.png)
Ask a few hours after the work is finished, by text, with a direct link. If nothing comes back, ask once more around day ten, then stop.
That is the short answer, and most articles about this stop there. Two things they leave out matter more than the wording. The first is that where and how the review gets submitted now affects whether it survives at all — and the right answer to that is the opposite depending on what kind of business you run. The second is that no business fixes this with better templates. It gets fixed by making the ask automatic, so it stops depending on anyone remembering.
The gap between meaning to ask and asking
Ask an owner what share of jobs get a review request and the answer is usually somewhere near ninety percent. Count the actual sends and it tends to land closer to thirty.
The gap is rarely a discipline problem. Asking sits at the end of a job, which is the busiest and least structured moment in the day — the invoice, the next call, the drive. It's the step with no deadline attached, so it's the step that quietly doesn't happen.
There's a second, quieter reason, and it shows up constantly in conversations with owners: asking feels like fishing for a compliment. The work went fine, the customer was happy, and the ask still doesn't get made, because it feels like a favour too far. That hesitation has nothing to do with the quality of the work and everything to do with the awkwardness of the moment.
Both problems have the same fix, and it isn't a better script. It's taking the decision out of the moment entirely.
Where the review gets submitted, and why it depends on your business
The standard advice is to capture the review while the customer is still in front of you — at the counter, in the driveway, phone in hand. A newer round of advice says never to do it. Both are too blunt, because neither describes what actually gets filtered.
What gets filtered is clustering: a run of reviews from the same device, the same network, or arriving inside a tight window. That pattern is indistinguishable from a coordinated push, and filters act on patterns rather than intentions. Whether your in-person ask produces that pattern depends entirely on where the work happens.
If customers come to you — a shop, a clinic, a salon, a dealership, an office — then asking at the counter puts review after review on one address and, very often, one WiFi network, several inside the same hour. That is the cluster, exactly as described. Businesses running counter-side capture routinely see batches of entirely genuine reviews disappear days after they were posted.
The adjustment costs nothing. Keep the verbal ask, which still converts better than anything else available. Move the submission off the premises:
"If you're happy with how today went, a Google review makes a real difference to us. We'll text you the link shortly so you don't have to go looking for it."
Then send it an hour or two later, from their own home on their own connection.
If you go to the customer — home services, mobile repair, in-home care, anything where the job happens at their address — the picture reverses. Every review comes from a different address on a different connection, spread across the day. A customer standing in her own driveway, reviewing on her own phone over her own cell service, is the most ordinary-looking review Google will process that week. There is nothing to cluster. Ask in the driveway. It converts better than anything you can send later.
Three rules hold either way:
Never hand over your own phone. Multiple reviews from one device is the clearest signal in the set, and it's the one that takes a whole batch down with it.
Never put the customer on your hotspot or guest WiFi to do it. That collapses a mobile business back into a fixed one and manufactures the exact pattern you'd otherwise never produce.
Never send in bulk. A backlog cleared in one afternoon is a burst regardless of where anyone was standing. More on that below.
Google's published policy doesn't address location at all. It goes as far as fake engagement — "content that is not based on a real experience" — and the reviews in question are real experiences. But the documentation describes intent, and the filtering responds to signals. Those are not the same thing, and the businesses losing reviews are the ones that assumed they were.
What AI actually reads in a review profile
Something has shifted in the last two years that most review advice hasn't accounted for. Customers are no longer the only audience for review text. Assistants and AI search results read it too, and they read it closely.
When someone asks an assistant to recommend a plumber in their city, it works from what previous customers described. Forty reviews reading "great service, thanks" supply almost nothing to reason from. Fifteen reviews describing what was wrong, what was done, and how the job went supply enough context for a confident, specific recommendation — and a recommendation with reasons behind it is the one that gets made.
The distinction is depth, not word count. A padded review says no more than a short one. What earns the recommendation is specificity — the job named, the problem described, the outcome stated. A profile full of specific reviews is a profile that can be summarised accurately, and being summarised accurately is how a business ends up named in an answer instead of skipped.
Three practical consequences:
Detail is now doing work that the star rating used to do alone. A four-star review describing the job in three sentences carries more weight, with both readers and models, than a bare five star.
A perfect 5.0 isn't the target. A 4.9 supported by substantive reviews reads as credible. A wall of five-star ratings with nothing written beneath them reads as manufactured, to people and to software.
Photos add context nothing else does. A customer photo of the finished cabinet, the new roof, the colour — it documents the work in a way text can't, and almost nobody thinks to ask for one.
Google is more restrained about what any of this does for ranking. Its own guidance says "more reviews and positive ratings can help your business's local ranking," alongside relevance and distance, and states plainly that "there's no way to request or pay for a better local ranking."
When to ask for a Google review
Gratitude has a short shelf life. The hour after the heat comes back on is a different state of mind from three days later, when the system simply works and always did.
- Same-day repair or service call — 2 hours after completion (text)
- Install or big-ticket job — Day 2, after a night with it (text, then email day 3)
- Appointment-based work — 3–4 hours later, or next morning (text)
- Anything delivered later — When it's received, not when it's finished (email)
- Emergency or after-hours — The following day, in daylight (text)
- Long-cycle projects — Around two weeks after completion (email)
The principle beneath the list: ask once the thing that was sold has proved itself, not at the moment the work stopped. For a repair that's a couple of hours. For an installation it's the next day. For photography it's when the gallery arrives, not when the shoot ends. For a wedding it's a fortnight later, once everyone is home.
Never on the morning of an event. Never mid-honeymoon. Never at 3am after an emergency call.
How to ask for a review by text
One message, a couple of hours after leaving:
Hi Marcus — Dale here from Northside Heating. Just checking the AC is running right for you now.
If it is, would you mind leaving us a Google review? Takes about 30 seconds and it genuinely helps a small shop like ours.
[review link]
If anything's off instead, reply here and I'll sort it.
Four things in that are deliberate.
It comes from a named person rather than a business. It confirms the work is actually finished before asking a favour. The link sits in the message — every additional tap costs responses, so the short link from the Google Business Profile belongs directly in the text, tested on a phone while signed out before it goes anywhere.
The final line does the most work. It gives a dissatisfied customer a faster route than a public review, and most people take it, because a fixed problem beats an audience. That is not review gating, and the difference matters — everyone still receives the same link.
How to automate Google review requests
Everything above is a timing problem, and timing problems are what automation is for. A person has to remember, at the end of a job, on a busy afternoon. Software doesn't.
What a review request automation actually does. The architecture is the same regardless of which tool runs it:
- A trigger — a job marked complete, an invoice paid, an appointment closed out in the CRM or field service app.
- A delay — two hours, or until 9am the next morning. This is the piece that manual asking never gets right.
- Two gates — messaging consent on the record, and a quiet-hours window so nothing sends before 8am, after 8pm, or on a Sunday.
- The send — text first, email as the second touch on a separate day.
- A stop rule — the contact exits permanently on a review, a reply, or day fourteen. This has to live inside the automation, because it's the rule people forget, and forgetting it is how a loyal customer gets asked eight times a year.
- A record — every ask logged, so ask rate becomes a number rather than a guess.
Where these builds usually go wrong. The trigger is the hard part, not the messages. Most field service software marks a job complete at a moment that doesn't match when the customer actually experienced completion — the tech closes the ticket in the van, or the office closes it two days later during invoicing. An automation firing off the wrong event sends perfectly written messages at meaningless times, and that failure is invisible until someone checks.
The second common failure is a burst. Connecting an automation to a CRM with two years of closed jobs in it, without a date filter, will send hundreds of requests in an afternoon. Which brings us to the next section.
Whether it's worth automating. Below roughly ten jobs a week, a person with a saved message and a calendar reminder can run this by hand. Past that, manual asking degrades in exactly the way the thirty-percent figure describes — not all at once, just quietly, on the busy weeks, which are the weeks with the most jobs in them.
Anyone can generate the message templates now. Asking an AI for a review request script takes ten seconds and the output is fine. What doesn't come out of a chat window is the build: the trigger wired to the right event, the gates, the stop rule, the logging, and something watching it to confirm it's still firing next month.
Want this running rather than documented? A Pebble Digital Blueprint maps the whole follow-up system for a specific business — triggers, timing, gates, and where the current process leaks — and hands over the plan. See what a Blueprint covers →
How many follow-ups
One.
A single reminder converts. A second costs more customers than it earns reviews. Around day ten, one message closes the loop:
Hi Marcus — last time you'll hear from me about this, promise. If the AC is still running well, here's the link: [link]. If it isn't, reply and I'll get someone out.
That opening line does two jobs. It converts, because it removes any suspicion of an ongoing campaign. And it commits the business — a third message after that sentence is a broken promise over a review.
Then the hard stop. Day fourteen, the contact is finished and doesn't re-enter on the next job. This matters most for businesses people return to. A salon client asked at every visit gets asked eight times a year by a system that resets, and that is how a regular quietly starts going elsewhere.
Why not to ask everyone at once
A business sitting on six months of happy customers shouldn't clear the backlog on a Tuesday.
This is the one clustering risk that applies to every business, mobile or fixed. Reviews arriving in a burst are the clearest available signal that something is being organised rather than earned, and filters are built around precisely that shape. A steady trickle outperforms a spike, and it isn't close. Working through a backlog over several weeks — oldest first, a handful at a time — feels slow, and it's still faster than having the batch removed and starting again.
This is also the single most common way a new automation causes damage. A date filter on the trigger is not optional.
Reply to every review, and reply quickly
Within the hour where possible. It feels excessive; it isn't, and it matters most on the negative ones.
Google's guidance states that "positive reviews and helpful replies can help your business stand out." In practice responsiveness appears to carry more weight than that phrasing suggests, and the profiles that climb tend to be the ones where every review gets answered the same day without exception.
Replies are also read by people, which is the part that gets forgotten. The reviewer is one person. The reply is written for the hundred who read it afterwards while deciding whether to call. Better audience, one sentence.
Two sentences, the customer's first name, the actual job named. Twelve identical thank-yous read as automated, because they are.
For a negative review where the customer is right: apologise, take responsibility, give a direct number, sign it. Don't explain the invoice.
For one where the customer is wrong: a single line acknowledging the disagreement, then move it off the platform. "Our recollection differs, so I'd rather talk it through than go back and forth here — I'm at [number]." Sentences spent defending the business read as defensive to everyone watching, including when the business is entirely right.
What actually gets reviews removed
Google's prohibited content policy is short and worth reading directly. Three things are genuinely out:
Paying for reviews in any form. The policy names "payment, discounts, free goods and/or services." A discount offered for a review sits in the same category as buying them outright, and twenty reviews mentioning the discount is not subtle.
Filtering who gets asked. The wording is that a business may not "discourage or prohibit negative reviews, or selectively solicit positive reviews from customers." Surveying customers and sending the Google link only to the happy ones is exactly this. Asking everyone while offering a direct reply as an alternative is not.
Writing them. Not from the business account, not from a relative's, not as a draft handed to a customer to post.
Asking is explicitly permitted. The same policy confirms a business may "solicit or encourage the posting of content that does represent a genuine experience."
One useful thing to make easier without crossing that line: give people a starting point rather than a script. "Even a line about what we fixed and how it went is plenty" addresses the blank-page problem, which is the real reason people who meant to leave a review never did. Sending every customer identical wording is the version that causes trouble — twenty reviews built from one sentence stop looking like twenty customers.
The number worth tracking
Not the star average. It moves too slowly to be diagnostic and largely reflects how long a business has been open.
Ask rate — the share of completed jobs that received a request — is the number that moves. It's controllable this week, and review count follows it by roughly a month. It's also the number that produces the uncomfortable conversation, since most businesses convinced they're at ninety percent are closer to thirty.
Response rate and reply rate come next. When response rate is low, the cause is almost always timing rather than wording.
Ask rate is also the strongest argument for automating the process, because an automated ask rate is a hundred percent by definition. There's no busy week, no forgotten job, no end-of-day decision to make.
Where to start
Three steps, in order:
- Pull the Google review short link from the Business Profile and test it on a phone, signed out.
- Add the verbal ask to the end of every job — and send the link a couple of hours later if customers come to you, or let them submit on the spot if you went to them.
- Count ask rate for two weeks. Nothing else.
Two weeks of counting usually settles the question of whether this needs a system. Most businesses find the number is lower than expected, and that the gap sits on the busiest weeks — which are the weeks with the most jobs to lose.
Get the asking off your to-do list
Pebble Digital builds and monitors the follow-up systems behind this — review requests, quote follow-up, missed-call recovery — for home service businesses and other appointment-based work. A Blueprint maps the whole thing for one specific business: the triggers, the timing, the gates, and where the current process is leaking. The plan is yours whether or not we build it.
Written by Cade Bagley of Pebble Digital, who works in local SEO and automation for home service businesses. Last updated September 2026.