A site growing during a search update is not evidence that the update rewarded it. CodeQuest.work recorded +41% clicks and +49% impressions in the 28 days after the July 2026 core update completed — yet decomposition found almost no contribution from the update itself. Among the pages present in both periods, 37% gained position while 45% lost it. 55% of the click growth came from pages published during the window, and the growth on existing pages came from impression expansion, not from ranking higher. Separating an update from your own growth takes four checks: look for a step in the daily curve, split new pages from surviving ones, count the position distribution rather than the average, and decompose impressions from CTR.
Observation: Google Shook Three Times in July
July 2026 combined one confirmed core update with two bursts of volatility Google never acknowledged. The timeline first.
| When | What | Confirmed? |
|---|---|---|
| through Jul 9 | July core update completes rollout — the 4th confirmed update of 2026 | Confirmed |
| Jul 18–19 | Weekend volatility picked up by 14+ SERP trackers simultaneously | Unconfirmed |
| Jul 23–24 | A second round of broad volatility reported | Unconfirmed |
July gave everyone three separate reasons to feel volatility. In a month like that, whether you rose or fell, the temptation is to attribute it to the update. This report tests that temptation against our own data.
Sources: Search Engine Roundtable / Globital / Google Search Status Dashboard
Measured: +41% Clicks in the 28 Days After Completion
The subject is CodeQuest.work, our own site of developer tools and technical articles. Using the July core update's completion date (July 9) as the boundary, we compared the 28 days after, the 28 days before, and the 28 days before that, in Google Search Console.
| Period | Clicks (vs prior) | Impressions (vs prior) | CTR |
|---|---|---|---|
| 5/15–6/11 | (baseline) | (baseline) | 5.33% |
| 6/12–7/9 | +55% | +79% | 4.63% |
| 7/10–8/6 (post-update) | +41% | +49% | 4.38% |
Clicks +41% and impressions +49% against the prior 28 days; +120% and +167% against three months earlier. On the numbers alone, it's tempting to claim we landed on the winning side of July's volatility. But one thing is already off: CTR fell monotonically, 5.33% → 4.63% → 4.38%. If we had risen into higher positions, CTR should normally have risen with it.
Check #1: Is There a Step in the Daily Curve?
If an update moved your rankings, the daily data forms a step: the level shifts at the rollout completion or the volatility dates. Here are the weekly impressions, indexed to 100 at the first week.
100 → 107 → 133 → 183 → 188 → 231 → 264 → 321 → 288 → 320 → 366
There is no step at July 9, none at July 18–19, and none at July 23–24 — just one continuous ramp running since late May. This is not the shape of a curve lifted by an update; it looks like the continuation of something already underway.
Check #2: Positions Actually Fell
Next we isolated the pages present in both periods — the survivors — and counted the position change page by page. The distribution matters, not the site average: an average gets dragged around by shifts in query mix and hides what actually happened. The survivor set numbers in the hundreds of pages; here is how it split.
Position change across surviving pages (share)
- •Gained position: 37%
- •Lost position: 45%
- •Flat: 18%
Losses outnumber gains. In the very window where traffic grew 41%, the existing pages tilted toward losing position, not gaining it.
The survivors' clicks rose 19% and impressions 27%. Impressions growing while position falls means the query surface widened — we were not promoted; we were matched against more queries. Picking up new long-tail queries lowers average position, raises impressions, and dilutes CTR. That accounts for the monotonic CTR decline we saw at the start.
For how to read position and impressions separately in GSC, see How to Verify SEO Results in Google Search Console.
Check #3: 55% of the Growth Came From Pages We Published in the Window
Finally we allocated the click delta across newly appearing pages, surviving pages, and pages that dropped out. This is the decisive split: leave new pages in the comparison and a site will almost always read as 'grown.'
| Segment | Share of click growth |
|---|---|
| Newly appearing | 55% |
| Surviving | 46% |
| Dropped out | -1% |
More than half of it was simply the pages we shipped. And on the surviving side, the single largest gainer has an explanation that has nothing to do with the update.
The biggest gainer was a change we made ourselves in late July
The top surviving gainer (page C) moved from position 10.4 to 6.0 and accounted for roughly a quarter of the entire growth across surviving pages on its own. It is the page where, on July 30, we found a high-impression query missing from the title and rewrote the title to match. The update did not lift it; closing a title–query mismatch did. No other page under the same update behaved this way.
Meanwhile, some pages were cut down
Underneath the aggregate growth, specific pages clearly lost. Tool A fell from position 7.7 to 21.1 — off page one entirely — and article B lost 42% of its clicks alongside a 7.0 → 9.0 slide. When the sitewide number is up, losses like these vanish completely. Winning on the total and protecting each asset are different things.
Four Checks Before You Credit an Update
The checks we ran transfer directly to any site. They need nothing but GSC data — no third-party tools.
1. Look for a step in the daily curve
An update's effect shows up as a level shift at the rollout boundary. A smooth ramp through the date is not update-driven. Aggregate to 7-day totals so the weekday cycle doesn't fool you.
2. Separate new pages from surviving pages
A site that keeps publishing will grow in any before/after comparison, update or not. Restrict to pages present in both periods, or you'll mistake your own publishing cadence for algorithmic approval. Ours was 55% new pages.
3. Count gains and losses — don't read the average position
GSC's average position drops merely because you match more queries. A site that looks like it 'lost position' may simply have widened its surface. Counting how many pages rose and how many fell is closer to the truth.
4. Decompose impressions from CTR
A genuine position gain lifts impressions and CTR together. Impressions up with CTR down means surface expansion, not promotion. CTR moving alone points first at a title or meta change, not at the algorithm.
Only after all four hold can you say the update moved you. In our case, not one of them did.
But These Four Checks Only Look Inside Google Search
Everything above runs on GSC data alone. Which means any change that happened on a surface GSC doesn't cover is, by construction, invisible to this decomposition. AI surfaces are exactly that. And the two things usually lumped together as 'AI' are measured in completely different places.
| Surface | Where it happens | Where you measure it |
|---|---|---|
| AIO (AI Overviews, etc.) | Inside the Google results page | Blended into GSC's numbers — but not separable |
| GEO (ChatGPT, Perplexity, etc.) | Outside Google entirely | Invisible to GSC. Only visible in GA4 referral sources |
First, an honest limit. GSC has no filter that isolates AIO. You cannot read 'how many times we were cited in an AI Overview' out of GSC, which means the impressions and clicks decomposed above still carry AIO's influence mixed in. Isolating AIO's contribution is not possible with this method today.
The outer surface, GEO, is checkable in GA4. Counting AI referral sessions across the same two periods:
AI referral sessions (measured in GA4)
| Source | Change in sessions (vs prior period) |
|---|---|
| All AI sources | +4% |
| (reference) Google search | +34% |
AI referrals were essentially flat. The total held while the mix shifted underneath — ChatGPT down, Gemini up. That said, AI referrals are a small fraction of all sessions here, far too thin a base to call a trend. The only claim they support is that this period's growth happened on the search surface, not the GEO surface. The conclusion stands.
The operational implication: don't collapse AIO and GEO into one 'AI strategy.' The tactics overlap, but the instruments differ — and the moment you collapse them, you lose the ability to say which surface moved. Being able to state 'search grew, AI didn't' here is purely a consequence of measuring the two separately.
Google states plainly that "optimizing for generative AI search is optimizing for the search experience, and thus still SEO." Note the scope: that is Google's position on Google's generative AI search. No equivalent statement exists from OpenAI, Perplexity, or others — on those surfaces we have observed referrals without any official guidance.
What This Observation Shows — and What It Doesn't
What holds
- •Growth during an update window is not, by itself, evidence of the update's effect
- •On a site that keeps publishing, the majority of a before/after delta can be new pages
- •A sitewide rise can conceal individual pages falling off page one
What this cannot claim
- •It cannot claim the July volatility had no effect — only that it wasn't needed to explain this site's growth. Other verticals and other scales may differ (n=1)
- •Because publishing and volatility ran concurrently, we cannot rule out that new pages simply offset update-driven losses. The tilt toward losses among survivors is consistent with that reading too
- •Google never confirmed the July 18–19 and 23–24 movements. Trackers reacting is not the same fact as an algorithm update shipping
When the May core update landed, Google stated plainly that a ranking drop doesn't necessarily mean your page is broken. This observation shows the mirror image: a rise in traffic isn't necessarily evidence that an update rewarded you.
Related measurements: For how to read Google's official statements, see the May 2026 core update report; for the conditions behind new-page launch velocity, see SEO as the Base, AIO/GEO on Top. This report is the third: whether that growth can be credited to the update at all.
Capture a Baseline Around the Volatility
Attribution gets sharper the better your pre-volatility baseline. Enter a URL for a 45-item audit of technical foundation, structured data, meta tags, and E-E-A-T signals — your comparison point when the next shake arrives.
Written by
今井政和SEO Director / Frontend Developer
SEO Director with 20+ years of web industry experience. Creator of Direbase and the official WordPress plugin "ORECTIC SEO CHECK." Author of a book on web strategy inspired by Edo-era merchant principles.
@imai_directorFAQ
If my site grew during an update, was that the update's doing?▾
How do I test whether an update actually moved me?▾
Why do impressions and clicks rise while position falls?▾
If the sitewide numbers are up, can I ignore individual pages?▾
How should I treat volatility Google never confirmed?▾
Related Articles
Related Articles
How to Verify SEO Results in Search Console | Practical Improvement Cycle Guide
Learn how to read impressions, clicks, CTR, and position in Google Search Console to properly verify your SEO improvements. A practical guide to the improve → verify → discover cycle.
Don't Be Fooled by AI Search Stats | Examining the I/O 2026 Hype at Source
A principal-level report tracing the sensational numbers around AI search back to their sources — the causation caveat behind +35%, the real range of CTR drops, and the GSC anonymization trap — to build a judgment that isn't fooled by stats.
What Is GSC's Gen AI Performance Report? How to Measure AI Overview & AI Mode Impact
A guide to the Gen AI Performance Report added to Google Search Console in June 2026. Covers how to read AI Overview and AI Mode impression data, why clicks can't be measured, Pew Research's real-world click data, and our own GSC observations on AI search measurement demand.
What Is an SEO Penalty? How to Check Manual Actions, Recover & Prevent
An SEO penalty is a drop in search visibility caused by guideline violations. Learn the difference between manual actions and algorithmic drops, how to check in Google Search Console, common causes, the reconsideration-request recovery process, and prevention.
Verify Why You Grew — by Decomposition, Not Feel
Enter a URL for a 45-item audit covering technical, structured data, and E-E-A-T signals. Run it on both sides of a volatility window and compare.