Google has now updated how links and carousel link cards appear inside AI Mode and AI Overviews more times than most marketers can count. Each iteration comes with the same stated rationale: make it easier for users to click through to publishers and creators. Each time, the implied subtext is that the previous format was not doing that job well enough.
That pattern tells you something. Not about Google's design competence, but about the structural tension at the centre of AI-generated search responses. The AI answers the question. The link is an afterthought. No amount of carousel redesign resolves that fundamental ordering.
What Repeated Link Redesigns Actually Signal
When a product team iterates on the same UI element repeatedly, it usually means the metric they are optimising for is not moving. In this case, that metric is clicks from AI responses to external sites. Google has been transparent that driving those clicks is the intention. The fact that this keeps getting revisited suggests the behaviour change is not following the design change.
This is not a small problem for brands that have invested in earning citations inside AI Overviews. If citation earns you a mention but not a visit, the SEO value model breaks. Traditional search operated on a clear chain: ranking led to impression, impression led to click, click led to session, session led to conversion. AI search does not follow that chain reliably. The citation is real. The click may not be.
What makes this harder to diagnose is that Google has not released click data from AI Mode or AI Overviews in any usable form. Google Search Console shows some AI Overviews impression data, but the click attribution picture remains incomplete. Brands are left inferring behaviour from traffic patterns rather than measuring it directly.
Citation Without Clicks Is Still a Signal Worth Having
Before writing off AI citations as valueless without the click, it is worth being clear about what citation does achieve. When your brand or content is consistently surfaced inside AI responses, it contributes to something closer to brand exposure than traditional organic traffic. A user reading an AI Overview that cites your site has encountered your brand in a context that implies authority. That matters, even without a click.
In B2B and considered-purchase categories, buyers often conduct multiple AI-assisted research sessions before visiting any website directly. They may type your brand name into a search bar days later, arriving as direct or brand-search traffic. That journey is invisible in most attribution models. The AI citation is not the end of the funnel - it is part of the research phase that precedes it.
This is why measuring AI search purely by referral traffic is structurally flawed. It captures only the subset of users who clicked the link card in the AI response. It ignores brand uplift, latent demand, and downstream search behaviour. For accurate picture of AI search contribution, you need to look further into the funnel than your analytics dashboard typically goes.
How to Measure AI Visibility When Click Data Is Missing
The absence of reliable click data from Google on AI Mode and AI Overviews is not a reason to stop measuring. It is a reason to measure differently. Start by isolating the referral traffic you can see. Sessions from ChatGPT, Perplexity, and Gemini appear in GA4 as referral traffic with identifiable source domains. Google AI Mode referrals are trickier - they often arrive as organic sessions with query strings that have been stripped, or fold into direct traffic depending on how the user clicked.
Beyond referral sessions, track brand search volume over time alongside your AI citation frequency. If you are building a programme around earning AI citations, you need a baseline of branded search volume before the programme starts. Increases in brand-direct and branded organic traffic in the months following a visibility uplift in AI surfaces are a reasonable proxy signal, though not a clean attribution.
The most robust approach is to extend your measurement into the CRM. If a lead arrives and lists a research journey that included AI-assisted search, that qualifies for AI search attribution. Not all leads will volunteer this, but structured intake questions - what prompted your search, which tools did you use - cost nothing to add and build a qualitative picture over time. Revenue attributed through CRM interviews is harder to scale than analytics data, but it is considerably more accurate.
What This Means for Content and Citation Strategy
If clicks from AI responses are structurally low - not because your content is weak, but because AI interfaces are designed to answer rather than redirect - then the strategic goal shifts. You are no longer optimising purely for the click. You are optimising to be cited in the answer, with the click being a bonus rather than the primary outcome.
That changes how you think about content structure. Concise, extractable answers to specific questions perform well in AI citation because the AI can surface them directly. Longer, exploratory content may drive deeper engagement if the user does click through, but it is less likely to be directly quoted. Both have a role, but they serve different objectives in the AI search funnel.
Schema markup, entity consistency across your site and third-party sources, and clean crawler access remain foundational. But the goal of those signals is to make your content reliably machine-readable and trustworthy to AI systems - not to guarantee a click. Thinking of schema as a click driver misses the point. It is an eligibility signal. The click, if it comes, is downstream of eligibility.
The Wider Implication for AI Search Programmes
Google's repeated link format changes in AI Mode and AI Overviews are a live experiment in whether AI search can generate publisher traffic at meaningful scale. The honest reading of the situation is that it cannot do so at the same rate as traditional blue-link search. That is not a temporary problem. It reflects the purpose of the product.
For brands running structured AEO or GEO programmes, this does not undermine the investment. It reframes what success looks like. Citation frequency, brand mention consistency, downstream branded search uplift, and CRM-attributed leads from AI-assisted journeys are all legitimate success metrics. Referral sessions from AI surfaces are useful to track, but they should be one measure among several - not the primary one.
The brands that will do this well are the ones that accept that AI search measurement is fundamentally different from organic search measurement, and build their reporting accordingly. Waiting for Google to release clean click data before committing to a measurement framework is not a strategy. That data may never arrive in the form traditional SEO reporting expects.