Google's AI Overviews, which are AI-generated answers that can show up when you do a Google search, now includes a right-hand display of links. Google is also testing relevant in-text links in AI Overviews as an experiment so users can see cited information throughout the response.
The update is available on desktops, and also mobile devices if you tap the site icons at the upper right of the screen.
Hema Budaraju, senior director of product management for Google Search, told Business Insider that the update will make it easier for people to visit sites that interest them.
Google launched AI Overviews in May at its annual I/O conference . The tool offers a summarized version of Google Search results at the top of the page for some queries. While the previous version included links in its responses, they were less front-and-center. In the past, you could scroll down or click a down arrow at the bottom of a paragraph to see the sources.
The changes make Google's AI search product more closely resemble OpenAI's SearchGPT, which also provides in-text links and an expandable sidebar of sources. Powered by the GPT-4 family, SearchGPT responds to inquiries with up-to-date information from the web. The tool is in testing and available to a small group of users and publishers, according to the company's announcement.
Google's AI Overviews, however, is far more widely available, and there are still some key differences between the two. Unlike AI Overviews, SearchGPT is set up like a chatbot, so you can ask it follow-up questions in the chat rather than making a new search. OpenAI has also shared plans to integrate the best features of SearchGPT directly into ChatGPT .
Meanwhile, Google is embedding AI into its already-established search engine. Google also offers a chatbot experience with its virtual assistant Gemini.
The deals come amid concerns from publishers around the impact AI products could have on their readership, as well as any scraping of their online work without permission to train the large language models powering AI products. AI-powered answer engine Perplexity faced backlash, for example, when it was accused of regurgitating news content without providing attribution to publishers.
While Google also seems to be taking steps towards expanding attribution, it hasn't taken the same approach in signing deals with US publishers. However, Google is paying Reddit for the ability to train its AI on the company's vast trove of online posts. The tech giant also announced in 2023 that it signed licensing deals with 1,500 news publications in the EU due to the EU Copyright Directive.
But as companies continue to experiment with what makes a strong generative AI search product, publishers have been watching to see how the concept of linking to the information cited is implemented. And from a user standpoint, AI models are known to hallucinate, so products that reveal where the AI is getting its information from can be helpful when fact-checking purposes, or to further explore a topic.
Budaraju said in a statement that Google will continue testing different ways of presenting information, while "prioritizing approaches that drive traffic to relevant websites."
"Showing links to supporting web pages directly within AI Overviews is driving higher traffic to publisher sites," Budaraju said in a statement about the experiment.
Google's update to AI Overviews comes as it rolls out AI Overviews to six new countries, including the UK, India, Japan, Indonesia, Mexico, and Brazil. The updated version with the links is available in all of the countries it's launched in, as well as for Search Labs users in over 120 countries and territories.
Google is also experimenting with new tools in Search Labs for English queries in the US, like the ability to save AI Overviews results and simplify the language in the responses.
Get updates delivered to you daily. Free and customizable.
It’s essential to note our commitment to transparency:
Our Terms of Use acknowledge that our services may not always be error-free, and our Community Standards emphasize our discretion in enforcing policies. As a platform hosting over 100,000 pieces of content published daily, we cannot pre-vet content, but we strive to foster a dynamic environment for free expression and robust discourse through safety guardrails of human and AI moderation.
Comments / 0