The most reliable, physically observable mechanism is scroll depth, not a hidden algorithmic change to how each position is scored. When an AI Overview renders above the traditional organic results, it occupies vertical space that the first organic result used to occupy. Position 1 is still labeled position 1, but a user now has to scroll past the Overview to reach it. Everything below shifts down by the same amount. This is the same “pixel depth” mechanism that has driven click-through behavior for over a decade, well before AI Overviews existed, applied to a new and often larger SERP feature.
The established mechanism: pixel depth, not position number
Pre-AI SEO research has long documented that click-through rate correlates more tightly with actual pixel position on the rendered page than with the nominal organic rank number. A position 1 result pushed far down the page by image carousels, ad blocks, featured snippets, or knowledge panels has historically shown lower CTR than a “true” position 1 on a sparse SERP, because users interact with what’s visually in front of them, not with an invisible rank number. AI Overviews extend this same principle: they are frequently a large, multi-paragraph block, sometimes with expandable sections and source links, which can push the entire ten-blue-links block substantially further down the viewport than any single prior SERP feature typically did.
Mechanistically, then, the effect on CTR distribution across positions one through ten is best understood as a magnitude increase on an already-known phenomenon: more scroll distance to reach any given organic position generally means fewer users reach it before either being satisfied by the Overview or abandoning the search. This compresses the CTR curve overall (lower CTR at every position relative to a no-Overview SERP) rather than dramatically reshuffling which position gets relatively more or less of the remaining clicks.
What is not established, and where fabrication risk is highest
Any specific claim of the form “position 1 now gets X% instead of Y%” should be treated with real caution. Google has not published a redistributed CTR-by-position table for AI Overview SERPs, and while several third-party SEO tools and industry studies (from vendors like Semrush, Ahrefs, and similar rank-tracking platforms) have published their own observed CTR estimates on AI Overview SERPs, these are studies of a specific sample of tracked keywords and clients, not a universal, Google-confirmed figure. Numbers vary meaningfully between these studies depending on methodology, query set, and time period sampled. If you encounter or want to cite a specific redistribution percentage, treat it as a third-party estimate from a named source and label it as such rather than presenting it as settled fact about how Google’s systems behave.
It’s also worth being precise about what “mechanistically” can honestly mean here. There is no evidence that Google’s ranking algorithm treats a given URL differently in terms of relevance scoring because an AI Overview is present on that SERP. The organic ranking underneath the Overview is still produced by the same underlying ranking systems. What changes is the display layout the user encounters, and layout-driven CTR compression is a well-established, non-speculative phenomenon; it does not require assuming any change to the ranking computation itself.
How this compares to prior SERP-feature displacement, and where it differs in degree
Featured snippets, “People Also Ask” accordions, and image or video carousels have all previously pushed organic position 1 further down the page than its nominal rank number implied, and each of those features generated the same basic scroll-depth compression when introduced. What differs with AI Overviews is largely magnitude and frequency of occurrence rather than a new type of mechanism. A featured snippet historically occupied a relatively compact, single-answer block; an AI Overview is frequently a substantially larger, multi-paragraph block, sometimes with expandable sections, source chips, and follow-up prompts, occupying more vertical space than most prior single-purpose SERP features did individually. It’s also appearing across a broader range of query types than featured snippets historically covered, since Google has expanded AI Overview coverage well beyond the narrow, clearly-factual query set that snippets were mostly confined to. Both of these factors, larger size per instance and broader query coverage, compound the same underlying scroll-depth mechanism rather than introducing a mechanistically different one, which is why practitioners who lived through the featured-snippet-driven CTR compression of the mid-2010s are recognizing a familiar pattern here, just operating at larger scale.
A common measurement mistake this creates
A frequent analytical error is comparing current CTR at a given position against a benchmark curve built years ago, before AI Overviews existed at meaningful scale, and concluding that “CTR has collapsed” or that “Google changed the algorithm” for a specific position. This conflates two different things: the ranking computation for that URL likely hasn’t changed at all, but the SERP layout the user encounters has. Attributing the CTR change to a ranking-quality signal (and responding by chasing content or authority improvements) when the actual driver is a layout-and-scroll-depth effect wastes effort on the wrong fix. The correct response to a CTR drop concentrated on AI-Overview-present queries, with stable underlying rank, is treating it as a display and demand-capture problem, not a relevance or quality problem to be solved with the traditional ranking-improvement playbook.
A hypothetical illustration
Consider a hypothetical example: a site called Foxglove Garden Supply ranks position 1 for “how often to water succulents” and has a historical CTR baseline of 28% at that position, built from years of pre-AI-Overview data. Suppose an AI Overview begins appearing for that query, pushing the ten-blue-links block, including Foxglove’s position 1 result, several hundred pixels further down the viewport. If Foxglove’s team then observes CTR at that same nominal position 1 drop to 14%, hypothetically, and mistakenly concludes Google penalized their page or that content quality declined, they’d be chasing the wrong fix, rewriting content or building links, when the actual driver is scroll-depth compression from the Overview’s vertical footprint, with the underlying ranking computation for that URL likely unchanged. The correct read in this hypothetical is to segment CTR analysis specifically for queries where an Overview is confirmed present, compare that segment against Foxglove’s non-Overview queries, and treat the gap as a display and demand-capture effect rather than a relevance problem.
Practical implication
For measurement and reporting purposes, the useful move is to stop expecting CTR-by-position benchmarks calibrated on pre-AI-Overview SERPs to still hold for queries where an Overview is present. If your historical CTR curve was built from a mix of Overview and non-Overview SERPs, blending it going forward will understate the compression effect on Overview-heavy queries and overstate it on Overview-free ones. Segment CTR analysis by whether an AI Overview is present for each query when doing forecasting or diagnosing a decline, rather than applying one blanket curve across all query types. This segmentation, grounded in the observable scroll-depth mechanism, is a more defensible diagnostic approach than trying to reverse-engineer a universal redistribution percentage that no single authoritative source has actually published.
