Skip to content

field note

August 2026 spam update: what Google said and what to check

Google updates

Shahid AliUpdated August 22, 2026all posts

August 2026 spam update: what Google said and what to check

At 09:27 Pacific on 18 August 2026, Google’s Search Status Dashboard logged a new ranking incident, and a minute later the note went up: “Released the August 2026 spam update, which applies globally and to all languages. The rollout may take a few days to complete.”

Google Search Central added one line on LinkedIn: “This is a normal spam update.” No blog post went with it. Its own spam updates page still carries a December 2025 date. So the primary record is two sentences, and everything below is me reading those two sentences next to Google’s written policies and what the trade press managed to get out of Google on day one. I have tried to keep the three apart.

The facts, and where each one comes from

This is the third spam update of 2026. The March one started on 24 March and was marked complete the next day, the fastest rollout the dashboard has recorded. The June one ran from 24 to 26 June. Before those, the August 2025 spam update took 27 days, so “a few days” is a Google estimate, not a promise, and I would not plan around it.

Spam update rollout time, per the status dashboard Aug 2025 27 days Mar 2026 1 day Jun 2026 2 days Aug 2026 2 days, 16 hours Scale: one day = 24px. Source: Google Search Status Dashboard incident history.

Google announced no new spam policy this time. Search Engine Journal checked and the spam policies page is unchanged since 15 May. The rules being enforced are the ones already published.

Google was asked what the update covers and gave the trade press a partial answer, reported on the day it launched: it targets sites violating “some of” the spam policies; it does not target link spam, it does not target the site reputation abuse policy, “and some other policies”; Google would not say what share of queries it touches; there will be periodic refreshes; and recovery “can take many months”. That last part matches Google’s own spam updates page, which says a site may improve “if our automated systems learn over a period of months that the site complies”.

That is the whole factual record as of the morning of 19 August. If you read a post this week that tells you exactly which sites got hit and why, ask which of those sentences it came from.

It is not an AI-content update, and here is the actual wording

I keep seeing this framed as Google going after AI writing. The policy text does not support that. Here is the scaled content abuse definition, quoted from the page dated 15 May 2026:

Scaled content abuse is when many pages are generated for the primary purpose of manipulating search rankings and not helping users. This abusive practice is typically focused on creating large amounts of unoriginal content that provides little to no value to users, no matter how it’s created.

The first example under it is “using generative AI tools or other similar tools to generate many pages without adding value for users”. The second is scraping feeds and search results. The third is stitching pages together. Generative AI is one method in a list of methods. The offence is “many pages” plus “not helping users”. A thousand thin pages out of a spreadsheet macro in 2016 and a thousand thin pages out of a language model in 2026 read the same to that policy, and Google’s guidance since February 2023 has been that content is judged “however it is produced”.

There is one genuinely new thing since the last time most people read that page. On 15 May 2026 Google widened the opening definition of spam to include “attempting to manipulate generative AI responses in Google Search”. So the same policies now cover AI Overviews and AI Mode, not just the ten links. The June update followed that change by five weeks. Google has not said whether August continues it. I would assume the surfaces are all one thing now and stop treating “ranking in AI Overviews” as a separate game with looser rules, because Google has stopped treating it that way in writing.

So if you publish with AI, the honest version of the risk is: not that you used a model, but that you used it to produce pages at a scale that only makes sense if the goal was to occupy results, and none of those pages says anything a reader could not get from the first result they already saw. If that describes a folder on your site, this update is aimed at that folder. If it does not, it is not aimed at you.

Which policies to check yourself on this week

Given what Google said there, I would not spend this week auditing backlinks or third-party subfolders. Those have their own updates. The policies I would read against my own site, in the order I actually find them on client sites:

Scaled content abuse, above. Count pages, then ask what percentage of them a person would miss if they vanished.

Doorway abuse. Google’s own example is “multiple domain names or pages targeted at specific regions or cities that funnel users to one page”. Local SEO is full of this: one template, sixty town names, nothing on the page that changes with the town except the town. I wrote about what separates a location page that indexes from one that does not, and the test is the same one the policy applies. If nothing but the place name changes, it is a doorway.

Expired domain abuse. Buying a domain with history and filling it with unrelated content. Google’s examples are blunt: casino content on a former primary school site, affiliate pages on a former government domain.

How to read Search Console until the rollout ends

This is where people lose money, and it has nothing to do with spam. The update-agnostic version of this routine is written up in full as what to do during a Google update; the short form follows.

Do not act on this week’s Performance data. During a rollout, positions move, settle, and move again. A page that drops fifteen places on Tuesday and is back on Friday has not been hit; it has been re-evaluated twice. If you rewrite it on Wednesday you have destroyed the only clean read of what the update did to it. Search Engine Journal put it mildly: a single day’s data may not capture the full picture. I would put it less mildly. A single day’s data during a rollout is noise with a date on it.

What I do instead:

  1. Write down 18 August. Search Console has no annotations, so this goes in a spreadsheet, and I put it against every property I touch.
  2. Wait for the dashboard to say complete. Not a tweet, the dashboard.
  3. Compare the seven days after completion with the seven days before 18 August, matched by weekday. Impressions first, then clicks. Impressions tell you whether Google is still showing the page at all.
  4. Look at pages, not the property total. Spam updates rarely shave a flat percentage off a site. They take out a class of pages. If traffic fell 20% and 90% of the loss sits in one folder, that folder is the finding.
  5. Check indexing separately from ranking. Spam systems can drop pages from the index, not just push them down. A group of URLs moving from Indexed to “Crawled - currently not indexed” in the same week is a stronger signal than any position change. The bulk URL inspection tool on this site gives per-URL status for a folder in one run, and I took a snapshot of every template-driven folder I look after on 18 August so I have a before.

If it turns out you were hit

Two things to have straight before you touch anything.

First, this is algorithmic, not a manual action. Nothing will appear under Manual actions in Search Console, there is no reconsideration request to file, and Google’s language for recovery is “months”, with periodic refreshes of the update along the way. That is slower than most people expect and it is worth knowing before you promise anyone a timeline.

Second, the fix is by class of page, not by URL. Find the folder or template. Decide honestly whether it deserves to exist. Most scaled content does not, and deleting a thousand pages that never earned a click is a cleanup, not a loss; it usually helps what remains. For pages worth keeping, add the thing they were missing: first-hand information, original numbers, a real position, checked facts. Not more words. If a page cannot carry any of that because there was never anything to say, it belongs in the first group. Whatever you remove, return a proper 410 or a 301 to the nearest genuinely relevant page. Not a thousand redirects to the homepage.

What I am doing on client sites

Nothing, until it is marked complete. The date is written down on every account and the indexing snapshots are taken. That is the plan for the week, and it is a boring plan on purpose.

I can be calm about this one because the work I do is the opposite of the pattern it targets: fewer pages, each with a reason to exist, checked one at a time until Google agrees. If your site is built that way you can be calm too. If it is not, this week is a cheap time to find out.

I will add the before-and-after numbers to this note when the dashboard says the rollout is done.

Update, 20 August 2026: still rolling, and the early volatility was not this update

Day two, and the dashboard still shows no completion. “A few days” has already outlived the March update, which was done in one.

Two things from yesterday worth adding to the record. First, the temperature. Day two read as heated across the forums, and that matches what the tracking tools showed within the first 24 hours. Heat during a rollout is normal and it is exactly why the plan above says wait; a heated day two tells you nothing about where your pages land on day five.

Second, and more useful: several people looked at their Search Console data, saw movement starting three to five days before 18 August, and decided Google had been quietly rolling this out early. John Mueller answered that directly on LinkedIn: “We don’t roll them out beforehand.” Google’s position, repeated for years now, is that the announcement goes out as close to the actual start as they can manage. So if your charts moved on the 14th or 15th, that movement belongs to something else, probably one of the unconfirmed adjustments Google ships constantly and never names.

This matters for the spreadsheet method above. The clean “before” window ends on 17 August, not wherever your volatility happens to start. If you backdate the update to match your own wobble, your before-and-after comparison will blame the spam update for a change it did not make, and whatever you fix in response will be aimed at the wrong system. Keep the line at 18 August. That is the only date Google has put in writing.

Nothing on my client accounts has changed and I have changed nothing. Still waiting for the word complete.

Update, 22 August 2026: the rollout is complete, and here is the window to compare

The dashboard now carries the closing line: “The rollout was complete as of August 21, 2026.” The incident entry reads 2026-08-18 09:27 to 2026-08-21 01:49 Pacific, which is two days and sixteen hours. In Eastern time that is 12:30pm on 18 August to 4:50am on 21 August.

So “a few days” held. Longer than March, which was done inside a day, longer than June at roughly two days, and nothing like the 27 days of August 2025. That spread is not unusual for spam updates, and I have since timed the last ten rollouts from the dashboard in how long Google updates take to roll out. The chart above is updated. If you were waiting to look at your data, the wait is over.

The comparison window, precisely:

  • Before: 11 to 17 August, the seven full days ending the day before the start.
  • After: 22 to 28 August. Not 21 August. The rollout ended at 01:49 Pacific on the 21st, which means part of that day’s traffic was still being served by a partly rolled out system, and Search Console days are Pacific days. Throwing the 21st away costs you nothing and keeps the after window clean.
  • Match weekday to weekday. Both windows above are Tuesday to Monday, so a like-for-like comparison is already built in.

One thing to be careful about this month: Google’s data anomalies page records a logging error that suppressed Generative AI in Search impressions from 13 to 17 August 2026, and another that hit Discover clicks and impressions on 13 August. Both sit inside the before window. If your property has AI or Discover data in it, filter those out of the comparison or your “before” is artificially low and the update will look like a win it did not earn.

Also worth knowing while you read charts this week: the crawl stats report is missing 15 and 16 August for many properties. That is a reporting gap, not a crawling gap, and it is not related to this update. Nor is the other thing people are pinning on the update this week: favicons disappearing from search results, which Google has confirmed as a bug on its own side.

What I am doing now, in order: pull the two windows per property, look at impressions before clicks, split by folder rather than site total, then compare the indexing snapshot I took on 18 August against a fresh run of the bulk URL inspection tool to see whether anything moved out of the index rather than just down the page. A correction to what I wrote on 19 August, when I said I would add the before-and-after numbers to this note. I will not, and I should not have promised it: the numbers are client numbers and they are not mine to publish. My own property is far too small to read anything into either. What I can do is describe the pattern across accounts once I have it, without numbers I do not own, and that will be its own post rather than another update here.

If nothing in your data moved: that is the most common outcome of a spam update, and it is the correct outcome for a site that was never doing the things the spam policies name.

One caution for anyone reading traffic in this window. Google confirmed on 26 August 2026 that it is routing result links through a google.com/goto redirect. It does not touch rankings, but it can change what your analytics records as the referrer, so a shift between organic and direct in late August has two candidate causes now, not one.

Sources