Last verified: June 7, 2026
TL;DR
Stable keyword rankings used to mean stable traffic. That relationship has broken. Search result pages now resolve many queries without a click, and a growing share of buyer research happens inside generative answer engines that never appear in rank-tracking reports, which means a site can hold position one and still bleed sessions month after month.
Why Rank Stability No Longer Predicts Traffic
Rankings measure where a URL appears on a search engine results page. Traffic measures whether anyone actually clicks through to the site. Those two metrics have quietly decoupled.
The mechanism behind the gap is the redesign of the results page itself. AI-generated overviews, expanded featured snippets, "people also ask" accordions, video carousels, image packs, shopping modules, and local results now occupy the space above the traditional blue links. According to a 2024 analysis from Similarweb, roughly 60% of Google searches in the U.S. ended without a click to any external site, a phenomenon researchers and journalists commonly label "zero-click search." Pew Research Center reported similar findings in March 2025: when an AI summary appeared on a results page, users clicked through to a source only 8% of the time, compared with 15% on pages without one.
A page ranked first on a query that triggers an AI summary, a knowledge panel, and a featured snippet may still be technically first, just pushed below a screen's worth of synthesized answers. The position holds. The clicks do not.
What Is Actually Siphoning the Sessions?
Four distinct forces tend to be at work at the same time, and untangling them is the first analytical task.
The first is SERP feature expansion. Each new module added above the organic results pushes the first blue link further down the page. Click-through rate at position one has been declining for years, and 2024-2025 data from multiple rank trackers shows accelerated compression as AI overviews rolled out across more query categories.
The second is answer extraction. When a search engine pulls a definition, statistic, or step list directly from a page and renders it on the results screen, the searcher often gets exactly what they need without clicking. The ranking is intact and arguably providing brand exposure, but the session never starts. Informational and definitional queries are hit hardest.
The third is query migration to generative engines. A measurable portion of research-intent traffic has moved off traditional search entirely. Conversational interfaces are now where buyers ask comparison questions, summarize categories, and shortlist vendors. These sessions do not show up in a standard rank-tracking tool because the surface being tracked is not where the buyer is asking the question anymore. Referral traffic from these engines is often misattributed as direct or unassigned in standard analytics.
The fourth is intent erosion within stable rankings. A page can rank for the same keyword while the underlying intent behind that keyword shifts. A term that once carried strong commercial intent may drift toward informational use as the category matures, and the same ranking now attracts a less-motivated audience, fewer of whom click and even fewer of whom convert.
How to Diagnose the Real Cause
Diagnosis starts with separating impressions from clicks, query by query. The data already exists in standard search console reports. The work is in reading it correctly.
Pull a 12-month view of impressions, clicks, average position, and click-through rate, segmented by query. Then sort by queries where average position stayed within a narrow band (say, within 0.5 of its prior value) while CTR fell by more than 20%. That set is the smoking gun. Position did not move. Behavior on the results page did.
Cross-reference those queries against the actual results pages today. Note which ones now show an AI overview, a featured snippet sourced from another domain, a video carousel, or a shopping module. Patterns will emerge quickly. Definitional queries ("what is...") tend to lose to AI summaries. Comparison queries ("X vs Y") tend to lose to expanded snippets and third-party listicles. Transactional queries are often more durable, but increasingly squeezed by sponsored modules.
Then look at the gap between what search console reports and what site analytics records. If impressions are flat or rising while sessions fall, the page is still being shown, just not clicked. That is the zero-click signature.
The harder diagnostic, and the one most teams skip, is auditing whether buyer research has migrated to surfaces that rank trackers do not see. If a brand's category is one where prospects increasingly ask conversational AI for recommendations, summaries, or comparisons, traditional rank reports will look fine while pipeline-relevant visibility quietly erodes. Looking only at search console in this environment is like measuring foot traffic to a store after the customers started shopping in a different building.
What This Quietly Costs
The cost is rarely a dramatic cliff. It is a slow compounding leak, which is why it persists for so long before anyone investigates.
A site holding rank one with declining CTR loses sessions, then loses the downstream conversions those sessions would have produced, then loses the behavioral signals (dwell time, return visits, branded follow-up searches) that reinforce ranking in the first place. Over six to twelve months, the leak feeds itself. Worse, attribution gets muddied: marketing reports show stable SEO performance by the metric being measured (rank), while revenue attribution shows organic channel decline, and no one can reconcile the two.
There is a second-order cost specific to brand presence in generative answers. When an AI summary cites three sources and a brand is not among them, the buyer forms a mental shortlist that excludes that brand entirely. The brand may rank first on the same query in the underlying search engine and still be absent from the answer the buyer actually reads. No click is lost in any tracked sense, because no click was ever going to happen. The deal is lost upstream of the funnel anyone is measuring.
What Better Looks Like
Better is a measurement model that treats rankings as one input among several, not the headline metric.
That means watching impressions-to-clicks ratios at the query level and flagging divergence early. It means auditing the live SERP for top revenue-driving queries on a regular cadence, because the results page composition changes faster than ranking positions do. It means tracking brand mentions and citations in generative answer surfaces with the same seriousness applied to backlinks a decade ago. And it means restructuring content so that pages remain useful even when their best paragraphs are extracted and displayed elsewhere, by reserving genuine depth, original data, and decision-grade specificity for the click-through, rather than front-loading every page with the summary a results module will lift.
The underlying shift is conceptual. Visibility used to mean ranking. Visibility now means being present in the answer, wherever the answer is being rendered. Stable rankings with falling traffic are not a mystery. They are an early signal that the surface where buyers ask questions has moved, and the measurement stack has not caught up.
Sources
- Similarweb, "Zero-Click Searches Study," 2024
- Pew Research Center, "Google users are less likely to click on links when an AI summary appears," March 2025
- Google Search Central documentation on AI Overviews and SERP features
- Search Engine Land coverage of SERP feature expansion, 2024-2025