Google Expands AI Headline Generation to Search Results
Google is implementing an experimental feature that utilizes artificial intelligence to generate and display alternative headlines for news articles and other web content within its traditional search results. This initiative follows a similar rollout in Google Discover and has prompted various concerns from news editors and publishers regarding accuracy, potential misrepresentation of content, and loss of editorial control.
Google has confirmed it is conducting a "small" and "narrow" test involving AI-generated title links for news articles and other content types displayed on its search engine results pages (SERPs).
This test expands upon Google's existing practice of generating custom title links, which became more frequent in 2021, and its previous implementation of AI-powered headlines in Google Discover and push notifications. While the company has indicated that the test primarily affects news sites, it is not exclusively limited to them. Google spokespeople involved in addressing this initiative include Jennifer Kutz, Mallory De Leon, and Ned Adriance.
Google's Rationale: Enhancing Relevance and Engagement
According to Google, the primary objective behind generating these AI-powered titles is to "identify content on a page that would be a useful and relevant title to a users’ query."
The company states this approach aims for "better matching titles to users’ queries and facilitating engagement with web content."
For its AI-generated headlines in Google Discover, Google previously reported that the feature "performs well for user satisfaction" and helps users explore topics covered by multiple creators. Google described the AI titles in Discover as reflecting "trending topics" across various sites rather than merely rewriting individual article headlines.
Publisher Concerns Mount Over AI-Generated Headlines
News editors and publishers have raised several significant concerns regarding the implementation of AI-generated headlines in Google Search.
Accuracy and Misrepresentation
Critics report that these AI-generated headlines sometimes contain factual inaccuracies, misleading claims, or confuse details between different stories.
- Google's AI generated the headline "US reverses foreign drone ban" for a story that explicitly stated the ban was not reversed.
- Another instance saw a headline about "Steam Machine price & HDMI details emerge" when no such details had been released.
- Google shortened "I used the ‘cheat on everything’ AI tool and it didn’t help me cheat on anything" to "‘Cheat on everything’ AI tool," which altered the original meaning.
- A headline "Microsoft is rebranding Copilot in the most Microsoft way possible" was altered to "Copilot Changes: Marketing Teams at it Again."
Loss of Editorial Control
Publishers express concern that these AI headlines misrepresent their content and affect their ability to market their own work effectively.
- Publishers note that Google sometimes truncates genuine headlines, making them difficult to read.
- The system can also generate generic titles for original, detailed stories.
- Staffers at The Verge have observed that these AI-generated headlines do not adhere to their editorial style.
Impact on Trustworthiness
An editor with 15 years of experience noted that this practice deviates from Google's usual headline modifications, which typically involve shortening existing headlines or selecting between provided options. Concerns have been raised that altering headlines and their meaning could affect the trustworthiness of journalism.
Clickbait Filtering Ineffectiveness
The AI system has been observed failing to filter out human-generated clickbait headlines. An example provided was a headline promising a "Star Wars Outlaws Free Download" for an article that detailed only a single game code giveaway in a limited regional promotion.
Lack of Disclosure
Google does not include any disclosure indicating when an original headline has been altered. This means publishers may need to manually check for changes to their content's representation in search results.
Precedent and Unclear Future
The current test in Search follows a similar experiment in Google Discover, where AI headline rewrites were initially tested using "small and narrow" language. That feature later transitioned into a standard offering due to reported user satisfaction.
Google's historical practice involves rewriting title tags in search results using rule-based systems. However, the current test utilizes generative AI to create entirely new text, a significant departure from extracting existing content.
Google spokespeople have stated that if this experiment were to be widely implemented, it would not utilize a generative AI model for creating these headlines, though an alternative method was not specified. The test has not received approval for a wider launch. Crucially, Google's documentation for title links does not provide an option for publishers to opt out of these rewrites. Google declined an interview request for further explanation regarding the Discover feature.