A small but pointed study has put a number on something practitioners have been trying to quantify for months. The click-through rate from Google AI Overviews and AI Mode is approximately 0.9%. That figure will prompt two very different reactions depending on where you sit.
Some will read it as vindication for every sceptic who has argued that AI search features cannibalise organic clicks without returning value. Others will recognise that 0.9% is not nothing - and that the more uncomfortable finding from the same study is what happens when you try to avoid AI features altogether.
What 0.9% Actually Means in Practice
Before treating 0.9% as a settled verdict on AI search value, it helps to put it in context. This is a small study. It reflects a snapshot of behaviour across a limited sample, not a statistically robust benchmark across all query types, sectors, and geographies. The number is directionally useful. It is not a fixed law.
That said, 0.9% is low by any reasonable comparison to traditional organic click-through rates, which vary widely but routinely exceed 2-5% for the top few positions in standard results. The question is not whether 0.9% is impressive. It is whether that click volume adds up to something meaningful at scale, and whether the citation and recommendation value that comes with AI feature appearances has worth that does not show up in a CTR figure.
Visibility in AI Overviews can shape how a brand is described, summarised, and recommended - even when users do not click. If your brand is being cited as an authority or a recommended provider, that influence on perception exists regardless of whether the user visits your site from that specific surface. CTR measures one outcome. It does not measure all outcomes.
The Opt-Out Finding Is the Real Story
The study's most significant finding is not the 0.9% headline. It is this: when sites opted out of appearing in Google's generative AI search features, they also saw a drop in clicks from standard web results. That is a material observation for anyone who has been considering avoidance as a strategy.
The mechanism is not confirmed, but the implication is clear enough to take seriously. Google's AI systems and its traditional ranking systems are not operating in entirely separate silos. Signals that feed one may influence the other. A site that withdraws from AI feature eligibility may not simply stay neutral in traditional results - it may lose ground there too.
This matters because opting out has been discussed in some quarters as a reasonable hedge - a way to protect traffic from zero-click AI summaries while maintaining standard rankings. The data from this study challenges that assumption directly. The trade-off may not be as clean as it appears.
Why Withdrawal Is Unlikely to Be the Right Move
There is a legitimate frustration behind the opt-out instinct. If AI features summarise your content and users get what they need without clicking, the value exchange feels one-sided. You produce the content. Google surfaces it. You see less traffic. It is a reasonable grievance.
But withdrawing from a system that may account for a growing share of how users form buying intent - and which may now influence traditional result performance too - creates a different kind of risk. You are not stepping out of a side feature. You are potentially stepping back from where evaluation decisions are increasingly being shaped.
The more productive response is to treat AI feature appearances as a visibility layer that requires its own measurement discipline. Track what you can see. Build a baseline. Understand what types of queries trigger AI Overviews in your sector, and whether your content appears in them. That is more useful than opting out and hoping the problem goes away.
Measuring What You Can Actually Measure
Google Search Console now surfaces some AI Overview data, including impressions and clicks associated with AI features. The data is limited, but it gives practitioners a starting point. If you are not already segmenting AI Overview traffic separately in your reporting, that is the first practical step.
Beyond GSC, AI-referred traffic from surfaces like AI Mode requires careful UTM discipline and source attribution work in GA4. Referral paths from AI features do not always behave like standard organic referrals. Sessions can appear as direct traffic or with ambiguous source tagging depending on how the user navigates from an AI-generated response to your site.
Connecting this through to CRM data is where the attribution picture gets genuinely useful. If AI-referred sessions convert at a different rate to standard organic sessions - higher or lower - that changes how you weight the channel. A 0.9% CTR on a query type that converts at 8% is a different proposition to a 5% CTR on a query type that converts at 0.5%. Volume alone is not the right lens.
What Practitioners Should Actually Do With This Data
First, do not make opt-out decisions based on CTR data alone. The study suggests there are second-order consequences that are not captured by the AI feature click figure itself. Any site considering withdrawal from AI Overviews should model the potential impact on standard results performance before acting.
Second, treat 0.9% as a provisional benchmark rather than a fixed rate. The figure is likely to vary considerably by sector, query intent, and how competitive the AI Overview slot is in a given category. Run your own analysis against GSC data before drawing conclusions about your specific situation.
Third, broaden what you count as a return from AI visibility. Citation quality, brand description accuracy, recommendation frequency in multi-source queries - these are meaningful outputs that do not appear in CTR data. Building a measurement framework that captures them alongside click and conversion data gives a more honest picture of what AI search appearances are actually worth.
The Bigger Picture on AI Search ROI
Studies like this one are valuable precisely because they inject some data into a conversation that has been dominated by speculation. 0.9% is a real number from a real observation, and it deserves to be taken seriously - not as a reason to abandon AI search investment, but as a calibration point for realistic expectations.
AI search features are not a replacement for well-structured organic programmes. They are an additional visibility layer with their own mechanics, their own measurement challenges, and - as this study suggests - their own dependencies on the signals that also drive traditional search performance. The opt-out finding reinforces something practitioners should already suspect: these systems are increasingly integrated, not parallel.
The sensible position is to stay visible, measure rigorously, and build the attribution infrastructure that lets you connect AI appearances to commercial outcomes. That is more defensible than either ignoring the channel or withdrawing from it based on a single CTR figure.