Why your Google reviews disappeared, and how to keep the next ones

You earned the reviews. One morning they're gone. No email, no warning, no explanation on the profile.

You're not imagining it, and you're rarely the cause. Google removed over 292 million policy-violating reviews in 2025, up from 240 million the year before. Its systems now scan the whole platform continuously and pull reviews in batches, real ones included, then reinstate only the ones that pass a second check. The fastest way to lose a batch is to earn them in a pattern that looks coordinated: all from one connection, all in one hour, all saying the same thing.

Here's what triggers a removal, what you can recover, and how to ask so the next batch stays up.

Reviews vanish in batches now, not one at a time

Google used to remove reviews the way you'd expect: a fake one gets reported, a human checks it, it comes down. That's not how 2026 works.

The moderation now runs continuously across every profile, powered by Google's Gemini models. When it spots a pattern it doesn't trust, it doesn't pause to ask. It pulls the whole cluster, holds it for analysis, and republishes only what clears. In one episode in February 2026, 4.3 million reviews disappeared globally in a single wave. About 3.8 million came back after verification. The other 500,000 stayed gone.

That batch behaviour is the part that catches honest businesses. A fake review getting pulled is fine. The problem is your twelve genuine ones getting swept up because they shared a signal with a spam pattern.

The four patterns that get real reviews removed

None of these require you to do anything dishonest. They're the everyday habits that happen to look like manipulation to a machine.

One connection, many reviews. A tablet on the counter or a QR code on your store Wi-Fi means every review leaves from the same IP address and the same location. To Google, a dozen reviews from one connection reads like one person posting twelve times. The fix isn't to stop asking. It's to let people review from their own phone, on their own data, after they've left.

A sudden spike. You normally get two reviews a week. A wedding, a launch, a quiet push from a supplier, and suddenly fifteen land in a day. Google reads the spike as unnatural and can wipe the batch. One roofing company had twenty homeowners review inside a fortnight after a developer encouraged it; the system flagged the pattern as bought reviews and the business dropped 40%. Every review was real.

Reviews that all sound the same. When several reviews repeat the same phrasing, the same name, the same structure, Google clusters them as a coordinated set. This is about repetition across reviews, not length. A genuinely short review is fine. Ten reviews that all say a near-identical line are not.

Scripted content and named staff. As of April 2026, Google's policy explicitly bans directing customers or staff to include specific content in a review, including naming the employee who served them. A customer naming you because they want to is still fine. Being told to do it is the violation, and clusters of name-dropping reviews now trigger a closer look at how you collected them. We cover this in full in what Google permits on review automation.

What you can get back, and what you can't

Some of a wiped batch returns on its own once Google's second pass clears it. That can take days. There's no button to speed it up.

For reviews that stay down and shouldn't have, Google does run a reporting flow from your Business Profile, Google Maps, and Google Search, all routing to the same place. It's worth filing for a review you can prove is legitimate. Be realistic about the odds: when removals are algorithmic and made in bulk, appeals are slow and often go unanswered. The reliable lever isn't recovery. It's not getting swept up next time.

How to ask so your reviews survive

The pattern Google trusts is the opposite of a one-day blast from one device. It's steady, spread out, and written by the customer.

Send the ask after they leave. A text or email a few hours later means the review comes from the customer's own phone and location, not your counter. That one change removes the single-connection signal that sinks tablet and Wi-Fi-QR setups.

Keep the flow steady. A handful of reviews a week, every week, beats fifteen in an afternoon. Consistent beats burst, and it's also what a real customer flow actually looks like.

Ask an open question. Script nothing. "Mind sharing how it went?" gets you a genuine, varied review. It also keeps you clear of the content and staff-naming rules. The review request templates are built to stay inside both.

Don't gate. Routing happy customers to Google and unhappy ones to a private form is the fastest way to get a whole profile's reviews unpublished. It's banned outright.

How RealGoodWords keeps you out of the blast radius

This is the version of asking that Google's own playbook describes, and it's how the product works by default.

Requests go out after the visit, by SMS or email, each one to the customer's own Google review page from their own device. No counter tablet, no shared connection, no single IP behind every review. The flow is paced to stay steady rather than spike. Every ask is one open question, so the customer writes their own words, with no script and no staff name put in their mouth. And everyone gets the same link, so there's no gating to put the whole profile at risk.

You can't control Google's moderation. You can control whether your reviews look like a real customer base or a coordinated batch. That's the whole difference.

If you want to see where your profile stands today, the free Local Standings tool at realgoodwords.app compares you to three nearby competitors. No signup, no card.