SEO Case Study: 12 Months of Google Updates
In short: One client, 12 months, 4.37 million impressions, eight annotated Google updates. Clicks grew 4x. Not one update produced a visible step change. What did explain performance: roughly 3.7% of pages holding 42.5% of all top-three query positions, and a click-through rate quietly falling while the traffic line went up.
Research note: a single anonymised client property, Google Search Web data, “Final” data state, 26 August 2025 to 26 August 2026, no filters applied. Headline totals are exact as reported. Figures derived from the client distribution charts are calculated from the plotted percentages and marked as approximate. Where we read a value off the time-series chart, we say so.
1. What Happened? SEO Case Study: 12 Months of Google Updates
We took one live property, pulled a full year of Console’s data, and overlaid every confirmed Google update onto the timeline: the August 2025 spam update, the num=100 removal, the December 2025 core update, the March 2026 spam and core updates, the May 2026 core update, the June 2026 spam update, and the August 2026 spam update.
The question was simple. Twelve months, eight algorithmic events – which ones actually moved this site?
The headline numbers
| Metric | 12-month total |
|---|---|
| Clicks | 79,371 |
| Impressions | 4,374,617 |
| Average CTR | 1.81% |
| Average position | 7.4 |
On the face of it, a healthy year for our client partner. The trendline rises steadily from roughly 90 daily clicks in late August 2025 to roughly 330 by August 2026 – a fourfold increase in a period during which a great many sites went backwards.
Then we looked at the two things that actually explained the year, and neither of them was an algorithm update.
2. Why Does It Matter? SEO Case Study: 12 Months of Google Updates
Finding 1: Eight updates, no step changes
Every confirmed Google update from August 2025 to August 2026, overlaid on daily clicks and impressions. The line climbs steadily through all of them. No update produces a visible drop.
This is the finding worth sitting with, because it contradicts how most organisations behave.
Look at the shaded bands. The August 2025 spam update, the December core update, the March spam and core updates, the May core update, the June spam update – the line passes through all of them without a drop. Growth for Szymaniak Digital client here is gradual, continuous, and largely indifferent to the algorithmic calendar.
There is no step up after a core update. There is no cliff after a spam update. The rises that do appear – early November 2025, the run from June 2026 onward – do not align with any annotation.
- The one genuine boundary is not a ranking event at all.
The num=100 removal in mid-September 2025 sits at the very start of the window. That was a measurement change, not a performance change: it removed bot-generated impressions from deep result pages. It marks the point where the impression data becomes comparable to itself, which is why we would not compare anything before it with anything after it.
The practical consequence: for this property, over a full year, algorithm updates were close to irrelevant as an explanatory variable. Yet updates absorb an enormous share of the industry’s attention, client anxiety and reporting time. When a client asks “did the update hit us?”, this is the shape of data that usually answers no.
Finding 2: The traffic went up while the efficiency went down
Read the two lines against each other rather than separately.
Reading approximate values off the chart, the property opened the period around 75 daily clicks against roughly 5,000 daily impressions, and closed it around 330 daily clicks against roughly 32,000 daily impressions.
| Approx. start (Sept 2025) | Approx. end (Aug 2026) | Change | |
|---|---|---|---|
| Daily clicks | ~75 | ~330 | ~4.4x |
| Daily impressions | ~5,000 | ~32,000 | ~6.4x |
| Implied daily CTR | ~1.5% | ~1.0% | down ~one third |
Impressions grew half again as fast as clicks. The site is being shown far more often and converting that exposure less well each time.
This is the pattern that a clicks chart alone will hide, and it is the pattern most likely to be misread in a board report. Absolute growth of 4x is a genuinely good result. It is also entirely compatible with a deteriorating position in the results page – more visibility, less of it clickable.
Two explanations fit, and they are not mutually exclusive. The site may be expanding into weaker long-tail territory where it ranks lower and earns fewer clicks per impression. Or the results pages themselves are absorbing more clicks before the user reaches an organic link. Distinguishing between them requires segmenting by query type and by SERP feature presence – which is the analysis this property needs next.
Finding 3: Roughly 3.7% of pages hold 42.5% of top-three positions
This is where the year is actually explained.
The same property, two distributions. 42.5% of queries rank in the top three. Only 3.7% of pages do.
| Position band | Share of queries | Share of pages |
|---|---|---|
| Top 3 | 42.5% | 3.7% |
| 4–10 | 37.1% | 62.3% |
| 11–20 | 8.2% | 19.0% |
| 21+ | 12.2% | 15.0% |
Working from the plotted values, the property has approximately 300 pages and approximately 23,600 ranking queries.
Which means:
- Approximately 11 pages sit in the top three.
- Those 11 pages carry approximately 10,050 top-three query positions – around 905 top-three queries per page.
- The ratio of top-three query share to top-three page share is 11.5 to 1.
A fraction under 4% of the page estate is producing the large majority of the site’s competitive visibility. This is a Pareto distribution far more extreme than the usual 80/20 – closer to 96/42.
Two implications, pulling in opposite directions.
The concentration is a risk. Eleven pages are carrying the property. A technical fault, a content change, a competitor’s improvement, or a shift in how those particular queries are answered could remove a large share of visibility in a single move. Most organisations have no idea how few pages their performance depends on, and therefore no protection around them.
The concentration is also the opportunity, and it sits in the second row. 62.3% of pages – roughly 187 of them – rank in positions 4 to 10. These are pages Google already considers relevant enough to place on page one. They are one meaningful improvement from the top three, where the query volume actually lives.
Compare the effort profiles. Chasing an algorithm update is unfalsifiable, uncontrollable and mostly reactive. Moving 187 already-ranking pages from position 6 to position 3 is specific, measurable, and entirely within your control.
Finding 4: The clicks-per-query number sets a realistic ceiling
Across the year the client property averaged approximately 3.4 clicks per ranking query and about 185 impressions per query. At page level, roughly 264 clicks and 14,600 impressions per page per year.
These are useful planning numbers because they let you forecast honestly. If a new page performs at site average, it earns a few hundred clicks a year. Publishing 50 more pages at average performance adds roughly 13,000 annual clicks. Moving 30 existing page-one pages into the top three plausibly does considerably more, for less SEO production investment.
That is the trade every content budget is implicitly making, and almost nobody calculates it.
3. Who Is Affected?
- Any organisation whose reporting leads with a clicks line.
If your monthly report shows clicks and nothing else, you cannot see the efficiency story in Finding 2. A site can grow clicks 4x while its click-through rate falls a third, and the report will read as unambiguous success.
- Sites with large page estates and small winners.
Publishers, e-commerce and content-led SaaS in particular. The more pages you have, the more likely a tiny subset is carrying everything, and the less likely anyone has checked which ones.
- Teams whose planning cycle is organised around algorithm updates.
If your quarter is shaped by reacting to core and spam updates, this dataset suggests the effort is misallocated. Eight updates, twelve months, no visible step changes.
- Organisations planning content volume.
Finding 4 makes the arithmetic explicit. Publishing more pages at average performance is usually a worse investment than improving pages already on page one. And this is the work we are currently implementing for our partner client here at Szymaniak Digital SEO Consulting.
4. What Should Businesses Do?
Step 1 — Find your eleven pages (Analytics – one day)
Export your queries and pages by position band and calculate the same two distributions. You are looking for one number: what percentage of your pages hold top-three positions, and what share of your top-three queries do they carry?
Then protect them. Named owner, change control, monitoring on the specific URLs, and a rule that nobody edits them without review. Most organisations apply change control to checkout and none to the eleven pages generating their organic visibility.
Step 2 – Work the 4-10 band deliberately (Content + Development – one quarter)
This is your highest-return SEO work, and it is a finite list rather than an open-ended brief.
- Export every page in positions 4–10. In this case, 187 of them.
- Sort by impressions, descending. Impressions tell you where the demand already is.
- Take the top 30 and work them properly: does the page actually answer the query it ranks for, is the intent right, is the title earning the click, is the content current, does it have internal links from your strong pages.
- Use the eleven top-three pages as link sources. They have the authority; spend it deliberately rather than leaving it in a footer.
- Re-measure at 90 days. Position movement within the band is your leading indicator, before clicks respond.
Step 3 – Add efficiency metrics to reporting (Marketing – one week)
Stop reporting clicks alone. Add:
- CTR trend at constant position band — the single best early warning of SERP-level erosion
- Impressions-to-clicks growth ratio — if impressions grow faster than clicks, say so explicitly
- Position distribution, quarterly — movement between bands is the real progress measure
- Page concentration ratio — what share of pages carries what share of top-three positions
Step 4 – Annotate, then stop blaming (Analytics – ongoing)
Keep the update overlay. It is genuinely useful — but mostly for ruling causes out. When a client asks whether an update hurt them, an annotated twelve-month view answers it in seconds, and the answer is usually no.
Establish the discipline: an update is a candidate explanation, not a default one. Look for a step change aligned to the annotation window. If there isn’t one, the cause is elsewhere, and continuing to look at the algorithm is time not spent on the 187 pages.
Step 5 – Recalculate your content investment (Marketing – one day)
Take your own clicks-per-page and clicks-per-query averages and run the comparison honestly. For most sites with a mature property, improving what already ranks beats publishing more. Do the deeper analysis before the next content budget is signed off.
5. What We’re Watching Next
- Whether the CTR decline continues or stabilises.
A third of the click-through rate in twelve months is a steep slope. If it continues, absolute click growth will stall even as impressions rise. This property should be tracking CTR at constant position monthly.
- Whether the concentration tightens or spreads.
Eleven pages out of 300 is a fragile structure. The healthy direction is more pages entering the top three – which is exactly what the 4–10 programme is designed to test.
- What the August 2026 spike actually was.
The final fortnight shows daily impressions reaching roughly 90,000 and clicks around 900 – close to triple the surrounding baseline, with CTR broadly holding. It coincides with the August 2026 spam update annotation, but a single sharp spike with sustained CTR looks far more like a demand event than an algorithmic one. We would not attribute it to the update without segmenting the queries behind it, and neither should anyone else.
- Whether impression inflation becomes the norm.
As generative surfaces expand, impressions counted against clicks not delivered may become a structural feature rather than a temporary distortion. If so, CTR stops being a performance metric and becomes a measurement artefact – and the industry will need a new denominator.
What this study cannot tell you
One property, one vertical, one year. The absence of update effects here does not mean updates never matter – it means they did not visibly move this site, which had no apparent quality or spam exposure. A property with genuine policy problems would look entirely different.
The chart-derived figures are estimates read from plotted values and should be treated as approximate. The distribution percentages, headline totals and their ratios are exact as reported.
Most importantly, this analysis says nothing about why the site grew. Continuous, gradual growth uncorrelated with algorithmic events is consistent with steady content and SEO investment.
About Szymaniak Digital
Szymaniak Digital is an enterprise AI SEO consultancy working with brands on visibility across classical search and AI-mediated discovery.
The most common analytical error we encounter is not a wrong conclusion – it is a right conclusion drawn from the only chart anyone looked at. A clicks line going up is compatible with efficiency going down, with visibility concentrated in a handful of fragile pages, and with an entire content programme aimed at the wrong end of the estate.
If you want your position distribution, efficiency trend and page concentration analysed properly – rather than another report on whether the latest update hit you – that is work we do.
Sources and method
Szymaniak Digital analysis
- Anonymised live client property, Google Search Console Web data, “Final” data state
- Date range 26 August 2025 to 26 August 2026, no dimension filters applied
- Google update annotations: August 2025 spam,
num=100removal, December 2025 core, March 2026 spam and core, May 2026 core, June 2026 spam, August 2026 spam - Distribution figures derived from plotted percentages; time-series values read from the daily chart and marked approximate
