Check this page with an assistantOpens a chat asking it to summarise this article and name the evidence behind each claim.
Claude opens with the prompt on your clipboard: Anthropic does not support prefilled prompts on the web, and we would rather copy it than ship a button that drops it.
One Description, Three Different Problems
Traffic is down and rankings look unchanged. That sentence covers three distinct situations with opposite remedies, and the summary chart most people look at cannot tell them apart because it shows clicks and nothing else.
The three are ranking loss, on queries you were not tracking; demand decline, where the searches themselves reduced; and click-through erosion, where you appear as often as before and get clicked less. Only the first is a visibility problem. Only the third is affected by how your result looks. Neither of the last two is fixed by rewriting the page.
Getting this wrong is expensive in a specific way: the default response to any decline is to rewrite content, and that is the correct response to roughly one of the three. The other two absorb a quarter of work and produce nothing, after which people conclude that SEO stopped working.
You can name which of the three you have, on which pages, with the numbers that show it.
Separating Them in Search Console
Read impressions, average position, and click-through rate together over matched periods. Each cause has a distinct signature across those three, and any one of them alone is ambiguous.
| Impressions | Position | CTR | Diagnosis |
|---|---|---|---|
| Down | Down | Flat | Ranking loss |
| Down | Flat | Flat | Demand decline, or queries resolving elsewhere |
| Flat or up | Flat | Down | Click-through erosion |
| Up | Down | Down | Query mix shifted; you are appearing for different things |
| Flat | Flat | Flat | Measurement change; check consent and tagging |
The last row catches a case people rarely consider. If Search Console shows no change and your analytics shows a decline, traffic did not fall, measurement did, and the investigation belongs in your tagging rather than in your content. That check costs two minutes and saves entire projects.
The Part You Cannot Know
Click-through erosion is real and measurable. Its cause is not. No search engine reports clicks for AI surfaces, and Google's generative AI report shows impressions without click data, which means "AI Overviews took your clicks" cannot be verified from any available source.
That is worth stating plainly because the claim is now the default explanation for any CTR decline, and it is being sold against. Someone quoting a percentage of clicks lost to AI Overviews on your site is quoting a number that does not exist in your data or anyone else's.
The honest position is that you have measured erosion and can list the plausible causes without ranking them. That is less satisfying than an explanation and more defensible than one, and it is the construction described in the evidence ladder: never promote an observation into a cause without evidence for the bridge.
The Confounder List
Before attributing erosion to anything, rule these out. Each produces the same signature, several are more common than the popular explanation, and all are checkable.
- Query mix shift.
Aggregate CTR falls when the proportion of informational queries in your mix rises, without any individual query changing. Check CTR on a fixed set of queries rather than in aggregate.
- Branded versus non-branded mix.
Branded queries convert to clicks at far higher rates. A shift in the ratio moves blended CTR on its own, which is why the two should be reported separately, per branded search.
- Results page layout changes.
More ads, a bigger map pack, a shopping module, or additional features push organic results down without changing your position number. Position is a rank, not a location on screen.
- Your own result changed.
A rewritten title, a lost rich result, or a changed description alters CTR directly. FAQ rich results disappearing in May 2026 did exactly this to a lot of pages, per structured data in 2026.
- Device and country mix.
Mobile CTR differs from desktop, and markets differ from each other. A change in the mix moves the blended number with nothing happening to any segment.
- Seasonality.
Compare year over year before concluding anything, per seasonal SEO. A large share of alarming charts are the same shape as last year.
What Each Cause Actually Needs
The point of the diagnosis is that the three have different remedies, and applying the wrong one is worse than doing nothing because it consumes the quarter.
Ranking loss is the case where content work applies. Identify which pages and queries lost position, look for the shared trait across them, and decide whether they genuinely deserve to win. That is the sequence in page-level authority and core update recovery.
Demand decline cannot be fixed on the page. Fewer people are searching, and the response is either to accept a smaller channel or to find where that demand went, which is frequently a different query formulation rather than an absence. Click-through erosion is a results-appearance problem, and the levers are titles, descriptions, and whether your result answers the query more compellingly than the ones around it, per SERP appearance.
Reporting It Without Overclaiming
The finding is the cause plus its confidence. "Click-through erosion on informational queries, cause not attributable, these five confounders checked" is a stronger report than a confident single explanation, because it survives being questioned.
It also prevents a specific commercial failure. A programme that attributes a decline to AI Overviews has told the client the problem is unfixable and external, which is convenient in the short term and removes any basis for the next invoice. Naming what you can act on and what you cannot is better for both parties.
Keep the analysis reproducible: same periods, same segments, same method each time, recorded alongside your change log. A decline that starts on a deploy date is an engineering incident, and you will only notice that if the dates are written down, which is the argument in SEO split testing.
Questions About Traffic Declines
- My rankings are the same but traffic is down. What happened?
Three different things produce that description and they need opposite responses. Ranking loss on queries you were not watching, demand decline where fewer people are searching, and click-through erosion where you are shown as often as before and clicked less. Search Console can separate all three, and the summary chart everyone looks at cannot.
- Did AI Overviews take my clicks?
Possibly, and you cannot verify it. No search engine reports clicks for AI surfaces, and Google's generative AI report shows impressions without click data. That means the popular explanation for click-through erosion is a hypothesis rather than a measurement, and anyone stating it as fact is stating something the data does not contain.
- How do I tell demand decline from a ranking problem?
Look at impressions and average position together. Impressions falling while position holds points at fewer people searching. Position falling while impressions hold points at losing visibility on queries that are still being searched. The two look identical in a clicks-only view, which is why clicks-only reporting causes so much misdiagnosis.
- Should I rewrite the page?
Not until you know which cause you have. A rewrite is the correct response to roughly one of the three, and the most expensive wrong answer to the other two. Demand decline cannot be rewritten away, and click-through erosion caused by a changed results layout is not a content problem.
- How far back should I compare?
Year over year for anything seasonal, and matched periods of equal length on the same weekdays for everything else. Comparing an arbitrary month against another month imports seasonality as though it were an effect, which is the single most common error in this analysis.
