Hospitality · Guide

How to automate review responses for a restaurant

· 8 min read

Short answer: A restaurant automates review responses by connecting its review platforms — Google Business Profile, TripAdvisor, Booking.com — to a workflow engine like n8n, letting an LLM draft each reply in the restaurant's own voice, sending the draft to a manager's phone for one-tap approval on Telegram or WhatsApp, and publishing the approved reply back automatically. A do-it-yourself build takes 5–7 working days and roughly €40€80/month in tooling; done-for-you, it goes live in about four weeks from €800/month. In our single-venue Riga pilot, first reply time dropped from 4 hours 12 minutes to 38 seconds.

Below is the exact procedure, what each step needs, where humans must stay in the loop, and an honest manual-vs-automated comparison so you can decide whether to build it yourself or have it operated for you.

Why response speed is the lever, not response polish

Two things happen when a restaurant answers reviews fast and consistently. First, owner engagement is a signal Google weighs for local ranking — venues that respond promptly and to everything tend to surface higher in the map pack than venues with months of silence. Second, future guests read your replies before they book. A two-day-old complaint with no answer reads very differently from the same complaint with a calm, specific owner response posted the same afternoon.

The bottleneck has never been writing skill. Managers know what to say. The bottleneck is that reviews arrive on three platforms, nobody owns the checking, and replying is the task that always loses to service. Automation removes the checking and the drafting; the manager keeps the judgement.

The six-step system

Step 1 — Aggregate every new review in one place

A scheduled n8n workflow polls Google Business Profile through its official API and pulls TripAdvisor and Booking.com reviews from your public listings (a scraping service like Jina Reader handles the platforms without reply APIs). Every new review is deduplicated and written to a database — Supabase's free tier is enough for any single venue — with platform, rating, text and timestamp.

Step 2 — Classify by rating and sentiment

Not every review should follow the same path. 4–5 star reviews go to the fast lane: draft, approve, publish. 1–2 star reviews are flagged so the manager sees the complaint before any draft. Refund demands, allergy or safety incidents, and anything that smells legal skip drafting entirely and escalate straight to a human. This routing is what makes the system safe rather than reckless.

Step 3 — Draft the reply in your voice

An LLM (Gemini Flash on the free tier, or Claude when budget allows) drafts the response against a one-page tone guide: how you greet and sign off, phrases you always and never use, how you reference dishes and staff, your escalation script for complaints. The draft must mention something specific from the review — the dish they named, the server they praised — because generic thank-you boilerplate is worse than silence.

Step 4 — Approve from your phone in one tap

The draft lands on the manager's Telegram or WhatsApp with Approve / Edit / Reject buttons. Approval takes seconds between covers. Edits are logged and folded back into the tone guide, so the drafts get closer to your voice every week. This human step is non-negotiable: the AI proposes, a person accountable for the venue disposes.

Step 5 — Publish back automatically

Approved replies post to Google via the Business Profile API and to TripAdvisor and Booking.com through each platform's owner-reply channel. No copy-pasting, no logging into three dashboards, no "I'll do it after service."

Step 6 — Measure it weekly

Because every review and reply is logged with timestamps, you get the two numbers that matter without any extra work: response rate (should be near 100%) and median time to first reply (should be same-day or better). If either drifts, you see it Monday morning, not at the quarterly review.

Stack: n8n (self-hosted) + Gemini Flash or Claude + Jina Reader + Google Business Profile API + Telegram bot + Supabase. Time to ship (DIY): 5–7 working days. Tooling cost: ~€40€80/month.

Manual vs automated, honestly compared

Manual respondingAutomated (with human approval)
Time to first replyHours to days; backlogs of unanswered reviews are commonDraft in under a minute; published as fast as the manager taps Approve
Response ratePartial — whatever survives the week's prioritiesNear 100%; nothing is missed because nothing depends on remembering
Platforms coveredUsually Google only; TripAdvisor and Booking.com checked sporadicallyGoogle + TripAdvisor + Booking.com in one queue
Tone consistencyVaries by who replies and how tired they areLocked to a written tone guide; improves with every edit
Manager time per weekHours, or it simply doesn't happenMinutes of approvals from a phone
Monthly costStaff time (the most expensive input you have)€40€80 tooling DIY, or from €800 done-for-you

What results to expect

One controlled data point, stated plainly: in Streamflow's single-venue Riga pilot (four weeks), first review reply time dropped from 4 hours 12 minutes to 38 seconds, booking conversion improved 22% over baseline, and the manager reclaimed approximately 12 hours per week. Those are figures from one pilot, not industry averages — your numbers will depend on your starting backlog and review volume. The mechanism, though, is the same everywhere: faster, fuller responses feed local ranking, ranking feeds bookings, bookings feed reviews. It compounds.

The backlog is the real cost. A restaurant with a year of unanswered reviews isn't neutral — it's signalling to every reader, and to Google, that nobody is home.

Build it yourself or have it operated?

If someone on your team enjoys n8n and APIs, the six steps above are a one-week project — start with Google only, add TripAdvisor and Booking.com once the approval loop feels natural. If nobody wants to own the plumbing, this is exactly what our Engine A review automation service does: built, monitored and tuned for you, from €800/month for a single location ( €1,500€3,000/month for multi-property groups), live in about four weeks.

Not sure whether reviews are your highest-leverage automation? The full list of ten hospitality automations ranks the options, and the ROI calculator puts a yearly number on your current reply speed. Riga-area venue? There's a dedicated page for Riga hospitality with local pilot details.

Common questions

Does Google allow automated responses to reviews?

Yes. Google provides an official Business Profile API for posting owner replies, and automation through it is permitted. What Google's content policies require is that replies are genuine, relevant and non-spammy — which is exactly why a well-built system keeps a human approval step: the AI drafts, a manager approves, the reply publishes under the owner's account.

Can AI reply to negative restaurant reviews without making things worse?

Yes, if the system routes them correctly. 1–2 star reviews should never auto-publish: the AI drafts an apology that acknowledges the specific complaint, offers to take it offline, and never argues — then a manager reads and edits before approving. Refund demands, allergy incidents and anything legal-sounding should escalate to a human with no draft at all.

How quickly should a restaurant respond to a new review?

Within 24 hours is the standard worth holding; same-day is better. Fast, consistent owner responses are a positive engagement signal for local ranking and they are read by future guests deciding where to book. An automated pipeline produces a draft in under a minute, so the only delay left is how fast a manager taps Approve — in Streamflow's single-venue Riga pilot, first reply time dropped from 4 hours 12 minutes to 38 seconds.

Do I need a developer to set up review response automation?

For a do-it-yourself build: you need someone comfortable with n8n, API credentials and a database — a technically confident manager can do it in 5–7 working days following the steps in this guide. If nobody on the team wants to own the plumbing, an agency operates it for you; Streamflow's done-for-you Engine A starts at €800/month for a single location, live in about four weeks.

Prices shown in your local currency for reference; contracts invoiced in EUR.

Does review response automation work for TripAdvisor and Booking.com, not just Google?

Yes. Google is the cleanest integration because it has an official reply API. TripAdvisor and Booking.com reviews are collected by scraping the public listing (for example with Jina Reader) and replies are published through each platform's owner-reply channel. One pipeline, three platforms, one approval queue on your phone.

What should a restaurant's review reply tone guide contain?

Five things: how you greet and sign off; three phrases you always use and three you never use; how you reference specifics (dish names, service, atmosphere); the exact escalation script for complaints (acknowledge, apologise for the experience, move it to email or phone); and two real example replies you are proud of. One page is enough — the LLM drafts against it for every review.

Want this running for your restaurant in four weeks?

Free 30-minute audit. We look at your current review backlog and reply times across Google, TripAdvisor and Booking.com, and tell you exactly what an automated pipeline would change.

Book a Free AI Clarity Call →