What 16 months of Search Console actually shows.

The dataset & method

The corpus is five businesses PYC runs organic search for: a signage e-commerce store, a prestige-marque automotive specialist, a domestic glazing repairer, a regional auto locksmith, and a B2B compliance-software brand. The window is 16 months of Search Console (March 2025 – June 2026) plus lifetime GA4 for conversions. In total, 50,391 organic clicks across 3.34M impressions.

Every figure is exported from the client's own consoles and sanitised by identity only — names, domains and locations removed, no metric altered. This is a field study, not a controlled trial: n = 5, one operator's book. Read it as a strong direction, not a law. The full per-month data is in the appendix.

Definitions

Six terms this study leans on, defined so nothing rides on jargon.

Topical authority

The trust a search engine assigns a site once it has covered a subject completely and long enough to be a safe bet to rank for it. Built from coverage plus historical data, not from any single page.

Non-brand query

A search that does not contain the company name — category demand, not brand navigation. The traffic that grows a business rather than the traffic it already had.

Foundation phase

The early period where coverage and history accrue with little visible return, before the curve breaks upward. In this data it ran roughly 12 months.

Average position

The mean ranking of a site’s queries in Search Console. Lower is better; a move from 46 to 24 is page 5 to page 2.

CTR (click-through rate)

Clicks ÷ impressions. Low CTR at high impressions means a site is being shown but ranked too low to be clicked — visibility without the click.

Core update

A broad, confirmed Google ranking change. Dates are public (Search Status Dashboard); their effect on any one site is inferred, not published.

The curve is real — and it is slow

All five accounts, charted against the confirmed Google updates — the two that fell included, so nothing is hidden. In the compounders, a long flat foundation precedes the breakout: coverage and history accrue for months with little to show, then the line hooks upward and peaks at the end. Hover any point for its exact month and value.

Clicks decouple from value

This is the finding that matters most, because it is the one that gets programmes cancelled. Two of the five accounts saw organic clicks fall over the window. Both are arguably the healthiest in the book. Put click direction next to average position and the business outcome, and the story inverts:

Every account improved average position — the two whose clicks fell included. The declines were an early, high-volume guide maturing and the query mix shifting to commercial terms; the accounts kept ranking better and kept producing customers.

Visibility without the click — “the great decoupling”

The decoupling in Finding 2 is not only an ageing guide. It is the signature of the AI-results era. Since Google’s AI Overviews became prominent through 2025 and into 2026, practitioners have named a pattern now visible across Search Console: impressions hold or rise while clicks fall — the AI answer resolves the query without a click. The industry calls it the great decoupling.

The scale, from independent studies (sourced below): organic CTR on AI-Overview queries fell roughly 61% (Seer, Sept 2025); Ahrefs measured a 58% lower CTR on the top-ranking page when an AI Overview is present, and about a 34.5% average drop across 300,000 searches. Early 2026 shows a partial rebound (CTR 1.3% → 2.4%, Dec 2025 → Feb 2026) — but the structural shift stands: the same query now earns the impression and not the click.

What the five accounts actually show

With the real monthly impressions now in hand — pulled from each property's Search Console over the same window — the decoupling turns out to be selective: present in one account, absent in the rest. That precision matters more than a blanket claim. First quarter of the window versus the last:

All five accounts, real Search Console impressions and clicks over the 16-month window.

The decoupling, drawn

One account shows it cleanly. Indexed to the opening quarter, impressions climb while clicks stay flat — the widening gap is the decoupling:

Impressions grew +230% while clicks moved just +18%; CTR fell from 1.38% to 0.50%. This is a B2B brand ranking for competitive, informational head terms — precisely the queries an AI answer resolves without a click. Visibility was built; the click did not follow.

What click-through rate exposes

Divide clicks by impressions and the in-progress account stops needing an excuse and starts having a diagnosis. Four of five accounts blended to a 1.10% CTR. One sits at little more than half that:

The compliance-SaaS account has plenty of impressions (771K) but a 0.57% CTR — it is being shown and not clicked, because its head-term positions (36 → 24) are still on page two or three, below the click-earning zone. That is not a failing programme; it is a foundation that has built visibility and not yet earned the position to convert it. CTR turns "trust me, it's working" into a number.

Non-brand is the prize, and it converts above its weight

Where the split was measured, the overwhelming majority of clicks were non-brand — category demand, not people typing the company name. And that demand converted: for the signage store, organic was 28% of sessions but 57% of orders.

Time is the input, retention is the multiplier

The inflection took roughly a year: the clearest curve stayed flat for ~12 months before it broke. That is the input the model demands — coverage and history, paid down before the return. And the return compounds only if you stay: the oldest account is five-plus years on one retainer, organic now around two-thirds of all traffic. Long enough that a churning vendor stack would have reset two or three times, losing the accumulated history each time.

The mechanism: why position leads and clicks lag

Put the five accounts together and a causal order falls out. Coverage produces entity-complete pages. As history accrues, the search engine's trust — topical authority — grows, and it shows up first as improved average position: the engine becomes willing to rank you before anyone clicks. Once position crosses into the click-earning zone (roughly page one to two), that trust converts into clicks. And clicks convert into customers only when the queries are commercial and non-brand.

So the chain is: coverage → position → clicks → non-brand conversions. Position is the leading indicator, clicks are the lagging one, and non-brand conversion is the outcome. An account can be winning on the leading indicator and the outcome while its headline clicks fall — because an early viral spike on an informational query inflated the click baseline that later normalised. That is exactly the two "declining" accounts. Read the leading indicator and the outcome; forgive the noisy middle.

Model it yourself

The thesis, made playable. Set the starting traffic, how long the foundation phase runs, and the monthly growth once it breaks — the curve, the time to 5×, and the 24-month total update live. Notice what actually drives the outcome: the foundation length and the growth rate, not where you start.

Do the numbers hold up?

Claims are cheap, so here the data is tested rather than asserted. The sample is small, which puts cross-account statistics off the table — but each account's 16-point monthly series is long enough for a per-account trend test and a curve fit.

A Mann-Kendall trend test asks whether each series really trends or just wanders. It confirms the split — and, importantly, that the two falling accounts are falling for real, not by chance:

Then, is "compounding" literally true — multiplicative, not just up-and-to-the-right? Fit the breakout window to a line and to an exponential and compare R². Glazing's breakout is exponential at R² = 0.988 (≈ ~68% month-on-month) — genuinely compounding. Signage, by contrast, fits an exponential at only R² = 0.37: it is a step-change to a new plateau, not a curve. Reporting that difference, rather than calling both a hockey stick, is the point.

Before / after each core update

Average monthly clicks in the two months before vs the two months after — the correlations, quantified (clicks, not position). The nuance is the whole point:

Limitations

What this study is not, stated plainly so you can weight it correctly.

1. 01 n = 5, drawn from one operator’s book. This is a field study, not a controlled trial — read it as a direction, not a law.
2. 02 Selection bias: these are engaged, funded retainers, not a random sample of businesses doing SEO.
3. 03 The update dates are sourced and confirmed; the correlations to each account are the author’s inference, labelled correlation not cause. A GSC before/after around each date could confirm or retract any one of them.
4. 04 Non-brand share was measured for one account (93%); it is not generalised to the portfolio. “Organic share” figures (68%, 77%) are share of traffic or customers — a different metric.
5. 05 Impressions were available for four of five accounts, so the blended CTR excludes the auto-locksmith.
6. 06 Sanitised by identity only. No metric was rounded to flatter or altered to fit; where a number moved the wrong way it is shown moving the wrong way.

References

The algorithm context is sourced to public records — not commentary on these anonymous accounts, which does not exist. Verified July 2026.

1. 01 Google Search Status Dashboard ↗ Google’s own authoritative record of ranking/core/spam update rollouts and dates.
2. 02 Search Engine Land — Google algorithm updates (library) ↗ Running, dated history of confirmed Google updates.
3. 03 Search Engine Land — 2025 in review: 3 core updates + 1 spam update ↗ Confirms the 2025 core-update and spam-update dates used here.
4. 04 Search Engine Roundtable — 2025 Google algorithm updates ↗ Practitioner tracking of confirmed and unconfirmed volatility across 2025.
5. 05 Search Engine Land — May 2026 core update rollout complete ↗ Confirms the May 2026 core-update window (May 21 – Jun 2, 2026).
6. 06 Search Engine Roundtable — May 2026 core update done ↗ Corroborates the May 2026 rollout and its felt impact.

AI-Overviews era

1. 07 Seer Interactive — AI Overviews impact on Google CTR (Sept 2025) ↗ Organic CTR on AI-Overview queries fell ~61% (1.76% → 0.61%).
2. 08 Ahrefs — AI Overviews reduce clicks ↗ ~58% lower CTR on the top-ranking page when an AI Overview is present; ~34.5% average drop across 300k searches.
3. 09 Search Engine Land — AI Overviews CTR shows early recovery ↗ CTR on AI-Overview queries rebounded 1.3% → 2.4% (Dec 2025 → Feb 2026).
4. 10 Honcho — why impressions are up but clicks are down ↗ Documents the “great decoupling” pattern directly in Search Console data.

Appendix · the full data

Organic clicks per month, March 2025 – June 2026, straight from each account's Search Console. No smoothing, no cherry-picked window — the same numbers the charts are drawn from.