Twenty real reviews in a weekend looks exactly like a purchase.

Every article on this topic wants to hand you a number: two a week, one a day, ten a month. Google has never published one, and the people quoting numbers are guessing. What actually matters is velocity relative to your own history. A profile that has collected two reviews a week for a year can absorb a busy week without a ripple. A profile that has collected two reviews a year cannot suddenly do two a day.

That's why the same fifteen reviews are safe for one business and radioactive for another. The filter isn't counting reviews. It's comparing this week's pattern to your baseline, and a baseline of silence makes almost any burst look wrong.

So the honest answer to 'how many at once is safe' is: barely more than your normal, sustained for long enough that the new pace becomes your normal. Growth reads as growth. A step-change reads as a purchase.

A spike gets Google's attention. What happens next depends on everything else the reviews have in common, because a purchased batch shares more than timing. Reviews that arrive together AND look alike are the ones that go down together.

The old deal was simple: suspicious reviews quietly disappeared and life went on. That deal expired. Google now maintains a public menu of profile-level restrictions for fake-engagement violations, and the pattern that triggers review deletion is the same pattern that starts a restriction file.

The ladder runs from losing the flagged reviews, to a freeze on receiving new ones, to your existing reviews being temporarily unpublished, to a public notice on the profile that fake reviews were removed. Repeat the pattern and industry reporting says the restriction windows stretch, in some cases roughly doubling. And here's the part that stings: the filter would rather delete ten real reviews than let two fake ones stand. When a burst gets swept, your legitimate praise goes with it, and removed reviews rarely come back.

One blast of enthusiasm can cost a year of collected trust. That's an expensive Saturday.

There's a softer version of the penalty that owners find more confusing than removals: every new review simply fails to publish. Customers swear they posted. The profile shows nothing. Local SEO circles call it review jail, and the practitioners who've tracked it report holds that can run for months, with Google support declining to share timelines.

If your reviews stopped landing right after an aggressive push, this is the likely explanation, and the fix is patience plus a clean pattern: keep serving customers, keep asking at a calm pace, and let the profile demonstrate normal behavior until the hold lifts. What doesn't help is asking harder, which extends the exact pattern that caused the hold.

You drip it. The backlog isn't going anywhere, and neither is the filter, so the move is to convert your pile of past customers into a steady feed instead of a flood. Work backwards from your most recent jobs, a handful of asks per week, freshest first because they remember the most.

And accept an honest truth about old jobs: the customer from last February barely remembers the details, and detail is what makes a review both convincing and filter-proof. Your best asks are always the most recent ones. The backlog is a bonus round, not the strategy.

The strategy is the habit: every completed job gets an ask within a day or two, forever. Do that and the velocity question answers itself, because your review pace simply tracks your job pace, which is the one pattern Google never questions.

Every trap in this guide comes from the same root: asking in batches, on your schedule, in ways that make the reviews arrive together and look alike. The fix is a system where the ask rides along with the work. Job wraps, customer gets a text that evening, they answer from their couch on their own connection, and the review that comes out names what actually happened because it was built from their own answers.

That's how small Talk sends. Requests go out per job, not per campaign, so your review pace mirrors your actual work pace. Customers respond from their own devices on their own time, so there's no shared-network fingerprint. And the guided draft pulls specifics out of people who would otherwise type 'Great service!', which is exactly the generic shape the filter sweeps. The result is a review stream that grows the way Google expects real reputations to grow: steadily, specifically, one honest job at a time.