Between May and August 2026, a UK employee wellbeing platform's commercial pages did exactly what an SEO engagement is supposed to make them do. The buyer-guide article on the head query went from position 15.7 to 5.9. The category variant reached 4.8. The corporate-wellness article reached 8.2. Page 2 and 3 became page 1, and impressions on the commercial cohort rose.
Clicks from those queries: approximately zero.
Not disappointing. Not below benchmark. Zero-shaped. In the 90-day window, the site's non-brand results at positions 5.5 to 10.5 collected 128,716 impressions and 20 clicks - a click-through rate of 0.02%. Even results measured at position 1 to 1.5 on their queries took 1,036 impressions and converted them into not one click.
Any CTR model you have ever used says page 1 pays. This site reached page 1 and it did not pay. This case study is about why, what "why" implies about where the next unit of effort should go, and the reporting discipline that kept a zero from being spun into anything else.
§ The autopsy of a SERP
The answer was sitting one query away, in the live results themselves. I pulled the actual UK SERPs for four commercial head terms. All four had the same shape: an AI Overview on top, ads, a People Also Ask block, product panels, then - somewhere below all of it - the organic results the client had spent three months climbing into.
Google now answers the commercial question above the results. The searcher who asked "best employee wellbeing platform UK" meets a synthesized answer, a row of ads, and a set of expandable follow-up questions before the first blue link. Position 6 in 2026 is not what position 6 was when the CTR curves in everyone's opportunity models were fitted. On these SERPs, for this site, the entire click volume that models promise to page 1 was being absorbed before the organic list began.
That is the finding I could not have reached from inside Search Console, and it reframed the whole engagement: ranking was no longer the binding constraint. The site had solved ranking. It was now losing a different competition - the one for the SERP surface itself.
§ Own-curve or no curve
The methodological point of this piece: when I calculated the remaining opportunity, I refused to use industry CTR bands, because the site's own measured curve falsified them. A generic model says position 5 earns some healthy single-digit CTR; this site's measured reality at those positions was 0.02%. Projecting clicks from the generic band would have manufactured an opportunity number two orders of magnitude too high, and every downstream promise built on it would have been fiction.
So the projection ran on the site's own curve, with a stated assumption: roughly 19,000 attributed non-brand impressions a month, a target CTR of 1 to 3% explicitly labeled as an assumption rather than a benchmark, and an honest near-term prize measured in tens of clicks per month, growing with authority. Nobody frames an engagement around "tens of clicks" because it sounds small. I do it because it survives contact with the client's own data, and because the alternative - a number the first month of reality disproves - costs the relationship more than the small number ever will.
§ Where the effort goes when ranking is solved
The diagnosis splits the problem into three pieces with three different owners, which is the practical payoff.
The AI-and-answer-surface layer - the Overview that owns the top of every commercial SERP, and the answer engines beyond Google - is where the searcher's attention actually went, so that is where visibility work now aims: being the source the answer cites, rather than the link below the answer. (On that front this client already had its first win; that story is case study #35.) The snippet layer - titles and metas that make an organic result worth clicking for the minority who scroll past the answer stack - was full of generic titles and missing descriptions, all fixable in-house. The authority layer - the links and earned mentions that decide who gets cited everywhere - is the long game. Three layers, three owners, no pretending one more rankings push would change the click column.
§ How to test whether your constraint has moved
- Build your site's own CTR-by-position curve from Search Console data, non-brand only. If it sits far below every benchmark at every position, stop using benchmarks - your curve is the finding.
- Pull the live SERPs for your top commercial queries and inventory what sits above the first organic result: AI Overview, ads, PAA, panels. That stack is your real competitor now.
- Separate the three layers before assigning work: answer-surface capture, snippet quality, authority. They fail independently and are fixed by different people.
- Never let a rankings win be reported as a traffic win until the click column agrees. Position is an input. It stopped being the output years before the tooling admitted it.
Numbers dated as of the August 2026 audit: positions and CTR from Google Search Console over a 90-day window, SERP composition from live UK pulls on four commercial queries, both sourced in the client deliverable. The site is anonymized as a matter of client confidentiality.
The wider claim, and I mean it precisely: for commercial queries in categories like this one, ranking has become a qualifying round. It gets you into the arena where answers are assembled. The prize moved above the results - and the work that wins it is not more ranking.