quiet they assume nobody wanted a table. They have never considered that somebody tried, somewhere they could not see, and was quietly handed to somebody else.
The Savvy Operator is bookable. Not busier, not better at marketing. Reachable by every route a stranger might take, including the ones that do not involve a human being at either end.
Here is the whole thing in a line.
A booking you can only take when the phone rings is not availability. It is a queue for your attention.
Last month we said the visible thing is never the thing, and the thing with memory is. Then that a machine cannot repeat a fact you have not written down. This is the third one, and it is the sharpest. A machine cannot book a table that exists only in your head
Thinks quiet means nobody wanted a table. | Knows some of them tried somewhere invisible. |
Has never seen their own booking flow as a stranger. | Ran the ten-minute test and screenshotted it. |
Assumes the provider has it handled. | Asked four questions in writing and kept the answer. |
Keeps the walk-in policy in their head. | Wrote it down where a machine can quote it. |
Will find out when a competitor mentions it. | Found out in September, for ten minutes. |
One last thing, and it is the reason this issue exists at all.
I published a wrong thing, I found out, and this is me saying so with the date on it.
The correction turned out to be worth more than the original. Chasing it down is how the partner-list problem surfaced, and that one nobody else has published.
If this newsletter is worth anything to you, it is not because I am always right. It is because you will hear it from me when I am not.
Your Next Move: Ten Minutes and Somebody Else's Phone
Find someone whose phone is not signed into your accounts. Ask them to look for a table at your venue for Saturday, in AI Mode, the way a stranger would.
Watch. Screenshot. Write down the date.
Then send four questions to your booking provider.
That is the whole job, and you will know more about how strangers reach you than you did this morning.
How We Can Work Together
Here is the honest problem with everything above. It is another job.
You already run a kitchen, a roster, a supplier list and a family. Nobody needs a fourteenth tab open.
So do not run it yourself. Set your AI up as a second brain for the business and let it hold the detail.
Not a toy. Not a chatbot you poke when you remember it exists.
A working folder that knows your menu, your prices, your hours, your reviews and your suburb. And the twenty questions your team answers on the phone every week.
Feed it those things once and it stops handing you ideas. It starts handing you a scoreboard.
It becomes the quiet operator in the back room. Three weeks ago it watched the machines deciding whether four people on a footpath ever hear your name. Then an hour you had written off. Last week a statement you had never fully understood. This week it checks whether a stranger can book you at all.
That is why I built Strategic AI Marketing. We set your AI up properly, on your real numbers, so it thinks with your business instead of guessing at it.
You do the cooking. It does the remembering.
Two ways to start this week.
One, and it is free. Get your AI Visibility Snapshot. Four assistants, twelve questions about your venue, written down and dated. You will know exactly where you stand by the end of the week. It is at strategicai.marketing/free-ai-marketing-snapshot.
Two, if the booking question is the one nagging at you. Reply to this email with one word: BOOKABLE, and tell me what you take bookings on.
I will send back the four questions to put to your provider, written so they can only be answered yes or no. Plus what each answer actually means for you.
No pitch deck. No discovery call unless you ask for one. One reply, one useful answer.
Next Week on The Savvy Operator: A Special Edition. One year, fifty-four issues, and every prompt I have ever given you in one place — sorted by what you are actually trying to fix. The one to keep.
Until then, remember. A machine cannot book a table that exists only in your head. Serve your community. Be the operator.
Talk soon,
Rowan Shead
The Editor
The Savvy Operator
Owner of Strategic Ai Marketing
PS. You know another restaurant owner who's staring at empty tables right now wondering what they're doing wrong. They're not doing anything wrong. They just can't see what you just saw. Forward this newsletter to them. It takes four seconds and it might save them thousands. |
PPS. Those 37 restaurant owners I've worked with? They didn't come to me talking about "digital marketing strategy." They came to me talking about the hollow ache of a half-empty dining room on a Friday night. The 3am calculator spiral where you keep re-running the numbers hoping they'll change. The feeling of captaining a ship that takes on water faster than you can bail. |
If you're tired of fighting the digital war alone |
The Savvy Operator
References
[1] Google. Agentic restaurant booking in AI Mode expanded on 10/04/2026 from the United States to eight markets: Australia, Canada, Hong Kong, India, New Zealand, Singapore, South Africa and the United Kingdom. No Labs opt-in and no subscription required. Sourcing note, stated plainly: Google published market-specific announcements for the UK, Canada and South Africa. It published no Australia-specific announcement. The market list naming Australia comes from a post by Rose Yao, Google's vice president of product management, on 10/04/2026, and from technology press reporting the same day. Firm enough to act on. Not a Google blog post about Australia.
[2] Google. Use Google Search AI Mode to book restaurants in the UK, blog.google, 10/04/2026. Names booking partners TheFork, SevenRooms, ResDiary, Mozrest, Foodhub, Dojo, DesignMyNight and OpenTable. This is Google's United Kingdom partner list. No Australian partner list has been published by Google or anyone else.
[3] Company sources, checked 14/09/2026. TheFork closed its Australian operations in March 2024. Dojo is a United Kingdom card-payments business, not a reservations platform. Foodhub is a takeaway ordering platform rather than a table-reservation service. Mozrest is UK and European reservation middleware with no Australian presence found. OpenTable, SevenRooms, ResDiary and DesignMyNight operate in Australia.
[4] Google. Reserve with Google support documentation: available only in certain countries or regions, for businesses that work with a supported scheduling provider.
[5] Google. Actions Center reservations documentation for booking providers: the provider must hold a direct contractual relationship with every merchant in its feed, match merchant data to Google Maps locations, and meet Google's standard service definition. The integration is done by the platform, not by the venue.
[6] Google Business Profile support: if you do not work with a provider to manage your bookings, you can find providers available in your country through Reserve with Google. A custom booking link on a profile is a link to a page, not a bookable action, and does not expose live availability.
[7] SOCi. 2026 Local Visibility Index, restaurant breakout published 17/08/2026. The average restaurant brand appears in Google's local 3-Pack 24.3% of the time, is recommended by ChatGPT 5.3%, Gemini 7.1% and Perplexity 7.6%. These are the restaurant-specific figures. SOCi's widely quoted cross-industry numbers are different and are not about a venue.
[8] Yelp. Reservations and waitlists inside ChatGPT, announced 10/08/2026, United States and Canada. Not available in Australia.
[9] Google. Order food in Ask Maps, blog.google, 06/08/2026. Agentic food ordering through partners including Square and Toast, rolling out in the United States with more countries over time. The conversational features of Ask Maps are live in Australia. The ordering is not.
[10] AlsoAsked / Mark Williams-Cook. 19.2 million English-language queries measured across 2026. AI Overviews appeared on approximately 97% of People Also Ask results in the first week of September 2026, against 86% in August and 12% fourteen months earlier.
[11] Tyro. Eat Pay Love 2026. Average spend per visit, Sydney: $32.28. Australian data. The one-two-top-a-week figure in The Brutal Math is a deliberately conservative illustration, not a measured loss.
Strategic AI Marketing's own citation record — 0/12 on 26/05/2026 and every reading since — is published at strategicai.marketing/we-did-this/.