How to Tell a Reporting Bug From a Real Traffic Drop
A chart falling off a cliff and a real loss of business look identical for about the first ten minutes.
Tags get wiped by a plugin update. A developer changes a URL. A consent banner starts blocking the thing that counts visits. Google's own reports break. Every one of those draws the same shape as a genuine drop, and the cost of guessing runs both ways: you spend a month fixing a problem that was never there, or you sit on a real one because someone told you it was probably a glitch.
There is a test that separates them, it takes about ten minutes, and it costs nothing. Below is the test, and then August 2026 as the worked example — a month where a confirmed Google bug and a real ranking update landed within a fortnight of each other and a lot of people credited the wrong one.
Rather have someone look at the actual numbers with you? Book a free strategy call. Otherwise the whole method is below, and it costs nothing to run.
Three Things Happened in August, and Only One Was a Bug
They landed within about two weeks of each other, which is most of why the month was confusing.
A Confirmed Logging Error
Google confirmed that a logging error depressed impressions in the Generative AI in Search performance report — the Search Console report covering AI Overviews and AI Mode — for data starting August 13, 2026. In Google's words, the issue affected "data logging only," and John Mueller said plainly that it was "not representative of visibility changes in Search."
Worth being precise about the boundary, because a lot of coverage was not. What Google confirmed was a problem in that one report. Site owners also reported drops in the standard Search performance report over the same window, and those were not separately confirmed. The data appeared to normalize the following week.
So: one confirmed bug in one report, plus a cloud of unconfirmed reports around it. That distinction matters when you are deciding whether your own numbers were affected.
A Real Ranking Update
The August 2026 spam update ran from August 18 to August 21, globally and in all languages, according to Google's own Search Status Dashboard. It was the third spam update of 2026 and Google described the mechanics as nothing new.
Routine does not mean irrelevant. A spam update is a real ranking event, and a drop that starts inside those four days is a different problem from a logging error — it goes down the path in if it turns out to be real, not the reporting path. A drop that starts before August 18 was not caused by it, whatever the timing looks like at a glance.
A New Report Going Live
The Generative AI performance report finished rolling out. Google's own documentation is unusually specific about the date: "As of August 31, 2026, we've rolled out these insights to all websites worldwide." It had been arriving in phases before that, which is why some people had it in June and others did not have it in August.
A new report in the middle of a confusing month adds its own noise: a number that appears for the first time can look like a change when it is just a measurement that did not exist before.
The One Test That Settles It
Does the drop show up in more than one place?
That is the whole test, and it works because your numbers come from separate systems that fail independently. Search Console counts what Google served. Your analytics counts what loaded in a browser. Your form inbox counts what a human actually sent you. Your phone counts what rang.
A real drop in visibility moves several of those at once, because fewer people arriving means fewer sessions, fewer forms, and fewer calls. A reporting bug moves exactly one, because only one system is broken.
Note which two are hardest to break. Search Console and your analytics both depend on code and configuration, so either can go wrong on its own — which is why "check your analytics" is not a complete answer when analytics might be the thing that failed. Your form inbox and your phone log are different. A human being typed that message or dialed that number. A broken tag cannot fake one and cannot hide one.
That is why I keep call tracking and form submissions next to the traffic numbers in a monthly report rather than on their own page. They are not only outcome metrics. They are the control group.
So: drop in Search Console, leads flat, phone normal — you are almost certainly looking at a reporting problem. Drop in Search Console and the phone is quieter — something real happened, and the report is just where you noticed it first.
Four Questions Before You Panic
In order. The first two take about five minutes together and settle most cases.
- Is Google reporting a known issue? Check the Google Search Status Dashboard. It logs ranking, indexing, and serving incidents with start and end timestamps, and it is where ranking updates get recorded too. If your drop lines up with a logged incident, you have your answer for free.
- Did leads move? Forms, calls, bookings, orders — whatever counts as a lead for you. If those held steady while a chart fell, the chart is the problem.
- Did anything change on the site? A plugin update, a redesign, a URL change, a new consent banner, a developer touching the header. Tracking code lives in places that get overwritten by routine work, and this is the single most common cause of a drop that turns out not to be a drop. What a working tracking setup includes is the list to check it against.
- Does the shape look human or mechanical? Real declines are ragged and partial — down a third, spread unevenly across pages, worse on some queries than others. Mechanical failures are clean: a vertical drop on one exact date, or one segment reading zero while every other segment carries on untouched. Nothing about real human behavior is that tidy.
If It Turns Out to Be Real
Then stop looking at the total, because the total cannot tell you anything actionable. Three moves, in order.
Find out what specifically lost. In Search Console, compare the two periods by page and by query rather than by site. Almost always the loss is concentrated — a handful of pages, or one cluster of queries. A site-wide 30% decline that turns out to be four pages losing 80% is a completely different problem from an even decline across everything, and only the first one has an obvious fix.
Line the dates up against what changed. Your own release history first, then Google's. If your drop starts the day a ranking update finished rolling out, that is a strong signal. If it starts three days before, the update is not your cause no matter how much you would prefer it to be.
Decide whether it is worth chasing. Not every real drop is worth money to fix. If the pages that lost ground were not producing leads anyway, the honest answer is to note it and spend the hour somewhere that pays. Ranking for something you were never going to convert is a vanity problem wearing a serious costume. Every hour costs the same, so the only question is which one buys more.
The New AI Overviews Report, and What It Is Not Yet
The Generative AI performance report is the first client-facing data anybody has had on whether a site shows up inside Google's AI answers. That is genuinely new and it is worth pulling. It is also worth being honest about its limits, because the gap between "we have a number" and "we have a number that means something" is where reporting goes wrong.
What it gives you: organic impressions from Google's generative AI features over time, your best and worst performing pages inside them, and a breakdown by device, country, and date.
What it does not give you, per Google's own documentation:
- It covers AI Overviews and AI Mode only. Experimental Search Labs surfaces are excluded.
- It caps at 1,000 rows, the same limit as the standard performance report, so a large site is seeing a sample.
- Recent days are preliminary and can change within hours.
- Chart totals and table totals can disagree, because they aggregate differently.
- Impressions are the headline, which is a visibility number, not a revenue number.
So use it as a direction, not a decimal. It answers "are we appearing in AI answers at all, and is that going up or down" — a real question nobody could answer six months ago. It does not answer "what did AI search earn us," and any report that implies it does is dressing up an impression count as a result. That is the same move as leading a report with reach, and it ages just as badly.
It is also the right tool for the opt-out question. If a client asks whether they should block their site from AI Overviews, the answer for most small businesses is no — but now you can look at what they are actually getting before saying so, instead of arguing from a general principle.
Not sure what you are looking at?
Send a screenshot of the chart that worried you to saguarosocialmediaaz@gmail.com, with a note about whether your leads moved. I will tell you whether it reads like a reporting problem or a real one, and what I would check next. One paragraph back, no pitch, and I will not follow up unless you ask me to.
What This Means for Your Monthly Report
Three habits come out of a month like August, and none of them are extra work.
Say what you do not know. "Impressions fell in this report and nowhere else, Google confirmed a logging error covering those dates, and leads were flat" is a complete answer. It is more useful than a confident story in either direction, and a client can tell the difference.
Do not invent a fix for a problem that was not real. The temptation in a bad-looking month is to show activity — a batch of changes that proves someone was paying attention. If the cause was a bug, the changes are noise you will be untangling next month when you try to work out what actually moved.
Keep the lead numbers next to the traffic numbers. A report that answers what we spent, what we got, and what we are changing has the control group built into it. That was true before August and it is what made August readable.
A clean month is also worth reporting plainly. "The August update rolled out and our rankings did not move" is a real finding, and it takes ten minutes to confirm. The same logic applies on the paid side, where a change in how bidding works can move your leads without moving your invoice.
Common Questions About Data Drops
Five that come up whenever a chart moves.
My Search Console Traffic Dropped. How Do I Know If It Is Real?
Check whether the drop shows up anywhere else. Search Console, your analytics, your form inbox, and your phone measure different things in different ways. A real drop moves more than one of them. A reporting bug moves exactly one.
Where Can I Check Whether Google Is Having a Known Issue?
The Google Search Status Dashboard. It logs ranking, indexing, and serving incidents with start and end timestamps, and it is also where Google records the start and finish of ranking updates. It is free and it is the first thing to check.
Did Google Confirm the August 2026 Search Console Drop Was a Bug?
Partly, and the distinction matters. Google confirmed a logging error affecting the Generative AI in Search report for data starting August 13, and said it touched data logging only. The wider drops site owners reported in the standard Search report were not separately confirmed. Data appeared to normalize the following week.
Should My Site Opt Out of AI Overviews?
For most small businesses, no. Opting out removes you from a surface people already use and gains you nothing directly. The better move is to open the Generative AI performance report and decide from your own numbers.
How Much Can I Trust the New AI Performance Report?
Enough to watch a trend, not enough to bill against. It covers two AI surfaces, reports impressions rather than clean clicks, caps at 1,000 rows, and its most recent days are preliminary. Direction, not decimal.