How to Automate Review Requests and Replies for a Restaurant, Cafe or Local Service

For a restaurant, a cafe or any business people find on a map, the review page is the shop window. It is also the only part of the business updated entirely by other people, on their schedule, usually at the moment something went wrong.

Two jobs keep it healthy, and neither is complicated: ask every customer once, and reply to everything that gets written. Both take five minutes, and both stop happening the moment you get busy, which is exactly when they matter.

This guide automates the asking and drafts the replying. Setup takes about two hours.

What you are building

A short message that goes out after a visit or a job, asking for a review on one platform, with a direct link. Plus a second automation that watches for new reviews, drafts a reply in your voice, and puts it in front of you to approve. You still press send on replies, because replies are read by everybody who has not eaten with you yet.

Tools: your booking system, till or invoicing software, an automation tool (Zapier, Make or n8n), an AI assistant (ChatGPT or Claude), and the review platform you care about.

Step 1: pick one platform and get the short link

Ask for reviews in three places and you will get them in none. Choose the platform your customers actually check before deciding, which for most local businesses is the map listing, and put everything behind one link.

Every major review platform gives owners a direct review link in the business dashboard. Get it, shorten it, and keep it in the same note as the rest of this build. If a customer has to search for you, tap through two screens and hunt for the right button, most will simply not.

Step 2: choose the moment

The ask has to land close to the experience and after it is finished. Use the trigger you already have.

• ·A restaurant or cafe taking bookings: the morning after the reservation.
• ·A salon or clinic: a few hours after the appointment, once they have seen the result in daylight.
• ·A trades or services job: the day after the invoice is settled, never before.
• ·A shop with a loyalty scheme or an email list: after a second purchase rather than a first.

One ask per customer, ever, unless a year has gone by. A suppression list of everybody you have already asked is the difference between a review request and a nuisance.

Step 3: write the ask

The message has one job and should carry one link. Keep it shorter than feels polite.

Pick one of the three, send it to yourself, and read it on a phone. If the link sits below the fold on a phone screen, cut a sentence.

Step 4: wire the ask

One scenario, and the only fiddly part is the suppression list.

1. 1Trigger from the system that holds the moment you chose: a completed booking, a settled invoice, a till transaction with an email address attached.
2. 2Check the customer against a sheet of everybody already asked. If they are on it, stop there.
3. 3Wait until a civilised hour. Nobody wants a review request at ten past six in the morning.
4. 4Send the message through your email tool or SMS provider, filling in the customer name and what they bought or booked.
5. 5Add them to the suppression sheet with the date.

Run it against your own contact details for the first few days. A review request sent to somebody whose booking was cancelled is a very specific kind of embarrassing, and it is easier to catch in testing than to explain afterwards.

Step 5: reply to everything

Replying to reviews is public work. It is read by people deciding whether to book, and it is the cheapest way to show what happens when something goes wrong. Every review gets a reply, including the five-star one that says nothing at all.

Have the automation notify you when a new review appears, pass the text to your AI assistant, and produce a draft you can approve or rewrite.

Two kinds of review are not a drafting problem at all. One is abusive, or clearly about a different business, and that goes to the platform through its reporting route rather than into a public argument you cannot win. The other is a serious complaint about safety or hygiene, which needs a phone call from you the same day and a short public reply saying you have been in touch.

Read every negative draft before it goes out, without exception. The model writes a calm reply, which is right most of the time, but only you know whether the complaint describes a night you already know went wrong.

Step 6: use what customers wrote

Reviews are the only writing about your business in the words your customers actually use, which makes them the best raw material you have for content. Once a month, take everything that came in and pull out the themes.

Use those five things on your listing, your menu board and your posts. Never publish a quote you did not receive, and never tidy a customer quote into something they did not say.

What changes

The asking stops depending on whether the shift was calm. The replying stops being the job you catch up on once a quarter. And the page strangers use to decide about you gets tended every week by a system instead of by whoever remembers.

The asking half is Workflow 04: Get More Customers and the reply and content half sits in Workflow 06: Automate Your Content. Both are in the AutomateFirst playbook with the full prompt set and a checklist, seven workflows for one payment of $297.