Automating Google review requests is allowed. Review gating is not. That single distinction is what decides whether your review flow compounds for years or gets your listing filtered, and most businesses get it wrong because their review tool shipped with gating turned on by default. Google's own guidance tells businesses to ask customers to visit a review link or scan a QR code — it has no opinion on whether a human or a workflow presses send. What it does prohibit is deciding who gets asked based on how happy they are.
We build these flows for local service clients around Dallas — home services, clinics, legal, med spas — and the compliant version is genuinely simpler than the gated one. Fewer branches, fewer tools, better numbers. Here's the whole thing.
Is automating Google review requests against the policy?
No. Nothing in Google's review policies distinguishes between a request typed by a receptionist and one fired by a workflow. Google's guidance on asking for reviews is explicit that you can remind customers and hand them a link or a QR code.
The prohibitions are about manipulation, and there are three of them worth memorizing:
- No incentives. Offering a discount, a free service, a gift card, or a raffle entry in exchange for a review is prohibited content, full stop. This includes asking someone to change or remove a negative review in exchange for something.
- No pressure on the premises. You can't require or push customers into leaving a review while they're standing in your shop. This is the one that kills the front-counter tablet.
- No dictating content. You can't ask customers to mention a specific staff member, a specific service, or a specific keyword. Organic mentions are fine. Coaching them is not.
Notice what isn't on that list: sending at scale. Volume is not the violation. Selection is.
Review gating is the line, and it's probably already in your stack
Gating works like this. After a job, the customer gets a message asking how it went. Four or five stars sends them to your Google review link. One to three stars sends them to a private feedback form that never reaches Google. It feels responsible. It's prohibited, and it has been for years — Google's policy bars using software to screen customers and steer the unhappy ones away from the public profile.
The reason this keeps happening isn't malice, it's defaults. A large share of the review-management platforms sold to local businesses ship with a sentiment fork enabled out of the box, described in the UI as "protect your rating" or "smart routing." Owners buy the tool, connect their CRM, and are gating within an hour without ever seeing the word.
The enforcement risk is real but the mechanical risk is worse. A gated funnel produces a review profile with an impossible shape: ninety-plus percent five stars, almost no fours, a handful of ones from people who found the profile on their own. Real businesses don't look like that. That distribution is a signal, and it's a signal you generate yourself.
What tightened up in 2026
Google spent the first half of 2026 moving enforcement earlier in the pipeline — from removing bad reviews after the fact to screening them before they publish. The changes rolled out in stages through the spring and expanded globally by early summer. Four filters matter for anyone running an automated flow:
- Account history. A brand-new Google account with zero contribution history posting an immediate five star gets held for scrutiny. If you're handing a tablet to customers who've never written a review in their life, expect a chunk of them to evaporate.
- Location proximity. Google checks whether the reviewer's location signals are consistent with having actually been at the business. Reviews written from three states away for a plumber don't survive.
- Compensation language. Automated scanning for phrases that suggest a reward — discount, coupon, free, gift card, comped. A customer thanking you for the freebie in an otherwise glowing review can take the review down with it.
- Same-source clustering. A batch of reviews from one IP address in one afternoon is the textbook front-counter-tablet fingerprint.
Three of those four are self-inflicted by the review process, not by the review. Which is the point: you can lose reviews you legitimately earned because of how you asked for them.
If you're watching reviews appear and then quietly vanish, that's usually what's happening. It's the same class of problem as the visibility work in our Google Business Profile feature guide — the profile is fine, the process feeding it isn't.
The flow we actually build
The compliant build has fewer moving parts than a gated one. In n8n it's typically eight to ten nodes.
One trigger, one path
Pick the single event in the client's system that means the work is genuinely finished — job status flipped to Complete, appointment marked attended, invoice paid. One trigger, not three. When two paths can fire, the same customer gets asked twice and the whole thing starts reading as pressure.
The same message to everyone
No survey in front. No sentiment branch. No filter on ticket size, no "only ask the customers the tech says were happy." Everyone who hits the trigger gets the same request with the same link. This is the entire compliance story in one node, and every attempt to be clever here is a step back toward gating.
Timing that respects the memory
Somewhere between one hour and twenty-four hours after completion, while the customer can still see the result. Not a week later attached to a statement. And never while they're on the premises — that's both a policy problem and a conversion problem, because nobody writes a thoughtful review with a technician watching.
One reminder, then silence
A single follow-up three to four days out if nothing has posted, with anyone who already reviewed suppressed. Two touches total. We've tested more and it doesn't help; it just raises unsubscribe rates and makes the sequence feel like collections.
A rate limit you'll be glad you set
Cap sends per day. The first time a client imports a backlog or has an unusually heavy week, an uncapped workflow will fire two hundred requests in an hour and hand Google a spike to filter. Steady beats fast. This is the same discipline we apply to outbound email sequences — sending patterns are a signal whether or not you meant them to be.
Service recovery after the ask, never instead of it
You still want to know when a job went badly. Send a separate "how did we do?" follow-up to everyone, after the review request has already gone out. Unhappy customers get caught, your ops team gets the signal, and you never made the public invitation conditional on sentiment. Most clients wire that response into the same CRM alerting they already use for lead follow-up, so it lands where somebody will act on it.
What actually moves review volume
Across the local accounts we run this in, three things move the number and everything else is noise.
Trigger accuracy beats copywriting. The single biggest lift we get isn't the message — it's fixing a CRM where "complete" gets set inconsistently, so half the finished jobs never trigger anything. We usually find twenty to forty percent of eligible jobs silently missing before we touch a word of the copy.
One tap to the box. Use the direct write-a-review URL tied to the place ID. Every intermediate landing page, branded microsite, or "choose a platform" screen costs completions. Owners love the microsite. Customers abandon it.
A named human in the message. "Hi Sarah, this is Marcus who was out at your place this morning" outperforms the branded template consistently. It's not a trick, it's just that a review is a favor and people do favors for people.
Compliant flows typically land in the mid-single-digit to low-teens percent response range for SMS on service jobs, which is unglamorous and completely sufficient — a business closing forty jobs a month gets two to five new reviews a month, forever, without anyone remembering to ask.
Responding, without sounding like a bot
Responding is the free half of this that most businesses skip. It's a visible activity signal on the profile, and increasingly it's source text — the same content that gets pulled into local AI answers, which we dug into in the piece on AI-driven local discovery.
Draft with AI, publish with a human. Our rule on client accounts: anything four stars and up can be an approved AI draft; anything below four gets written by a person. A model handles a happy review fine. It cannot read the room on an angry one, and a tone-deaf reply to a one-star review does more damage than the review did.
And don't reply to all thirty at once after six months of silence. Same clustering logic as sending.
What to watch weekly
Two numbers. Requests sent versus reviews received, tracked week over week — that ratio tells you whether the flow is healthy. And your review count on the profile itself, checked against the previous week, because a count that goes down means something you earned got filtered. A rising delete rate is the earliest warning you'll get that a step in the flow is tripping something, and it shows up long before a policy warning does.
If you want the deeper local picture around those numbers, the fundamentals in our Dallas local SEO breakdown are the layer this sits on top of. Reviews accelerate a profile that's already correct. They don't fix one that isn't.
Get your review flow audited
If you're running review requests through a platform you didn't configure yourself, the honest first move is to look at the actual branch logic and find out whether there's a sentiment fork in it. Most owners have never seen that screen.
That's part of what we check in a free automation audit: we trace your review flow end to end, flag anything that gates or spikes, and show you the compliant version mapped to whatever CRM you're already on. If your setup is already clean, we'll say so and move on to something that isn't.
Get a Free Automation Audit — or see how we build these flows with n8n workflow automation.