If you opened Search Console this week and your AI impressions fell off a cliff between November 9 and November 11, 2026, take a breath before you touch anything. Google confirmed it was a logging error, not a ranking loss. The impressions were never counted properly, so the dip is a hole in the data — not a hole in your traffic. Your pages did not suddenly lose visibility in AI Overviews or AI Mode. Google’s counting broke for three days, and the numbers for those three days cannot be trusted.
That distinction matters. A real visibility drop calls for diagnosis and sometimes hard decisions about content. A logging error calls for exactly one thing: a note in your records, and patience. Mix the two up and you can spend a week “fixing” pages that were never broken — and end up making them worse. This article covers what Google actually said, why the timing made this one so easy to misread, and how to tell a reporting gap from a real drop in about fifteen minutes.
What happened on November 9–11, 2026
Google’s Search Console data anomalies page carries a short note about the incident. It says, verbatim:
“A logging error caused a decrease in impressions on the Generative AI performance report in Search for data from November 9 – November 11, 2026. This issue affects data logging only.”
Read that last sentence twice: “This issue affects data logging only.” The problem sat in the counting, not in the search results. Your pages were still appearing in AI Overviews and AI Mode the way they had been. The system that records those appearances and turns them into the impressions number in your report was the part that failed. The visibility was real; the record of it was not.
The note did not appear the moment the problem started. It showed up on the anomalies page between November 12 at 15:00 UTC and November 13 at 03:00 UTC — after the three affected days had already passed. That delay is normal for this kind of confirmation, and it is also why so many site owners spent November 10 and 11 staring at a falling graph with no official explanation. Search Engine Land’s report was published on November 12, 2026, and Search Engine Roundtable covered it the same day. Both pointed back to the same anomalies-page note, and both framed it the way Google did: a data bug, not a ranking event.
Three details in Google’s wording are worth pinning down. First, the window is specific: November 9 through November 11, 2026. Not the whole month. Second, the report is specific: the Generative AI performance report in Search — not the main Performance report, and not Discover or News. If your main Performance report looks normal for the same dates while the AI report dips, that is consistent with what Google described. Third, Google has not said the missing impressions will be backfilled. Treat November 9–11 in that report as a data gap, not a true reading. Do not estimate what the “real” number would have been and present it as measured. Mark the gap, explain it in one sentence, and move on. An annotation now appears in the report itself, but your own notes still matter, because the in-report annotation does not travel with the data when you export it into a spreadsheet or a client deck.
Why the timing was a trap
Here is where this incident stops being a routine data bug and becomes a genuinely nasty trap: Google’s November 2026 core update started rolling out on November 10, 2026 — right in the middle of the logging-error window.
The core update began on November 10, according to the Google Search Status Dashboard, at around 10:07 a.m. PT. It is expected to run for up to two weeks, finishing around November 24. Google’s standing guidance for core updates is to wait at least a full week after the rollout completes before judging what it did to your traffic. That guidance exists for a reason: rankings move around during a rollout, the dust takes time to settle, and an early reading is often wrong.
Now put the two events together. On November 9, your AI impressions start falling because of a logging error. On November 10, a core update starts rolling out. You open Search Console on November 11 or 12 and you see a sharp dip in a brand-new AI report, on the exact days a core update began. If you did not know about the logging error — and for the first day or two, nobody outside Google did — the obvious conclusion writes itself: the core update hurt your AI visibility. That conclusion would have been wrong, but it would have felt completely reasonable.
So the first job after an incident like this is not fixing anything. It is separating the two events in your head and in your records. The logging error is a three-day counting fault, November 9–11, confined to one report. The core update is a two-week ranking rollout, November 10 to around November 24, that touches the whole of search. They overlap in time. They do not overlap in cause, and they do not overlap in what you should do about them. For the core update, the right move is the boring one: wait. Our own November core update guide covers the timeline and the wait-and-see approach in detail. Do not let a three-day logging error convince you that the core update has already delivered its verdict on your site. It has not.
What the Generative AI report actually measures
Part of the confusion comes from the report itself being new. The Generative AI performance report in Search covers AI Overviews and AI Mode. It was announced on June 3, 2026, and it reached all websites worldwide on August 31, 2026. By November, most site owners had only had a couple of months with it.
It measures one thing: impressions. It shows impressions only. It does not show clicks. An impression in this report means your content appeared in an AI Overview or in AI Mode. It does not mean anyone clicked through to your site. You cannot use this report to work out a click-through rate, and you cannot use it to say how much traffic AI Overviews sent you. It is a visibility measure, and only a visibility measure.
The same impressions are also counted in the main Performance report. That becomes a diagnostic tool: because the same underlying appearances feed both reports, a logging error that hits one report’s counting but not the other’s leaves a fingerprint — the AI report dips while the main report holds steady. A real loss of visibility would show up differently, because the appearances themselves would be gone, and both reports would reflect that. And because there are no clicks in this report, there is no traffic number attached to the dip. Nobody lost sessions because of this error. A dip in impressions in a report that has no clicks is a dip in a record of appearances. It is not a dip in visitors.
Real drop vs. logging error: what each one looks like
The most useful thing you can do with this incident is turn it into a reusable test. The next time a Search Console graph falls, you will want to know within minutes whether you are looking at a real problem or another counting fault. The table below shows the pattern each one leaves across four signals. It is illustrative — a guide to the shapes to look for, not measured data from any one site.
| Signal | Real visibility drop | Logging error (Nov 9–11 pattern) |
|---|---|---|
| AI report impressions | Falls and stays down, or falls gradually over days and weeks | Falls sharply for a defined window (here, three days), then returns to the previous level |
| Main Performance report clicks | Falls in step with the impressions drop, on the same dates | Holds steady — no matching collapse on the same dates |
| GA4 organic sessions | Falls in line with the clicks drop, same dates, same scale | Holds steady — real visitors kept arriving at the normal rate |
| Server-log Googlebot hits | Often falls or changes pattern if crawling or indexing is involved | Normal crawling pattern continues |
A real drop is consistent: the same story shows up in every signal, because fewer appearances in search means fewer impressions, fewer clicks, fewer sessions, and often a change in how Googlebot treats the site. A logging error is inconsistent by nature: the appearances kept happening, so the downstream signals kept behaving normally, while the one broken counter told a different story.
That inconsistency is your best early warning. You do not need to wait for Google to confirm a bug to start suspecting one. If one report falls off a cliff while the other three signals carry on as normal, your first hypothesis should be a reporting fault, not a ranking disaster. One caution: do not treat any single row as proof on its own. GA4 sessions can dip for their own reasons — a holiday, a tracking hiccup, a slow news day in your niche. The test works because you are looking for agreement across signals. Three or four signals telling the same story is evidence. One signal on its own is a hint.
The 5-step “before you touch anything” checklist
When a graph falls, the urge to act is strong. Resist it for fifteen minutes and run this checklist first. It will save you from the most common mistake in SEO: fixing a problem that does not exist and creating a real one in the process.
Step 1: Check the Search Console data anomalies page. This takes about thirty seconds. Google posts known data issues there, including logging errors, delays, and outages. In this incident, the note appeared between November 12 at 15:00 UTC and November 13 at 03:00 UTC. If there is a note that matches your dates and your report, stop. You have your answer. Write it down, annotate the data, and do nothing else.
Step 2: Check for an annotation inside the report. Google now places an annotation in the Generative AI report itself for this incident. Open the report, look at the dates in question, and see whether a marker is there. Google does not annotate a report lightly. When you see one, read it before you read the graph.
Step 3: Compare the main Performance report’s clicks against the AI report’s impressions. Look at clicks, not impressions — clicks are the signal closest to real visitors. If clicks held steady while AI impressions fell, the appearances were still happening and people were still clicking through. That is the fingerprint of a counting fault. If clicks fell in step with the AI impressions, you may have a real problem, and it is time to dig deeper.
Step 4: Check GA4 organic sessions for the same dates. GA4 measures what happened on your site, not what Google recorded in search. If organic sessions for November 9–11 look normal — in line with the days before and after, allowing for the usual day-of-week wobble — then real traffic did not fall. A logging error cannot change how many people actually visited your site. Compare the same weekdays the week before and the week after, not a Monday against a Saturday.
Step 5: Check your server logs for Googlebot hits. If Googlebot kept crawling your site at its normal rate on November 9–11, Google had not stopped paying attention to your pages. You need one number: how many times Googlebot requested your pages on each of the three days, compared with the days around them. Most hosting dashboards will give you that in a couple of clicks. Normal bot activity plus normal sessions plus normal clicks plus a falling AI report equals a reporting gap, not a ranking event. The Google “crawled – currently not indexed” guidance is a useful companion read here: it covers what a real crawling-and-indexing problem looks like, which is a very different shape from a three-day impression dip with normal bot activity.
Only after all five steps should you even think about changing a page. If the five steps point to a logging error, the correct action is no action at all — except annotation. If they point to a real drop, you still should not rewrite anything during a core update rollout. Diagnose, document, and wait.
Annotate your data, or future-you will misread it
Here is the part almost everyone skips, and the part that causes the most damage over time. The incident is over. The numbers are back to normal. The temptation is to close the tab and move on. Do not. The three-day hole in your AI report will still be there in six months, sitting in the graph like a real drop, waiting for someone — you, a colleague, a client — to notice it and draw the wrong conclusion. The only thing standing between that future misreading and the truth is an annotation you write today.
Do it in three places. First, in Search Console itself, if your workflow uses annotations there. Google has already placed an annotation in the Generative AI report, but add your own note in your own words if your process allows it — something short, like “Nov 9–11, 2026: Google logging error, AI impressions undercounted, data gap only, not a ranking change.” Second, in GA4, where the people who look at it may never open Search Console. Third, in your client reports and internal documents: any monthly report covering November needs a one-line footnote — “AI impressions for Nov 9–11 are undercounted due to a confirmed Google logging error; treat those three days as a data gap.” Skip it and someone will eventually ask you to explain — or fix — a ranking loss that never happened.
Keep the note short, factual, and boring. Do not speculate about what the “real” numbers would have been. Google has not said the missing impressions will be backfilled, and presenting an estimate as a fact would be inventing data. Link to the anomalies page in your internal notes so the next person can verify it in one click.
The core-update timeline, and the don’t-rewrite rule
The November 2026 core update started rolling out on November 10, 2026, at around 10:07 a.m. PT, according to the Google Search Status Dashboard. It is expected to run for up to two weeks, finishing around November 24. Google’s guidance is to wait at least a full week after the rollout completes before judging what it did to your traffic — so the honest assessment window is early December, not mid-November. The logging error ran November 9–11 and was confirmed on the anomalies page between November 12 at 15:00 UTC and November 13 at 03:00 UTC. It affects the Generative AI performance report only, and only for those three dates. Do not judge the core update by data from November 9–11, and do not judge your AI visibility by a report that was undercounting for those three days.
The don’t-rewrite-pages-mid-rollout rule is the rule most often broken in weeks like this one. During a core update, rankings shift as the rollout progresses. A page that looks down on day three can recover by day ten. Changing content in the middle of that movement means you are reacting to noise, and you will never know whether the change helped, hurt, or did nothing. Wait for the rollout to finish, wait a full week after that, then compare the data with the period before the update started and make decisions with a stable baseline.
If you are managing clients, tell them now: the dip was a confirmed logging error for Nov 9–11, clicks and sessions were normal, and all content changes are on hold until the core update finishes and a clean week of data is in. Clients do not panic about problems. They panic about silence.
What to do now, and what not to do
The action list for this incident is short, and that is the point. A logging error does not need a recovery plan. It needs a note and a pause.
Do this: check the anomalies page and confirm the note matches your dates. Check the annotation in the report. Run the four-signal comparison — AI impressions, main report clicks, GA4 sessions, server-log bot hits — and write down what each one showed. Annotate the three days in Search Console, in GA4, and in any client report that covers November. Then leave your pages alone until the core update finishes and a full week has passed after that.
Do not do this: do not rewrite pages because the AI report dipped. Do not delete pages, merge pages, or change titles and headings in response to a counting fault. Do not present the November 9–11 AI impressions as a true reading in any report, and do not invent a backfilled number to fill the gap. Do not tell a client their AI visibility fell, because it did not — the record of it fell, and those are different things.
One last thing worth saying plainly: your site is probably fine. If your clicks held steady, your sessions held steady, and Googlebot kept visiting, then nothing about your site changed on November 9–11. Only a counter changed. The disciplined move this week is the boring one — annotate, wait, and check again when the data is clean.
Frequently Asked Questions
Did my site lose rankings in AI Overviews on November 9–11, 2026?
No — not according to Google. Google confirmed a logging error caused the decrease in impressions in the Generative AI performance report for November 9–11, 2026, and said the issue “affects data logging only.” The appearances in AI Overviews and AI Mode kept happening; the system that counts them failed. A ranking loss would have shown up in your main Performance report clicks and your GA4 organic sessions too. In this incident, those signals held steady, which is consistent with a counting fault, not a visibility loss.
Will Google backfill the missing AI impressions for November 9–11?
Google has not said the missing impressions will be backfilled. Treat November 9–11 in the Generative AI report as a data gap, not a true reading. Do not estimate what the numbers “should” have been and present that estimate as measured data. Annotate the three days as a gap, cite the anomalies-page note, and compare performance using the days before and after the window instead.
How is this different from the November 2026 core update?
They are two separate events that happened to overlap. The logging error ran November 9–11 and affected only the Generative AI report’s impression counting. The November 2026 core update started rolling out November 10, runs for up to two weeks (finishing around November 24), and affects rankings across search. The core update is still in progress, and Google’s guidance is to wait at least a full week after it completes before judging its impact. Do not use the November 9–11 AI report data to judge the core update — that data is incomplete.
Why did my AI impressions dip but my clicks stay normal?
Because the Generative AI report shows impressions only — no clicks — and the logging error hit the impression counting in that report alone. Clicks come from people actually arriving at your site, and that measurement was not affected. Steady clicks while AI impressions fall is the classic fingerprint of a reporting gap: the appearances and the visits kept happening, and only one counter broke.
Should I change or rewrite my pages because of this dip?
No. Do not rewrite pages in response to a logging error, and do not rewrite pages mid-rollout during a core update either. Wait for the core update to finish (around November 24), wait at least a full week after that, then judge your traffic on clean data. If you change pages now, you will not be able to tell whether the change helped or the rollout simply settled. Annotate the gap, keep monitoring clicks and sessions, and make content decisions on stable data.
Where can I verify this myself?
Check Google’s Search Console data anomalies page for the note about November 9–11, 2026, and look for the annotation inside the Generative AI performance report itself. Search Engine Land’s report and Search Engine Roundtable both covered the confirmation on November 12, 2026. Cross-check with your main Performance report clicks, your GA4 organic sessions, and your server logs for the same three days — if those are steady, you have confirmed the pattern on your own site.