Google has now opened up access to their generative AI labs test to a small portion of testers. This means a select handful of users are now able to test Google’s highly anticipated AI-powered results.
In the past, I have reviewed Google’s Bard chatbot and felt as though Bing was doing a better job overall. My biggest gripe with Bard has been the lack of citations, something that generative AI within Search is designed to do a better job of.
Once I saw access was rolling out to users, I was very eager to jump in and run some tests for a selection of queries where I’ve been wondering how generative AI would present their search results. Thanks to the help of Brendan OConnell, Senior Manager of Organic Search at The New York Post, I was all set to go.
Here are my initial thoughts on generative AI in Google’s search results with the Search Generative Experience (SGE).
How to gain access to SGE
Similar to how Bard was first made available, there is a waitlist that you’ll need to sign up to. You’ll need to be located in the US (or at least using a VPN) to be added to the list, along with not having a Workspace account, which can be a roadblock for many. Tip: use an old gmail address or sign up for a new one if you’re eager.
At the time of writing this article, Google has only sent out invites to test their Search Generative Experience (SGE) to a small portion of testers. So if you haven’t received an invite just yet, I would expect an email within the next week or so letting you know that access has been granted.
Moving on, now for the information that you’re most interested in. Beyond the previews shown at Google’s I/O event, here are some examples of how Google’s search results have changed, how it will impact publishers, my feedback for Google, along with some additional insights.
Search Generative Experience (SGE) vs. normal Google
The most surprising aspect to me about Google’s Search Generative Experience is that featured snippets are still available alongside AI answers. I was previously under the impression that featured snippets would be replaced by AI answers, so this is already a better case scenario for publishers.
It is also interesting to note that users seem to need to ask the generative AI to provide the answer by clicking ‘generate’ if it isn’t showing by default. While it isn’t currently showing for some queries (such as for queries about politics), it does appear to show quite broadly.
Based on the approach to featured snippets and allowing users to select ‘generate’ to show the AI answers, Google is trying their best to not shake up the ecosystem too much. If Google were to launch the SGE available to beta testers, it would create far less volatility than we previously thought.
Here’s what this looks like to generate an AI answer within the SGE:
Another good example of Google’s SGE in action is for a query where I currently rank within the featured snippet for an article in the US. All of the content within my article is unique to my expertise (I have a ton of experience in Marketplace SEO), so I’m confident that I’m the original source of information.
When searching for the query ‘”what is marketplace seo” within the SGE, I’m provided with the generative AI answer at the top of search results with quite an extensive summary, along with 3 links to articles that are supposedly the source for the AI component of the SERP.
The issue here is that the 3 links are seemingly presented as the source for the AI answer. While this may be true to some extent, this is not the full truth. Google’s SGE is taking content from my article and presenting it as their own. The reason I know this to be true is that I have explicitly experimented with the wording and Google only started using that text in the featured snippet recently.
The content in itself could seem as fairly generic for anyone that knows SEO (it includes content, links and technical), and Google seems to be using that to their advantage. But I would bet that Google wouldn’t use that exact same wording if it wasn’t on my web page.
Update: Google has now actually made a change to the SGE example shown above. Previously, Google was not showing my website within the sources list in the carousel at the top, only in the featured snippet below. Google was clearly using content from my website, so it is good to see that they have updated this specific result. Attribution for exact match text usage needs to become more precise however.
Whoa, Google corrected this one fast. Some of the content from my site was being used in the AI answer, but wasn't showing as a link at the top (just in the featured snippet). Google has now fixed this to show in both spots. Documented here: https://t.co/d0w60fca4L pic.twitter.com/au7yxFbW90— SERP Alert (@SERPalerts) May 31, 2023
New opportunities for eCommerce with SGE
While it’s easy to focus on the downsides of SGE from a publisher’s perspective (it could easily be seen as less attribution), there are some specific benefits to eCommerce sites that I’d like to highlight.
Free product listings are a feature of search that have been around for a while in US search results. For clients of mine that have strong coverage for free product listings, the CTR for these results is exceptionally high. With SGE, Google is making these listings even more prominent.
In my opinion, this presents a significant SEO opportunity for eCommerce sites that are doing a good job of managing their product feeds, structured data, and optimization tactics for free product listings. Google is however showing different product listings at the top of search results compared to the bottom, so stores may need to adjust their strategies to get the prominent placement.
Note: with the SGE, this is the first time where organic links actually take preference to ads. When generating a SGE answer with AI, Google doesn’t appear to be placing another set of ads above the result. Currently, ads will be placed at the top of search results for the vast majority of queries.
With a high portion of searchers moving to Amazon in recent years for product searches, SGE may bring more searchers back to Google and in turn generate more searchers and more organic clicks for sites that are most visible. For sites that are in the eCommerce space, make sure to track your merchant listings closely to stay on top of this dataset.
Overall, I am impressed with how Google has gone about their early integration of generative AI within search results. There is still plenty of room for improvement to the Search Generative Experience (with respect to correct attribution), but it seems to be on the right track from what I’ve seen so far.
Here is a summary of the key points mentioned within my article:
- Google has opened up early testing of their SGE via Labs to some users. If you haven’t got access yet and are on the waitlist, I would expect that to change over the coming week or so.
- Google looks to be using generative AI for a lot more queries than I had expected. But the catch is that users can select ‘generate’ for many of these queries which prompts the AI answer to appear.
- Featured snippets have not been removed from search results like I had initially thought. Featured snippets are presented directly below the AI answers if triggered. As a user, I’m not sure how I feel about having two prominent answers to questions in search results like this (but as an SEO, I’m not complaining!)
- Attribution for content used within AI answers on Google needs to get better. After taking a look at a bunch of different queries, I quickly found an answer for my site that wasn’t correctly attributed. This is going under the assumption that the 3 visible links at the top of SERPs are considered to be the primary sources for the AI answer.
- There is a major opportunity for eCommerce sites with the SGE that is important to highlight. Track the ‘merchant listings’ search appearance filter in Google Search Console and understand where the gaps are in your free product listings strategy.
As I complete more testing of Google’s Search Generative Experience I’ll be sure to update this article with any new findings. In the meantime, keep an eye out for an email from Google and let me know on twitter if you see anything interesting.
Frequently Asked Questions
What is Google’s Search Generative Experience (SGE)?
The Search Generative Experience (SGE) is Google’s integration of generative AI into their search results. SGE is currently a Google labs experiment that is only available to users based in the United States, with expansion to come in the near future. Google’s SGE is in direct competition to Bing Chat and ChatGPT.
How long is the SGE beta set to last?
Google’s SGE experimental beta will end in December of 2023. This is based on the notice that appears once SGE is enabled for accounts. It is expected that until Dec 2023 is when the experiment will end and will then transition to Google’s live search results, beyond the constraints of the beta. This information was first caught by Search Engine Roundtable.
How can I get early access to labs and SGE?
If you’re located outside of the US, you’re still able to gain access to labs with the right approach. You’ll need to use a VPN (ExpressVPN for instance) and set your location to appear like you’re in the US. You will also need an email account that isn’t a Google Workspace account. My suggestion is to use an @gmail.com address if you’re having issues.
What are the differences among Google AI Search and Bing Chat?
There are clear differences among Google AI Search and Bing Chat with respect to how citations for text usage operates. In particular, Bing Chat provides direct attribution when there is a match among text on a web page and what the AI answer is. Whereas Google AI Search does not apply direct attribution in its current form, instead using general sources for text usage.
More Reviews of Google’s SGE
- A few thoughts on my initial Google Search Generative Experience (SGE) Tests – Aleyda Solis
- 11 realizations from testing Search Generative Experience – Kevin Indig
- Test driving Google’s Search Generative Experience – Eric Enge
- Examples of the SGE in action, Topic Authority system, Microsoft Copilots and Plugins – Marie Haynes