Falling CTR isn't always bad content: how to tell the difference

If a page's click-through rate is sliding while its ranking holds steady, the content is rarely the first thing you should touch. Four things push CTR down without anything changing on the page itself: an AI answer intercepting the click, impressions growing faster than clicks, a Search Console reporting change, and a broad decline in click rates across all of search. Real content decay has a different signature. Clicks and impressions fall together, position slips, and competitors take the space you left. Read the signals in that order and you'll stop paying writers to fix pages that were never broken.

Why "the content got stale" is the default diagnosis, and why it's usually wrong

The reflex makes sense. CTR is the one metric in Search Console that looks like it's about your snippet, so a falling number reads as a failure of persuasion. Title's weak. Meta description's tired. Content must have decayed.

Then the data arrives. Ahrefs re-ran its AI Overviews study in February 2026 across 300,000 keywords, comparing aggregated Search Console CTR for December 2023 against December 2025. The presence of an AI Overview correlated with a 58% lower average CTR for the top-ranking page, up from the 34.5% the same team measured in April 2025. Pew Research Center's July 2025 analysis of 68,879 real searches from 900 US adults found people clicked a traditional result in 8% of visits when an AI summary was present, against 15% when it wasn't. Google called that methodology flawed; Pew stood by it. The panel data is now a year old, so treat it as a directional read on behavior rather than a current benchmark.

None of those declines were caused by anyone's title tag.

The four things that lower CTR without touching your content

An AI answer took the click

This is the one most teams already suspect, and it's real. But the diagnosis is more specific than "an AI Overview appeared."

Seer Interactive's April 2026 study tracked 53 brands, 5.47 million queries and 2.43 billion organic impressions from January 2025 through February 2026, and separated three states: no AI Overview, AI Overview present with the brand cited, AI Overview present with the brand not cited. Presence alone wasn't the mover. Citation status was. Being cited delivered about 120% more organic clicks per impression than not being cited on the same AI Overview SERP, though still roughly 38% below what a no-AIO SERP produced. Translated to volume, per million informational impressions Seer measured roughly 33,500 clicks with no AI Overview, about 20,700 when cited, and about 9,400 when an AI Overview appeared and the brand wasn't in it.

So "an AI Overview is on my query" and "an AI Overview is eating my clicks" are two different findings, and the gap between them is whether you're inside the answer.

Exposure also isn't evenly spread. In Seer's dataset, AI Overviews appeared on 36% of informational queries, 8% of commercial and 5% of transactional. By query shape the concentration is sharper: comparison queries triggered an AI Overview 95.4% of the time, review queries 86.3%, question-format queries 85.9%. If your declining page targets a "X vs Y" term, the SERP almost certainly has an AI answer on it. If it targets a transactional term, that's a much weaker hypothesis and you should look elsewhere first.

You won more visibility, and the math did the rest

This is the failure mode almost nobody checks, and it produces the most alarming-looking chart in the set.

CTR is clicks divided by impressions. If impressions grow and clicks hold flat, CTR collapses while absolutely nothing bad has happened. Seer's own data caught this in Q4 2025. Their month-over-month detection flagged October as the single largest CTR drop in the dataset for brand-cited informational queries, down 52.1% in one month, with another 36.8% in November. The raw numbers tell a different story: September logged 15.8 million brand-cited impressions and 398,798 clicks. October logged 33.1 million impressions and 400,271 clicks. Impressions more than doubled. Clicks went up by about 1,500.

A CTR chart falling off a cliff and a page earning citations on twice as many queries can be the same event. Seer are careful here and so should you be: they note they can't attribute that impression surge to Google's selection logic versus the brands' own SEO work without account-level investigation. The lesson is procedural rather than causal. Look at clicks and impressions as separate series before you interpret the ratio between them.

Ahrefs called the general version of this the great decoupling: impressions climbing while clicks flatten, producing a Search Console chart shaped like an open crocodile mouth.

Your measurement moved, not your performance

Two reporting changes in the last two years distorted CTR in opposite directions, and both sit inside the window most teams are still comparing against.

On 3 April 2026 Google disclosed a logging error that had over-reported Search Console impressions from 13 May 2025 until 27 April 2026. Clicks were unaffected. Because CTR uses impressions as the denominator, every CTR figure in that eleven-and-a-half-month window was understated. Google fixed the logging going forward and, per John Mueller, will not reconstruct the historical data. Impressions appear to fall and CTR appears to improve from late April 2026 onward, and neither movement reflects anything a user did.

Pointing the other way: Google stopped supporting the &num=100 parameter around 12 September 2025, which removed a large volume of scraper-generated impressions from deep positions. Impressions dropped, average position improved, clicks were largely untouched, and reported CTR went up.

Any CTR trendline that spans September 2025 or April 2026 is comparing two different measurement regimes, and the shape of the line is partly an artifact. Annotate both dates in your reporting before anyone reads a story into the curve.

There's a third quirk worth knowing. Search Console assigns position based on the topmost element a URL appears in, and every link inside an AI Overview reports as position 1. Get cited in an AI Overview and your average position can improve while your clicks fall. That combination looks like a snippet problem and isn't one. Mueller also confirmed in January 2026 that a URL appearing in both an AI Overview and a blue link on the same SERP counts once, not twice, so you can't explain away impression growth as double-counting either.

The floor moved under everyone

The most under-quoted number in the Ahrefs study is the control group. Position-one CTR for informational keywords without an AI Overview fell by roughly half between December 2023 and December 2025. AI Overview keywords fell about 78% over the same window. The 58% headline is what's left after Ahrefs corrects for that general drift.

Which means "an AI Overview intercepted my click" is itself sometimes an over-diagnosis. Part of what your page lost, it lost to a general contraction in clicking behaviour that shows up on SERPs with no AI answer at all. Seer offers a plausible read on the flip side: no-AIO informational CTR in their cohort rose from 2.93% in January 2025 to 3.97% in December, which they hypothesise is a selection effect. As AI Overviews absorb the low-engagement answer-seeking queries, what's left in the no-AIO pool skews toward people who intend to click.

What real content failure actually looks like

Clicks down. Impressions down too, or flat while the query's demand is steady. Average position slipping over weeks rather than bouncing. A live SERP check showing pages that weren't there six months ago now sitting above you. Google's own traffic-drop debugging guide is worth reading here, and its advice on small position shifts is the part people skip: if a page is already performing well, avoid radical changes.

The tell that separates decay from interception is impressions. Interception costs you clicks while leaving impressions intact. Decay costs you both.

The four-signal read

Pull position, impressions, clicks and AI-answer status for the affected query set, then match the pattern.

Two practical notes. Run this at query level, not page level: a single URL usually spans several of these patterns at once, and the page-level average hides which. And use a comparison window that doesn't straddle either reporting change, which as of mid-2026 means starting your baseline no earlier than May 2026 for impressions and CTR.

The manual check that takes five minutes

Search Console won't tell you whether an AI Overview fired on a specific query, so go look.

Take your top ten declining queries, run each in a clean browser session, and record three things: whether an AI answer appears, whether your domain is cited inside it, and how far down the page your blue link now sits. That gives you the citation-status split Seer's data says actually predicts CTR. If you're cited on most of them, your content is working well enough to be selected as a source and the click loss is structural. If an AI Overview fires and you're absent, you have a citation problem rather than a quality problem, and the fix is answer structure and entity clarity, not a longer post.

What Search Console can and can't tell you now

Google launched Search Generative AI performance reports on 3 June 2026, giving a dedicated view of impressions inside AI Overviews, AI Mode and generative AI features in Discover, broken out by page, country, device and date. Access widened later that month.

The limit matters for this diagnosis. The report shows impressions only: no clicks, no CTR, no query data, as Search Engine Land noted at launch. Google has said it will add metrics over time without committing to what or when. So you can now confirm that a page appears in AI features and roughly how often, which is genuinely new, but you still cannot compute what an AI appearance is worth or compare AI-feature CTR against blue-link CTR inside Google's own tooling. Anyone selling you a clean AI Overviews CTR number is modelling it, not reading it.

The test that actually settles it: what happens after the click

Every signal above describes the SERP. None of them describes your page. CTR is a measurement of the search results, not of your content, and the only honest way to test content quality is to look at the people who did arrive.

Compare the behaviour of this quarter's sessions against last year's on the same URL: scroll depth, time on page, navigation to a second page, conversion rate per session, return visits. Two outcomes, two different conclusions.

If the surviving sessions behave the same as they always did and there are simply fewer of them, the content is fine and you lost distribution. Rewriting it will change nothing except your budget. If the sessions that arrive now bounce faster, scroll less and convert worse than they did a year ago, that's a genuine content or intent problem, and it would be showing up whether or not an AI Overview existed.

That second test is the one most teams can't run, because click data lives in Search Console, behavioural data lives in analytics, and AI citation data lives in a spreadsheet someone updates manually. Joined per page, the three of them answer the question in an afternoon. Separately, they generate arguments.

When rewriting is the right call

Rewriting is the correct response when the page has lost position to specific competitors you can name, when search intent for the query has genuinely shifted format, when the page's facts have aged past usefulness, or when an AI Overview fires on the query and you're consistently not cited. That last one is a rewrite with a different brief: clearer direct answers near the top, self-contained sections, unambiguous entity naming, original evidence a summariser has a reason to attribute.

What doesn't justify a rewrite is a CTR line pointing down while clicks, impressions and position say nothing is wrong.

FAQ

My rankings are the same but my CTR dropped. Is my content bad? Probably not, on that evidence alone. Check impressions first. If impressions rose while clicks stayed flat, CTR fell for arithmetic reasons. If impressions held and clicks fell, something on the SERP is intercepting the click, most likely an AI answer.

Impressions are up and clicks are down. What does that mean? Usually that you're appearing on more queries, or in more SERP features, than you're converting into visits. Ahrefs named this pattern the great decoupling. It can also mean you've been cited in AI Overviews, which increases impressions without a proportional click return.

Does being cited in an AI Overview help or hurt? Both, relative to different baselines. Seer's 2026 data puts brand-cited CTR about 120% above uncited on the same AI Overview SERP, and about 38% below a SERP with no AI Overview at all. Citation is the best available position on an AI-answer SERP and not a return to pre-AI traffic.

Can Search Console show me AI Overview clicks? No. The Search Generative AI performance reports launched in June 2026 report impressions, pages, countries, devices and dates. Clicks, CTR and query data aren't included.

Why did my Search Console impressions change without any change in rankings? Two known reporting events. Google removed &num=100 support around 12 September 2025, cutting scraper-driven impressions. And a logging error over-reported impressions from 13 May 2025 to 27 April 2026, which Google fixed going forward without correcting the history.

How long should I wait before acting on a CTR drop? Long enough to distinguish a trend from noise, and long enough for the comparison window to sit inside one measurement regime. Four to eight weeks of query-level data is a reasonable floor, and don't compare across September 2025 or April 2026.

Where RankSage fits

Every diagnosis above needs three data sources in the same view: Search Console click and impression trends, AI citation status per query, and post-click behaviour per page. Most teams have all three and can't join them, which is why "is this content decay or zero-click interception?" turns into a meeting instead of an answer.

RankSage is being built to join them per page: GA4, Search Console, first-party behavioural data and AI citation data across ChatGPT, Claude, Gemini, Perplexity and Copilot, resolved into ranked actions rather than another dashboard. It isn't launched yet. If separating real content failure from click interception is a problem you're currently solving by hand, join the waitlist and we'll bring you into early access.