Google Removes Text Fragments from AI Overviews, Blocking Traffic Tracking

Google has quietly removed text fragments from the citation links displayed within AI Overviews, effectively eliminating a key method that publishers and SEO professionals used to track traffic originating from Google’s generative search feature. The change, which has been confirmed by multiple industry observers over the past several days, strips away the scroll-to-text functionality that previously allowed users to jump directly to the specific passage on a website that Google’s AI had summarized. For website operators, however, the loss cuts deeper: the text fragment parameter had become a reliable, if unofficial, signal for identifying visits generated by AI Overviews within analytics platforms such as Google Analytics. Without it, distinguishing AI Overview traffic from ordinary organic search traffic has become significantly more difficult, if not impossible, depending on the setup. The history of scroll-to-text within Google’s ecosystem goes back several years. Google originally introduced the feature for featured snippets, the compact answer boxes that appear at the top of search results for many queries. When a user clicked a link within a featured snippet, the browser would automatically scroll to the relevant paragraph on the destination page, guided by a text fragment appended to the URL. This mechanism was later extended to the citation links in AI Overviews, providing a similar convenience for users while simultaneously giving publishers a straightforward way to identify these referrals. The text fragment typically appeared as a hash-delimited segment in the URL, something like #:~:text=keyword phrase, which analytics tools could capture as a distinct page identifier or referrer signature. For months, this allowed SEO practitioners to isolate AI Overview traffic, measure its behavior, and assess its value relative to other search entry points. The removal of these fragments changes that calculus entirely. Dana DiTomaso, a well-known figure in the digital marketing community, was among the first to publicly note the change on LinkedIn, reporting that the text fragments had vanished from AI Overview links. Her observations were soon corroborated by Thomas C. and Alexander Außermayr, both of whom independently confirmed the absence of the text fragment parameter in links generated by the feature. The pattern of reports suggests a broad rollout rather than an isolated test, though Google has not issued an official statement about the change. One possibility is that the removal is temporary, part of Google’s ongoing experimentation with the presentation and behavior of AI Overviews. The company has been iterating rapidly on the feature since its wider release, testing different layouts, link formats, and interactive elements. A temporary adjustment for testing purposes would not be out of character. However, the more widely held interpretation among SEO professionals is that this move is deliberate and strategic. By removing the text fragment, Google closes a loophole that allowed external measurement of a surface it controls. The company has consistently refrained from providing dedicated analytics for AI Overview traffic within Search Console, despite repeated requests from the SEO community. A dedicated filter or report that shows clicks, impressions, and user behavior specifically for AI Overviews would be a straightforward engineering addition, yet Google has declined to build one. The removal of text fragments aligns with that same posture: it makes independent measurement harder, reducing transparency around how much value AI Overviews actually deliver to publishers. The implications for website operators are substantial. Without the text fragment, standard URL-based tracking no longer distinguishes AI Overview clicks from regular organic clicks. Both appear as direct search traffic in most analytics configurations. Publishers who relied on this signal to evaluate the return on investment of their SEO efforts, or to understand how AI Overviews affect user engagement metrics such as bounce rate, time on page, and conversion rate, now face a blind spot. Workarounds exist, but they are less precise and more complex to implement. Server-side log analysis, for example, can sometimes reveal the referrer header more granularly than client-side analytics, but this requires technical infrastructure that many smaller sites lack. Custom JavaScript solutions that inspect the document referrer or the navigation timing API can offer partial recovery, but they are brittle and may be affected by browser privacy features. Another approach involves tracking user behavior patterns, such as rapid scrolling to a specific page position, which might correlate with the scroll-to-text behavior that was previously explicit. Yet none of these methods match the simplicity and reliability of a clean URL parameter. The change also raises broader questions about the relationship between Google and the publishers whose content feeds its AI systems. AI Overviews extract and synthesize information from third-party websites, then present that information within Google’s own interface, reducing the incentive for users to click through to the original source. When users do click, the visit is valuable, but measuring it accurately has just become harder. This dynamic echoes earlier controversies around featured snippets and zero-click searches, where Google’s answers satisfied user intent directly on the search results page, reducing traffic to publishers. The difference now is that AI Overviews are more comprehensive and conversational, potentially displacing even more clicks. The timing of this change is notable. AI Overviews are still in a phase of rapid adoption and refinement. Google is simultaneously expanding the feature to more queries, more languages, and more regions while also refining the quality and accuracy of the answers. Within this context, controlling how the feature is measured and perceived matters greatly. If publishers cannot accurately track AI Overview traffic, they cannot quantify the impact of being cited, which in turn affects their content strategy and their willingness to optimize for inclusion in AI Overviews. Some may choose to focus on queries where AI Overviews are less dominant, while others may invest in structured data or other signals that might improve their chances of being referenced. But without clear measurement, every decision becomes a guess. The broader SEO industry has been grappling with the challenge of measuring AI-generated search surfaces for months. Several third-party tools have attempted to estimate AI Overview visibility and traffic, but these estimates rely on proxy data and small sample sizes. The removal of text fragments widens the gap between what Google knows internally and what publishers can observe externally. For Search Console to become the definitive source of truth for AI Overview performance, Google would need to add specific metrics, and the company has shown no indication that it plans to do so. Until that changes, the industry will continue to rely on indirect signals, inference, and educated estimates. What makes this moment particularly frustrating for many SEO professionals is that text fragments were not a tracking mechanism by design. They were a user experience feature, introduced to make browsing more seamless and intuitive. That they also enabled tracking was a side effect, not a primary purpose. Google’s decision to remove them may have been motivated by privacy considerations, product design goals, or a desire to simplify the URL structure, but the effect on measurability is the same regardless of intent. The practical outcome is that a useful measurement signal has been withdrawn, and no replacement has been offered. For publishers who have built workflows and dashboards around AI Overview traffic identification, the immediate task is to find alternative methods. Some may turn to Google Analytics 4’s built-in machine learning models, which can attempt to classify traffic sources based on behavioral patterns. Others may implement custom dimensions that capture the referring page URL or the navigation entry point more aggressively. A few may experiment with server-side tagging or data layer modifications that attempt to reconstruct the context of the visit. None of these solutions are perfect, and all require ongoing maintenance as browsers and platforms evolve. The long-term strategic response may be more significant. If AI Overview traffic cannot be measured reliably, then optimizing for AI Overview inclusion becomes a bet with unclear odds. Publishers may prioritize content that targets queries where AI Overviews are less likely to appear, such as highly specialized or locally specific topics. Alternatively, they may focus on building direct traffic channels that bypass search intermediaries altogether, including email newsletters, social media, and direct site visits. The value of being cited in an AI Overview may shift from a measurable traffic driver to a brand visibility signal, with softer, harder-to-quantify benefits. This is not an entirely unfamiliar scenario. The SEO industry has learned to operate with incomplete data before. The loss of keyword-level data in Google Analytics with the move to not provided, the deprecation of the Google Search Console average position metric for some interpretations, and the gradual reduction of data fidelity in Google Analytics 4 have all forced practitioners to adapt. But each time, the adaptation requires new tools, new methodologies, and new assumptions. The removal of text fragments from AI Overview links is another step in that same direction, making the analytics landscape slightly more opaque. For Google, the calculus may be straightforward. AI Overviews are a product designed to keep users within the Google ecosystem longer, providing answers directly while still offering pathways to external sites. Making those pathways measurable weakens Google’s control over the narrative around the feature’s value. If publishers could definitively show that AI Overviews send minimal traffic, the pressure on Google to share more revenue or provide more attribution would increase. By limiting measurement, Google retains the ability to define the story. The industry response so far has been measured but pointed. LinkedIn posts from professionals like DiTomaso, Thomas C., and Außermayr have sparked discussions, with many SEOs sharing their own observations and workarounds. The general sentiment is one of resignation mixed with determination. Most acknowledge that Google is unlikely to reverse the change or provide an alternative. The focus is shifting to how to live with the new reality rather than how to reverse it. One open question is whether this change applies universally across all AI Overview implementations, including those in different countries and languages. Early reports suggest broad coverage, but sporadic exceptions may exist. Publishers who notice text fragments still appearing in some links should consider whether those are cached pages, lingering remnants of an earlier version, or genuine regional differences. Consistency will likely increase over time as the change propagates fully through Google’s infrastructure. Another consideration is the impact on user experience. The scroll-to-text behavior was genuinely useful for users, saving them from manually searching a long page for the relevant passage after clicking a citation link. Its removal makes the post-click experience slightly less smooth, which could subtly reduce user satisfaction. Google may have determined that this cost is acceptable, or the company may be testing an alternative mechanism that achieves the same goal without exposing a tracking signal. If Google introduces a new, non-trackable method for deep linking within AI Overviews, the user experience could be restored while the measurement challenge remains. The SEO community will be watching closely for any such development. In the meantime, publishers should audit their current analytics setup to understand what data they can still capture. Identifying AI Overview traffic through URL fragments is no longer viable, but other approaches may still work. Comparing traffic patterns before and after the change can help calibrate expectations. Engaging with analytics vendors and platform providers about their roadmaps for AI-generated search traffic measurement may also yield new solutions over time. The broader context is that AI Overviews are still in their early stages. How they evolve, how users interact with them, and how they affect the broader search ecosystem will become clearer over the next year. The removal of text fragments is one data point in that evolution, suggesting that Google is tightening its control over the metrics that matter most to publishers. For SEO professionals, the response should be pragmatic: adapt measurement approaches, diversify traffic sources, and continue to advocate for more transparency from Google. The ability to measure and optimize is fundamental to the practice of search engine optimization, and any reduction in measurement capability must be met with creative and technical solutions. Google’s decision to remove text fragments from AI Overview links marks a significant moment in the ongoing negotiation between the search giant and the publishers that supply its results. It reduces transparency, increases uncertainty, and forces a recalibration of how AI Overview performance is assessed. For now, the industry is left to work with imperfect data, as it has done before and will likely do again. The change is subtle, but its implications are far-reaching, touching on questions of measurement, attribution, value, and trust that sit at the heart of the modern search ecosystem.