Google is placing links at the bottom of AI Overviews that look like ordinary web links but behave differently. Instead of taking a user to an external website, they prompt a follow-up question that routes the user into AI Mode. The destination is not your page. It is another AI-generated response.
This is not a minor UX detail. It is a structural change to the AI Overview format that affects how - and whether - cited brands receive traffic. If your site is referenced in an AI Overview but the visible links send users deeper into Google's own environment, the citation you worked to earn has become decorative.
What Is Actually Happening in the SERP
AI Overviews already compress the user journey. A searcher asks a question, gets a synthesised answer, and has less reason to click through to source material. The citation links that do appear - typically small numbered references or source panels - have always generated modest click volume compared to traditional organic results. That was already a concern for AEO practitioners.
What this latest change introduces is a second category of link: one that appears at the bottom of the Overview, looks like a normal hyperlink, but instead of resolving to an external URL, fires a follow-up query inside AI Mode. The user stays within Google. The conversation continues. Your website is not in that conversation.
The practical effect is that a user who might have clicked through to your site after reading an AI Overview is now being funnelled into a conversational interface that will generate its own synthesis - potentially drawing on your content, but again without delivering the click.
The Citation Visibility Gap Gets Wider
There has always been a gap between being cited in an AI Overview and receiving measurable traffic from it. Google Search Console now surfaces some AI Overview impression and click data, but the coverage is incomplete and the click volumes are typically low relative to the impression counts. That gap was already a challenge for attributing AI search value in a CRM.
This change makes that gap structural rather than incidental. If AI Overview links are increasingly pointing into AI Mode rather than to external pages, then the click-through rate from those Overviews will fall further - not because your content is less relevant, but because the exit route from the Overview has changed. You can earn the citation and still not earn the visit.
Platforms like Perplexity still resolve citations to actual source URLs, which is one reason AI-referred traffic from Perplexity tends to be more trackable than traffic originating from Google's AI surfaces. Google's direction appears to be moving further away from that model, not towards it.
What This Means for AEO Strategy
The instinct when faced with this kind of change is to double down on citation acquisition - get your brand mentioned more often, in more Overviews, so that even a smaller percentage of clicks still adds up. That logic is not wrong, but it is incomplete. If the architecture of AI Overviews is being used to retain users inside Google rather than export them, citation volume alone does not solve the attribution problem.
The more durable response is to focus on the type of queries where AI Overviews are less likely to fully satisfy intent - complex, specific, or high-consideration questions where a user still needs to visit a source to complete their task. These are the queries where the AI-generated answer creates appetite rather than satisfying it. Product-specific queries, technical comparisons, and queries with a clear action behind them tend to behave this way.
Structuring content to answer the follow-up questions that AI Mode generates is also worth considering. If Google's AI Mode is synthesising responses from your domain, the pages that answer second and third-level questions - not just the top-level query - are the ones most likely to be drawn on. Deep, well-structured content that anticipates follow-up intent has more surface area for AI extraction.
The Measurement Problem This Creates
If a user enters AI Mode via an AI Overview link and eventually clicks through to your site from within that AI Mode session, the referral path becomes difficult to reconstruct. Google Analytics 4 may record the session as AI Mode-referred, as a direct visit, or potentially not at all if the user abandons before clicking. GA4 already undercounts AI referral traffic in several documented scenarios, and this additional step in the journey makes the attribution chain longer and more fragile.
The practical implication is that if you are measuring the value of AI search visibility purely through referral traffic in GA4, you are likely to see that number plateau or decline even when your AI Overview citation rate remains stable or improves. The measurement does not reflect the visibility. That is not a tracking configuration problem you can patch with UTMs - it is a consequence of where the user journey now ends.
Connecting AI search visibility to CRM outcomes - lead source, pipeline stage, closed revenue - remains the most reliable way to understand whether AI search is contributing to your business. Direct website traffic attribution from AI sources will increasingly understate the real influence. Businesses that rely on last-click or even assisted GA4 data to evaluate their AEO programme will draw the wrong conclusions.
What to Do Now
First, audit where your brand currently appears in AI Overviews and note whether the links in those Overviews point to your domain or prompt AI Mode follow-ups. This requires manual checking across your target queries - there is no automated report for it yet. Document the pattern. Understanding whether you are in the cited source panel or in the AI Mode trigger links is a meaningful distinction.
Second, revisit your content architecture for the queries where you want AI visibility. Pages that answer narrow, specific questions are more likely to generate clicks even in a low-CTR environment, because the user cannot get everything they need from the synthesised response. Broad informational content that an AI Overview can fully reproduce is the category most at risk of delivering zero traffic despite strong citation rates.
Third, treat this as a reason to accelerate CRM-level attribution rather than defer it. If AI search influence on pipeline is going to become harder to read in GA4 session data, the brands that have already connected AI visibility signals to their sales data will be in a better position to make resourcing decisions. Those that have not will face an increasingly noisy picture and no clear way to judge what their AEO work is actually worth.