AI Overviews and Original Journalism: Navigating the Fine Line Between Attribution and Infringement

To go back to this article, head to My Profile and then click on Saved Stories.

Rogers Reece

Google's AI-Generated Search Summary Replicated My Unique Content

Recently, a search summary generated by Google's AI took content from an article I wrote for WIRED and repurposed it in a manner that has me worried about the fate of journalism.

I had been testing out AI Overviews, the latest AI tool from the company aimed at addressing internet searches. I posed several questions related to subjects I had just reported on, so it didn't surprise me to find a reference to my article at the very end of the answer box. However, I was unexpectedly struck by the extent to which the opening section of an AI Overview seemed to directly lift from my own work.

The image on the left captures a conversation I had with a product developer at Anthropic, where we discussed strategies for utilizing the Claude chatbot developed by their company. On the right, there's a snippet from Google's AI Overview, providing guidance on operating Anthropic's chatbot. Comparing both texts, it's akin to observing a student who has plagiarized my homework, making minimal effort to alter the wording.

Before activating the AI Overviews, my piece frequently appeared as the top highlighted snippet in Google search outcomes, providing an easily accessible link for users seeking guidance on utilizing the Claude chatbot. In the early stages of experimenting with Google's revamped search feature, my article continued to show up in the top results for pertinent searches. However, it was relegated below the AI Overview response, which utilized information from my article to compile a 10-point bullet list incorporating elements of my work.

During conversations through emails and a phone conversation, a representative from Google confirmed that the AI-generated summaries might incorporate text verbatim from websites. However, they justified the use of AI Overviews by stating they clearly point back to the sources they derive from. In my instance, the initial section of the response didn't credit me directly. Rather, my original piece was merely one among six references linked at the end of the outcome. Given the placement of these source links at such a distance, it seems unlikely that publishers would see a notable increase in visitors under these circumstances.

In a statement provided to WIRED, a representative from Google explained, "AI Overviews aim to reflect key information found in the highest-ranking internet searches, including those sources referenced within the summaries themselves. This feature isn't meant to substitute for the actual web content but rather to give individuals a quick glimpse of what information is available, encouraging them to delve deeper." However, upon examining the language and structure of the specific AI Overview in question, I find myself at odds with Google's assertion that it merely provides a "conceptual match" to my original content. In reality, it goes beyond that. Moreover, despite Google engineers not designing this tool to replace original articles, AI Overviews offer straightforward answers that obscure where the information originated and diminish the likelihood that people will follow the links to the original content.

"The Google representative noted that AI Overviews containing links tend to receive more clicks compared to standard web listings for the same searches," the spokesperson remarked. However, WIRED was not provided with any supporting data for this statement, making it difficult to confirm the effect of the AI functionality on click-through rates independently. Additionally, it is important to mention that the comparison was made between the traffic coming from AI Overview referrals and the conventional blue-link Google traffic, not against the selections made for a featured snippet, which presumably have significantly higher rates.

Upon contacting Google regarding the AI Overview response which sourced information from my article, the experimental AI search outcome for that particular inquiry ceased to appear. Nevertheless, Google continued to try and create a response above the highlighted snippet.

Numerous lawsuits involving artificial intelligence are still pending resolution. However, when I consulted a copyright law expert for their opinion, they expressed doubts about the success of any potential lawsuits I might consider. "I don't believe you would have a strong legal argument for copyright infringement," stated Janet Fries, a lawyer with Faegre Drinker Biddle & Reath. She emphasized that copyright law is designed to avoid hindering beneficial and useful contributions. Focusing on the nature of the content in question, she clarified that it's challenging to assert copyright over instructional or factual material, such as an advice column, in comparison to more inherently creative outputs, like poetry.

Certainly, I'm not pioneering the idea of tailoring chatbot interactions to the target audience, and I acknowledge that my approach of prioritizing factual content does add complexity to the matter. However, it's challenging for me to envision a scenario where Google could pinpoint that specific detail regarding Claude's chatbot in its AI Overview findings without having consulted my contributions initially.

Authored by Joseph

Authored by Matt

Authored by Dhruv Mehrotra

Written by Hannah Zeavin

In discussing the potential outcomes of a theoretical copyright lawsuit grounded on my research, another specialist in the field expressed ambiguity regarding the court's decision. "I believe there's still an opportunity to present this case to a jury," mentioned Kristelia García, an academic at Georgetown University Law Center. "To then ascertain whether they would deem it 'substantially similar,' which is the criterion used by the law."

Just to make things perfectly clear, everything mentioned here is purely speculative. I'm not considering taking any legal action against Google. The purpose of this investigation was to enlighten the public on the effects of artificial intelligence and to encourage meaningful conversation.

Regardless of whether Google has violated WIRED's copyright in this particular case, it seems clear to me that should Google decide to make its AI Overviews more widespread, this move could significantly alter the landscape of digital journalism, and perhaps not for the better. Nilay Patel, the cofounder and editor-in-chief of The Verge, frequently discusses the notion of "Google Zero," a hypothetical scenario where publishers realize that their once-steady stream of traffic from the internet's biggest source has vanished. Given Google's significant influence on the way individuals search online, the tech giant has the power to drastically reduce traffic, and even potentially cause the downfall of certain publications, by merely tweaking how its services operate.

Given the financial constraints already burdening the journalism sector, it's not surprising that executives at news outlets, worried about the latest advancements in artificial intelligence, are quickly securing agreements with prominent AI firms. Organizations such as The Associated Press, The Atlantic, the company overseeing Business Insider and Politico, the Financial Times, Vox Media, the group behind The Wall Street Journal, and the online media powerhouse Dotdash Meredith, are among those that have entered into partnerships with OpenAI. However, the average employees might not be too pleased with the idea of their creative work being traded in these deals.

The importance of trustworthy journalism is evidently recognized by companies dedicated to artificial intelligence, despite their lack of financial compensation to content creators for their work. Liz Reid, who oversees Google's search operations, recently pointed out in a blog post that the issue of misleading AI-generated summaries, such as a suggestion to use harmless glue on pizza to increase cheese adhesiveness, can often be traced back to user-generated content, particularly from discussion forums. The specific advice about pizza may have originated from a Reddit user going by the name “fucksmith.” As Google endeavors to provide concise, AI-powered responses to queries, it shoulders a greater responsibility to ensure the accuracy of information and to promote content of high quality.

If you plan on copying someone's homework, ensure it's from a top-performing student, not a troublemaker on Reddit.

Suggested for You…

Delivered to your email: Will Knight delves into AI innovations in his Fast Forward series.

Welcome to the chaotic world of automated advertising

What is the required number of electric vehicle (EV) charging points to supplant gasoline stations in the US

A charitable organization aimed to reform the technology industry but ended up losing grip on its governance.

Eternal Sunshine: Discover the Ultimate Sunglasses for Every Expedition

Rogers Reece

Knight Will

Kate Knibbs

Dave Paresh

Lauren Goode

Reece Rogers

Knight Will

Name: Louise Matsakis

Additional Insights from WIRED

Critiques and Instructions

© 2024 Condé Nast. All rights reserved. WIRED is eligible to receive a share of revenue from items bought via our website, thanks to our Affiliate Agreements with retail partners. Reproduction, distribution, broadcasting, storage, or any other form of usage of the content on this website is strictly prohibited without obtaining prior written consent from Condé Nast. Advertising Options

Choose a global website