The Google reviews bug is a preview of the AI problem

Google Business Profile threw another wobbly yesterday. Business owners logging in to reply to reviews were told "You have no reviews yet" — while the same listings publicly displayed hundreds or thousands of reviews. One case flagged by Google Product Expert Amy Toman showed a listing with 916 reviews on the live page and zero visible in the dashboard. Panic in the forums. Threads multiplying by the hour.

By evening Google said the underlying issue — reviews being hidden after fake-review disputes — was resolved, and everything should reappear within a day or two.

Fine. Bug fixed. Move on. Except this isn't really about the bug.

This is about the fact that a business's most important reputation asset can disappear for a day, come back, and the only signal you had that it was fixed was a forum post from a Product Expert on LinkedIn. Not an email. Not a dashboard alert. Not a status page anyone thought to check. That's the actual story. And it's the same story that's about to define AI search visibility.

The dashboard is not the source of truth

Here's what happened, mechanically. Google's review-dispute system did something to legitimate reviews. Owners saw one number in their dashboard and a different number on the public listing. The public listing was right. The dashboard was wrong.

Think about that for a second. The interface Google gives business owners to *manage* their reviews was showing a false version of reality. The customer-facing version was accurate. Owners who trusted their dashboard would have concluded thousands of reviews had been wiped. Owners who checked the live listing would have known the reviews still existed but couldn't be accessed for reply.

This isn't a new pattern. Search Console has diverged from actual search behaviour for years — the data lags, samples, and rounds in ways that make month-over-month analysis genuinely unreliable for smaller sites. GBP has had a rolling series of these problems: reviews vanishing, listings suspended, categories changing on their own. Every time, the response is the same: forum panic, then eventual acknowledgement, then a fix nobody's told about directly.

The dashboard Google gives you is a snapshot of what Google thinks about your business, filtered through whichever system is currently broken.

Why this matters more now, not less

For a decade this was annoying but manageable. If GBP glitched, you checked your live listing, you complained in the forum, you waited. The public-facing reality was the ground truth and you could always see it.

That model breaks in AI search.

When ChatGPT or Google's AI Mode cites your business, there's no equivalent of "checking the live listing." There's no dashboard that tells you which queries triggered a citation, what the model said about you, or whether the citation was accurate. The public-facing reality *is the model output* — and the model output is unstable, personalised, and different every time you ask.

We already have a version of the reviews bug playing out in AI citations right now. Businesses get mentioned, then don't. A model confidently attributes a quote to the wrong company. A local service gets cited for a region it doesn't serve. And the "dashboard" — whatever GEO tool you're paying £400 a month for — is showing you sampled data that may or may not match what a real user just saw.

The GBP bug is a small, contained, fixable version of a problem that's about to be enormous, distributed, and unfixable in the same way.

What this actually breaks

There are three practical consequences worth naming, because most agencies aren't naming them.

First, the assumption that you can monitor your visibility from a single interface is dead. GBP taught us this. Search Console taught us this. Every AI citation tool is about to teach us this again. The correct posture is triangulation — check multiple sources, assume any single one is wrong, and build workflows that surface disagreement rather than smoothing it over.

Second, the response time between "something is broken" and "you find out" is now the actual metric that matters. Not rankings. Not impressions. How quickly can you tell that your reviews vanished, your citation dropped, your category flipped, your AI mention now says something wrong about your product? Most businesses don't have a defensible answer to this. Their answer is "when a customer emails us."

Third — and this is the one nobody wants to hear — the fix is not more dashboards. More GEO tools, more monitoring, more Slack alerts fed by more APIs won't solve this. What solves it is a habit of manually checking your own visibility across surfaces, weekly or bi-weekly, from queries you didn't feed into a tool. Search your business by name in ChatGPT. Look at your live GBP listing while logged out. Ask Perplexity about your category. The friction is the point — it's the only way you catch the things the dashboards are lying about.

The uncomfortable through-line

Google Business Profile is arguably the most important single asset for local businesses in the UK. It drives more direct calls, directions, and enquiries than most local businesses' entire websites. And Google has spent the last two years shipping bugs to it faster than they've shipped features. Reviews vanishing. Reinstatements taking weeks. Legitimate businesses paused because the spam algorithm was too aggressive. Dashboards showing zero when the reality is 916.

If your entire local visibility strategy is downstream of a Google product that behaves like this, you have a concentration risk that's not going away. Diversifying isn't a slogan — it means having enough independent surfaces (your own site with proper local schema, Apple Business Connect, Bing Places, actual review platforms customers use, direct customer relationships) that a bad GBP week isn't a bad revenue week.

That was true when it was just GBP. It's about to be true across every AI answer surface too.

What to actually do

Don't panic-audit your reviews today. If they're gone, they're probably back by tomorrow morning. The immediate bug isn't the point.

The point is to build the habits before they're urgent. Once a week, check your live GBP listing while logged out. Once a fortnight, search your brand name in ChatGPT and note what it says. Once a month, review whether your Search Console data is telling the same story as your analytics, your rank tracker, and your actual enquiry volume. When they diverge — and they will — that divergence is the signal, not the noise.

The businesses that will do well in AI search aren't the ones with the fanciest citation dashboards. They're the ones who noticed things were wrong before the dashboard did.

The GBP bug isn't a bug story. It's a rehearsal.