Every SEO deck has a version of the same chart: a steep curve where position one takes roughly a third of clicks, position two takes half of that, and everything past position five rounds to noise. The chart is real. The problem is that it is an average of averages, drawn across millions of queries that behave nothing like yours.
Public CTR studies — the long-running datasets from Advanced Web Ranking, Sistrix, and Seer among them — have shown the same directional trend for several years: the curve is flattening at the top, not because ranking first got less valuable, but because "first" moved down the page. A result that ranks first organically can sit below an AI overview, a featured snippet, a local pack, four ads, and a People Also Ask block. Position one in the API is not position one on the screen.
The number you actually have
You do not need a study. You have your own curve sitting in Search Console, and it is the only one calibrated to your queries, your industry, and your brand recognition.
Pull the Performance report with Query and Position dimensions, export to a sheet, and bucket by rounded position. Compute clicks divided by impressions per bucket. That is your curve. It will not look like the study, and that is the point.
Two things almost always show up when businesses do this for the first time:
- Branded queries wreck the average. People searching your company name click your result at rates that can exceed 50%. Leave those in and every position bucket looks inflated. Filter them out and look at the non-brand curve separately — that is the one that tells you what new demand capture looks like.
- The curve is bimodal by intent. Informational queries where a snippet answers the question have a suppressed CTR at every position. Transactional queries where the searcher needs to reach a specific business have a much steeper curve. Averaging them together produces a number that describes neither.
Why this changes the forecast
The standard SEO business case multiplies estimated search volume by a borrowed CTR to project traffic. Both inputs carry error, and they compound in the same direction — volume tools tend to overstate, borrowed CTR curves tend to overstate — so the projection is optimistic twice over.
A more defensible version uses your own measured CTR by position bucket, applied to a volume estimate you have explicitly marked as approximate, and reported as a range rather than a point. "Moving this cluster from position 8 to position 3 is worth somewhere between 40 and 110 incremental sessions a month, based on our own CTR curve" is a claim you can defend in a quarterly review. "This will bring 300 visits" is a claim you will have to walk back.
What the flattening does not mean
It is tempting to read flat CTR curves as evidence that ranking does not matter anymore. The data does not support that. What flattened is the gap between the top few positions on feature-heavy SERPs. The gap between page one and page two did not flatten at all. Falling off the first screen is still the difference between meaningful traffic and effectively none.
The practical read is narrower than the headline: on query types where the SERP answers the question above the organic results, the marginal value of moving from three to one is smaller than it used to be, and the effort is often better spent winning a query where no feature intercepts the click.
How to tell those queries apart
Search Console will not label SERP features for you, but the shape of the data hints at them. Look for queries where you hold a strong average position and have healthy impressions but a CTR well below your own curve for that bucket. Those are the queries where something above you is absorbing the click. Sample twenty of them manually — an incognito search each — and you will usually find the same two or three feature types repeating.
From there the decision is straightforward. If a featured snippet is eating the click, restructuring the page to win the snippet is worth more than chasing a position you already hold. If an AI overview is answering the question outright, the query may simply be a poor target, and the same effort applied to a comparison or decision-stage query will return more.
The habit worth building
Recompute your CTR curve once a quarter. It moves — as your brand grows, as SERP layouts change, as your mix of queries shifts. A curve you measured eighteen months ago is a borrowed curve again, just borrowed from your past self.
The discipline this enforces matters more than the number. Once a team is looking at its own click data by intent type, the conversations change from "where do we rank" to "which of these rankings actually produce anything," which is the question worth answering.