Four UK Sites Through Google's August Spam Update: One Collapsed, Three Kept Growing
Daily Search Console data for four UK sites before, during and after Google's August 2026 spam update: which one lost nearly all its impressions, which three kept growing, and what the focused sites had in common.
Awais M.
Founder of GeoRankLocal
When Google announced its August 2026 spam update, we had four UK sites in Search Console: our own, a railway-carriage marketplace, a local padel venue and a brand-new council-tax data site. Four sites is not a study. But it is four sets of daily numbers from before, during and after an update, and almost nobody publishes those.
Three of the four showed no lasting dip. One lost nearly all its impressions for about a week, and that one is ours.
What the update was
Google rolled out an August 2026 spam update globally between 18 and 21 August, according to Search Engine Roundtable. A second one followed: the September 2026 spam update started on 24 September and finished on 8 October, 13 days and 16 hours later, the longest of any 2026 spam update. Google describes it as an improvement to its automated spam detection, says affected sites may take months to recover, and offers no reconsideration request for algorithmic hits.
Our exports end on 27 September, four days into the September rollout, so this article is about August only.
The four sites, day by day
Choose a site and read any day. The shaded band is the August rollout. The hatched band is the September one, past the end of our data.
Did Google's August spam update touch your sites?
Daily Search Console data for four UK sites. Pick a site and a metric, then move across the chart to read any day.
GEO and web-design agency site. Broad topics: free tools, blog posts, service pages.
4–17 Aug
114 /day
Before the update
22 Aug–4 Sep
54.8 /day
−52% vs before
14–27 Sep
277 /day
+144% vs before
What the numbers say
Impressions per day, averaged over two weeks before the update (4 to 17 August), two weeks just after it (22 August to 4 September) and the latest two weeks we have (14 to 27 September):
| Site | Before | Just after | Latest |
|---|---|---|---|
| georanklocal.com | 114 | 55 | 277 |
| sleepertrain.co.uk | 581 | 1,583 | 2,428 |
| Padel Chippenham | 50 | 66 | 104 |
| Council-tax data site | 222 (7 days only) | 3,870 | 9,045 |
sleepertrain.co.uk had its two softest days on 20 and 21 August (552 and 481 impressions, from 826 on the 19th), then jumped to 1,315 on the 23rd and kept climbing. Padel Chippenham shows nothing unusual. The council-tax site, launched in Search Console on 11 August, was on a few hundred impressions a day until 26 August and then took off.
georanklocal.com is the exception. Daily impressions had held between roughly 40 and 170 since mid-July. On 16 August they fell to 56, then to between 6 and 14 a day from 17 to 21 August. After a one-day rebound to 144 on 23 August they stayed mostly in single digits and teens, apart from short spikes, until 7 September.
Why ours collapsed, as far as we can tell
We wrote this up at the time in an internal diagnosis. The short version: georanklocal.com had no coherent identity in Google's eyes. In the 28 days before the update, one free tool page produced about 80% of our impressions, almost all from global generic queries like "find questions", mostly from India, the US and Pakistan, not the UK. The rest of the visible footprint was unrelated: an invoicing blog post and an app-privacy page that had nothing to do with a UK web-design agency.
The three sites that held up each have one clear subject: railway carriages, a venue in one town, council tax. Their pages answer the thing the site is about.
That is a pattern across four sites, not proof of cause. Google does not say what a spam update targets, and nobody outside Google can isolate its effect.
And then it came back, partly
By 8 September our daily impressions were back above their old level and by mid-September they were averaging 277 a day, 2.4 times the pre-update figure. We would not read that as the update reversing. Google says recovery takes months and our visibility now looks different: of the 78 clicks Google attributes to queries it lists for 31 August to 27 September, at least 50 came from three queries that contain our own tool names ("question finder", "georank score", "geo rank score"). Our commercial pages are still not ranking: the web-design-for-tradesmen page averaged position 72.
What to take from it
- One page carrying a whole site is a risk. If 80% of your impressions come from a single page, a quality re-evaluation can take your visibility with it.
- Focus helped, or at least did not hurt. The sites with a single subject and pages that served that subject kept growing. Off-topic pages are the kind of thing quality systems can penalise, so remove or move them.
- Watch the right numbers. Total impressions can collapse while the traffic that pays you barely moves. Check what you lost before you panic.
- Keep the daily data. Search Console keeps 16 months, but an export a month is all it takes to be able to answer "what happened and when" later.
Method and limits
- Source: Search Console daily Performance exports (Web search type), up to 27 September 2026. georanklocal.com had no export for 27 to 30 August.
- Four sites, all small, three of them new and growing, so growth can hide a dip. Spam updates affect sites differently and Google does not publish what any given update targets.
- Averages are of daily impressions in each window. The council-tax site launched in Search Console on 11 August, so its "before" window has only seven days of data and no percentage change is shown.
Sources
- Google's August 2026 spam update, Search Engine Roundtable
- September 2026 spam update complete, Relevant Audience
Read next: Same Google position, ten times the clicks
Awais M.
Founder of GeoRankLocal
Awais M. is the founder of GeoRankLocal, a UK-wide agency that builds AI-citable websites and manages ongoing GEO and SEO for businesses across the United Kingdom. He writes about generative engine optimisation, AI search and what the data says about how customers find local businesses.
Try our free tools
Put this into practice. Check your AI visibility, see if AI recommends you, and find the questions worth answering — all free, no signup.
Live GEO Score
Score your website across 5 AI-readiness categories in seconds.
AI Citation Checker
See whether ChatGPT, Gemini, and Perplexity recommend your business.
Question Finder
Find the real questions your customers ask — ready to turn into content.
All Free Tools
Browse every free GEO tool in one place. No signup required.
Built for your trade
We build AI-citable websites for UK tradespeople — priced, scoped, and written for the way your customers actually search.
Want to get your business cited by AI?
Get a free AI visibility audit from GeoRankLocal.
Related articles
241,000 Impressions, 1,042 Clicks: What Fact-Lookup Pages Earn in 2026
Seven sites, one pattern: pages that answer a single fact need hundreds of impressions per click. Searches that include the current year clicked about five times as often. Interactive data from our Search Console exports.
Google's Keyword Planner Said "No Data" for 39 of 69 UK Trade-Website Searches. Here's What That Means for 2026
We ran 69 trade-website phrases through Google Keyword Planner for the UK. More than half returned no data. What that says about how people look for services, and why volume tools miss part of the picture.
We Audited Two Real Websites for AI Search Readiness. Here's Exactly What Separates What Works From What Doesn't.
We ran our full AI-visibility audit on two real websites outside our usual trades niche — a free-tools utility site and an 86-location retail chain — and compared every finding against live, dated screenshots: what gets an AI system to cite a site, and what quietly keeps it invisible.
Best Electrician in London, Manchester, Birmingham, Leeds & Glasgow: Google vs Gemini, With Screenshots of Every Search
We tested "best electrician in [city]" on Google and Gemini across all five cities, with live screenshots of every search. Two of five cities showed zero overlap between what Google shows and what Gemini recommends — one showed the strongest overlap in our entire study.