A significant court filing submitted on September 17, 2026, in the consolidated copyright lawsuit currently unfolding in the federal courts of New York has brought internal communications and proprietary performance data into the spotlight. The brief, filed by a coalition of news publishers including The New York Times, offers a rare glimpse into the internal philosophy of AI developers regarding the role of source attribution and the erosion of the referral traffic model that has historically sustained the digital news industry.
Internal Skepticism Regarding User Behavior
The filing highlights internal admissions from OpenAI staff that cast doubt on the efficacy of link-sharing within chatbot interfaces. According to the court documents, a software engineer at OpenAI remarked in February 2023, "no matter how prominently we show the links, users won’t click." This candid assessment suggests that the developers were aware from the early stages of product development that the conversational nature of Large Language Models (LLMs) creates a friction-filled path for users to visit source websites.
This sentiment was echoed by Nick Turley, OpenAI’s head of ChatGPT, who addressed the functionality of the "browse" feature. The brief cites Turley stating that there was "no good reason to click" on a link to an original source once the AI had already synthesized an answer for the user. These statements form the crux of the publishers’ argument: that OpenAI’s design choices are not merely incidental to the technology but are deliberate features intended to retain user attention within the AI ecosystem at the expense of external content creators.
The Erosion of Click-Through Rates
Central to the publishers’ case is data allegedly sourced from Microsoft, which provides a stark comparison between user behavior in traditional search environments versus conversational AI interfaces. The data reveals that for The New York Times’ websites, the Click-Through Rate (CTR) in Bing Chat—the precursor to the current Copilot service—was 87% to 93% lower than in standard Bing Web Search results.
Similar patterns of decline were reported for other publishers involved in the litigation. The Daily News plaintiffs experienced a drop in CTR ranging between 83% and 91%, while Ziff Davis sites saw a decrease between 51% and 94%. These figures represent a critical point of contention. The publishers argue that these dramatic reductions in traffic provide empirical evidence of "substitution," a legal concept suggesting that the AI product acts as a direct, non-transformative replacement for the original copyrighted work.
While the brief notes that "common sense suffices" to demonstrate harm to their business models, the inclusion of these statistics serves to quantify the potential economic impact of generative AI integration into search engines. However, the exact methodology behind this data remains opaque. The filing does not specify the timeframe of the collection, the volume of queries analyzed, or the specific criteria used to define the high and low ends of the ranges provided. Furthermore, as Microsoft rebranded its chat interface to "Copilot" in November 2023, the relevance of this older Bing Chat data to the current iteration of the product remains a subject of anticipated debate in future filings.
A Timeline of the Copyright Dispute
The current legal standoff is the culmination of years of tension between media organizations and technology firms.
- Early 2023: OpenAI engineers and leadership discuss the impact of AI responses on external traffic, with internal memos noting that users are unlikely to click source links.
- Late 2023: Microsoft rebrands Bing Chat to Copilot, integrating it more deeply into the Windows ecosystem and the broader Microsoft search architecture.
- September 4, 2026: OpenAI and Microsoft file their respective motions for summary judgment, arguing that their training and deployment processes constitute "fair use" under existing copyright laws.
- September 17, 2026: News publishers file a combined summary judgment brief, introducing the internal communications and CTR data that allege evidence of market substitution.
- September 24, 2026: Judge Sidney H. Stein issues a procedural order establishing the timeline for further submissions, with oppositions due by October 23 and reply briefs due by November 20.
The Defense: The Fair Use Doctrine
OpenAI and Microsoft maintain that their use of copyrighted material is protected under the doctrine of fair use. In their summary judgment motions filed earlier in September, both companies argued that their AI models provide transformative value. According to reporting by Axios, OpenAI’s motion included references to academic studies suggesting that the outputs of ChatGPT do not function as commercial substitutes for news publications.
A spokesperson for Microsoft stated in response to inquiries from the press that their legal filings thoroughly outline "why Copilot is not a substitute for publishers’ journalism." The tech giant asserts that its tools are designed to facilitate information discovery rather than to cannibalize the audience of news providers. As the case proceeds, the defendants will likely challenge the publishers’ interpretation of the CTR data, potentially arguing that the shift in user behavior is an evolution in search technology rather than an act of infringement.
Broader Implications for the Digital Ecosystem
The debate over CTRs in the age of AI touches upon the foundational economics of the internet. For decades, the "search engine referral" model has served as the primary pipeline for news organizations to reach readers and monetize their work through advertisements and subscriptions. If generative AI models fundamentally alter this pipeline, the financial viability of journalism may be at risk.
The industry is currently grappling with how to measure visibility in an era where traditional "clicks" are no longer the primary metric of engagement. Microsoft has introduced "Citation Share" in its Webmaster Tools, a metric that tracks how often a site is cited by an AI model, but the company has explicitly stated that this does not correlate to traffic or visitor counts. Similarly, Google’s generative AI reporting tools provide impression data for AI overviews but omit click data, leaving publishers in the dark regarding the true conversion rate of AI-mediated search.
This divergence in measurement—moving from traffic-driven metrics to citation-driven metrics—represents a significant shift in power. If publishers can no longer rely on referral traffic, they may be forced to rely on licensing deals or other compensatory mechanisms to sustain their operations.
Looking Ahead
The legal proceedings overseen by Judge Stein are expected to intensify in the coming months. The upcoming October 23 deadline for oppositions will likely see Microsoft and OpenAI provide a technical rebuttal to the CTR data and the internal emails surfaced by the publishers. Whether these responses will provide context that nullifies the publishers’ claims or simply open new avenues for discovery remains to be seen.
Legal experts suggest that the outcome of this case could set a massive precedent for the future of the web. If the court finds that the "substitution" of traffic constitutes copyright infringement, it could necessitate a complete overhaul of how generative AI models are trained and how they present information to users. Conversely, a ruling in favor of the tech giants would likely solidify the transition toward a "zero-click" internet, where the synthesis of information is prioritized over the direct navigation to source material. As the November 20 deadline for final replies approaches, both the media and technology sectors await a decision that will redefine the digital landscape for the next decade.
