At the latest Google I/O, held on the 10th of May, Google announced further information regarding the upcoming AI-powered features for search. Google is calling this the ‘Search Generative Experience’ (SGE). SGE differs from Google’s AI chatbot, Bard, and instead will be used to complement search results by “producing informative responses that are corroborated by reliable sources”.
Google has implemented a ‘five point approach’ to generative AI search:
- Information needs – The aim is to reduce the steps it takes for the searcher to accomplish a task or goal, making the experience more fluid and seamless.
- Information quality – AI responses must be of a high level and quality.
- Safety constraints – Should Google provide answers that users would trust to be 100% accurate, when Google may not be able to verify the accuracy?
- Ecosystem – Google wants to provide traffic and credits to the source of the content, encouraging users to delve deeper into these sources.
- Ads – Can ads be relevant and provide additional information to the user, and be displayed in a way that is beneficial to the search?
SGE in action
The following screenshots are taken from Google I/O, showing a demonstration of how users can expect to see SGE results.
For this example, the query “good bike for a 5 mile commute with hills” is searched via Google. An integrated AI-powered snapshot appears above the usual results, with some important considerations relating to the query at the top. To the right, any sources that the AI has used for the displayed information are shown with a thumbnail image, title and URL. These are clickable to get through to the site.
Users then have the option to ask a follow-up question or choose from some suggested ‘next steps’ to help guide the search along.
Selecting any of these options will open up ‘conversational mode’, which allows users to type directly to the AI for further information, similar to AI chatbots like Bard and ChatGPT.
Limitations of SGE
One of Google’s five points is focused on safety and accuracy, which obviously is very important when users are searching for information that they expect to be correct and true. Google has announced that SGE will not answer every query, as if there is a ‘data void’ or any ‘information gaps’, it cannot be guaranteed that the results provided by SGE are accurate.
This will limit SGE from answering any queries that could potentially produce harmful, hateful or explicit content. For example, medical queries will not be answered using AI as there is a risk of false information appearing. It is not yet confirmed, but it is possible that this will apply to financial queries too, and likely many others that can’t be guaranteed as accurate and safe.
What does this mean for SEO?
Although it can be difficult to tell what sort of impacts SGE will have on SEO prior to its launch, it does seem as though Google is taking some steps to ensure that it does not have a notable negative impact.
Google has highlighted a desire to provide traffic to sites that the AI sources information from, rather than taking it away from the site entirely. Users will be able to click through to the cited websites, however, it’s worth noting that these links will not direct to the particular page that the information was taken from, just the domain itself. This lack of direct citation isn’t an ideal user experience, as when a user clicks the link to your site looking for further information on the displayed topic but is met instead with your homepage, they are more likely to bounce.
The AI-generated information presented for each search query will be pulled from the web, and it’s likely that the information selected will be from authoritative sites with high quality content, but this isn’t guaranteed. As well as this, underneath the SGE results the SERPs will display as normal, so it remains important to continue best SEO practices. However, if SGE and ads are taking up a lot of real estate in SERPs, even websites showing in top positions may see an impact on organic traffic.
Testing SGE
Users in the US are currently able to join a Google Labs waitlist to test the new features via labs.google.com/search under ‘Experiment with Search Labs’. Hopefully, this will become available to users in the UK soon.
