Pages Deindexed After a Google Update? Here Is What We Measured During the August 2026 Rollout

Google began rolling out the August 2026 spam update on 18 August 2026, and as of the morning of 20 August the Search Status Dashboard still shows no completion time. If your traffic moved this week, you are probably trying to answer one question: did Google drop my pages, or just rank them lower?

Those are two different problems with two different fixes, and the tool everyone tells you to check cannot reliably answer either one today.

So we measured it instead. On 11 July 2026 we froze a cohort of 289 URLs across 252 domains and 13 content categories, and confirmed 288 of them were in Google's index. On 20 August, two days into this rollout, we re-checked every one.

We also have a more uncomfortable dataset: this site lost about 97% of its own Google visibility on 5 August and has not recovered. We publish that below too, because it is the cleanest example we have of the distinction this whole article turns on.

What Google actually said

The full statement, which every write-up you will read this week is a paraphrase of, is 22 words long:

Released the August 2026 spam update, which applies globally and to all languages. The rollout may take a few days to complete.

That is the entire announcement. Google named no new spam policy alongside it, and its spam policies page still lists the same 16 categories it did last week. Anyone telling you this update "targets" something more specific is inferring, not reporting.

Some useful context from Google's own Search Status Dashboard, which is worth more than most commentary: this is the third spam update of 2026. The March 2026 spam update completed in 19 hours and 30 minutes. The June 2026 spam update took 2 days and 1 hour. There have been six confirmed ranking updates in 2026 so far, including two core updates that each ran about 12 days. A multi-day rollout is normal, not ominous.

What we found: 98.3% of indexed pages held

Here is the result of the re-check, run on 20 August 2026 against the cohort frozen on 11 July.

Measure	Result
URLs confirmed indexed at baseline (11 Jul)	288
Still retrievable on 20 Aug	283 (98.3%)
No longer retrievable	5 (1.7%)
Transient misses that recovered on re-check	0

Over 40 days spanning two confirmed Google updates, fewer than 2 in 100 previously indexed URLs left the index. That is the number worth holding on to before you start changing anything, and it is consistent with what the two most credible independent analyses of recent deindexing events found.

A necessary caveat, because you should not trust an analysis that omits it: this window runs from 11 July to 20 August and contains the June and August spam updates plus ordinary index churn. We cannot attribute these five drops to the August update, and neither can anyone else. Two days is not enough time, and a 40-day baseline is not a controlled experiment. What this measures is index stability, not causation.

The five that dropped, and why four of them are not what they look like

This is where the exercise got interesting, and where it stops being a statistic and becomes a method you can copy.

We checked all five by hand rather than trusting our own tooling, which matters, because an index checker reporting "not indexed" can be wrong.

One was a site migration. doctorsofrunning.com had a post at a dated URL (/2023/06/monday-shakeout-shoes-for-flat-feet.html) that now 301-redirects to a clean slug (/monday-shakeout-shoes-for-flat-feet/). The old URL is correctly gone from the index because the site moved it. Nothing to do with any algorithm update. If you changed your URL structure this year and are now reading a drop in indexed pages as a penalty, check this first.

Four return HTTP 200 and are genuinely absent. Three are IRS.gov retirement pages, including irs.gov/retirement-plans/roth-iras. One is a UpToDate clinical reference page. All four load fine in a browser. We confirmed each absence with at least two independently phrased queries (four for the Roth IRA page), while sibling pages in the same IRS section returned normally, so this is not a quirk of one query form.

We do not know why. We are not going to pretend we do. But it is a clean illustration of the thing that makes this problem expensive: a page can return a perfect 200, on one of the most authoritative domains on the internet, and simply not be in the index. Nothing on the server is broken. Nothing alerts you. You find out because you looked.

Why Search Console cannot answer this today

Every guide to this update, and Google's own AI Overview on the topic, tells you to check the Page Indexing report in Search Console. That is correct advice on a normal week and unhelpful advice this week, for a reason the SEOs living through the rollout have been saying out loud in the same threads that rank for it: the data is lagging and unstable right now.

Search Console's indexing and performance reports run a few days behind. During a rollout that started 48 hours ago, that lag covers the entire event. You are looking at a partial picture of a moving target and deciding whether to delete pages based on it.

A direct index check does not have that problem, because it asks a narrower question. Not "how is this page performing", which needs aggregation and therefore lag, but "is this URL retrievable from the index right now", which is a live lookup. For a handful of important pages you can do this by hand: search site: followed by the exact URL. For more than a dozen, a bulk index checker does the same thing across a list in one pass, free.

That distinction is the whole point. Use the live check to establish what is true today. Use Search Console next week, once the rollout completes and its data settles, to understand what changed.

Deindexing and traffic loss are mostly separate events

This is the finding that should change what you do next, and it runs against the instinct that a monitoring company like us is commercially incentivised to encourage. We will say it anyway, because the data is not ambiguous.

Indexing Insight, analysing 2 million monitored pages through the May 2025 indexing purge, found that sites lost between 15% and 75% of their indexed pages, and that the removed pages had essentially zero impact on clicks and impressions. One site in their data lost 75% of its indexed pages and its clicks went up.

Marie Haynes, reviewing affected sites by hand over the same period, reached the same conclusion: "In the majority of the cases I looked at, this issue does not appear to have affected clicks." The pages Google dropped were old archives, thin definition posts, paraphrased content and pages built for search engines rather than readers. Pages nobody was reading.

Two independent analyses, different methods, same answer. Losing indexed pages and losing traffic are usually not the same event. So if your traffic fell this week and your indexed page count did not, deindexing is not your problem and fixing it will not help. And if your indexed count fell but your traffic did not, you may be watching Google prune pages that were never earning anything, which is not a crisis.

We are the case study, and it is not a flattering one

We are not writing this from a comfortable distance. On 5 August 2026 this site lost most of its Google visibility, and it has not come back.

Here is our own Search Console data, and we would rather publish it than write another article implying we have this figured out.

Window	Impressions/day	Distinct pages surfacing	Distinct queries
28 Jul to 4 Aug	776	73	397
13 Aug to 19 Aug	24	62	7

Read those last two columns together, because that is the entire point of this article.

Impressions fell about 97%. The number of pages still surfacing fell about 15%. The pages did not leave the index. They are still there, they still get retrieved, they simply stopped being retrieved for anything. The query count is the tell: we went from being pulled into 397 distinct searches to being pulled into 7.

Everything we could check, we checked, and it all came back clean. site:searchoptimo.com returns the site in full. We rank first for our own brand name. URL Inspection said "Submitted and indexed" with robots allowed and fetches succeeding. Our crawl stats over the surrounding 62 days were 94.4% 200 responses with no server errors and no error spike on 5 August. Four sibling properties on the same Search Console account ran smoothly straight through the same days, which rules out a reporting artifact.

There is one more detail worth your attention, because it is the thing that told us this was not a technical fault. On 10 to 12 August it all came back. 1,171 impressions, then 2,476, then 1,682, the last of which was higher than any day before the drop. Then on 13 August it collapsed again to 34. Nothing on our end changed on any of those days. A site that is broken does not spontaneously work perfectly for 72 hours.

We do not have a happy ending for you. Our current read is a site-level algorithmic demotion, and there is no configuration change that fixes that. What we do have is the point this article opened with, proven at our own expense: we lost nearly all of our search traffic without losing a single page from the index. If we had spent August diagnosing this as a deindexing, auditing sitemaps and requesting reindexing on 73 URLs, we would have burned two weeks fixing something that was never broken.

Check which problem you have before you fix it. We are recommending that because we had to learn it the slow way.

What to do while the rollout is still running

Short version: record, do not react.

1. Establish today's baseline. List the URLs that actually matter, the ones that earn traffic or revenue, and check whether each is currently in the index. Save the result with today's date. This is not just prudence, it is what Google's own core update guidance asks you to do: confirm the update has finished, wait a full week, then "try comparing this week with a week before the core update started rolling out". That comparison is impossible unless somebody recorded the before. A week from now you cannot go back and get one.
2. Separate the two questions. Is the page in the index (a site: check answers this), and is it ranking where it was (that is a different investigation). Do not let one symptom stand in for both. Search Engine Journal has a useful breakdown of the five distinct causes that all present as "deindexing".
3. Rule out the boring explanations first. A URL migration, a stray noindex, a canonical pointing elsewhere, or a robots rule will all look exactly like an algorithmic hit and none of them is one. Our own top drop was a redirect. If you see crawled but not indexed in Search Console, that is a distinct status with a distinct cause.
4. Change nothing structural yet. Deleting pages, rewriting sitemaps or disavowing links mid-rollout, based on lagging data, is how a temporary fluctuation becomes a permanent problem. Rankings commonly move during a rollout and settle after it.

The one thing worth doing today costs a few minutes: know which of your pages are in the index right now, so that whatever the dashboard says next week, you can tell what actually moved.

The honest summary

Google is running a normal spam update, of a normal kind, at a normal cadence, and it is the sixth confirmed ranking update of 2026. Across 288 URLs we had confirmed indexed 40 days ago, 98.3% are still there. Of the five that are not, one moved house and four are unexplained absences of pages that load perfectly well.

None of that means nothing happened to your site. It means the odds favour a ranking change over a deindexing, that the two need different fixes, and that the only way to know which one you have is to check rather than infer. Our own site is the proof in the uncomfortable direction: 97% of the visibility gone, and every page still indexed.

If you want that check to happen on a schedule instead of on the day you finally think to look, that is precisely what SearchOptimo does: it re-checks your URLs on a fixed interval and tells you when one falls out, which is the difference between finding out today and finding out in six weeks. It would not have saved our rankings in August. It did tell us, within a day, that our pages were still indexed, which is the reason we did not spend two weeks fixing the wrong thing.

Rollout status as of 20 August 2026: the August 2026 spam update is listed as started on 18 August with no completion recorded on Google's Search Status Dashboard. We will update this post when it completes.