At Google I/O 2023, Google unveiled its latest plans for the future of search, powered by AI technology. Speculation was intense ahead of the conference with many expecting a bombshell announcement – some kind of AI search revolution.
Instead, Google teased the future of AI-powered search by unveiling an experimental project, called Search Generative Experience (SGE). The project makes big promises about the future of AI search but only gives limited access to users in the US (for now).
The most exciting minute and a half of Google I/O 2023 was the search giant’s teaser video for SGE. It’s a dizzying mix of rolling bass lines, quick cuts and slices of AI search in action – a fine introduction to the future of search.
And that’s all it is: an introduction, for now. Because the future of search isn’t quite here yet. Google’s Search Generative Experience is an experiment, only open to users in the US who request access via Search Labs.
Google hasn’t specified whether or when this will open to users in other territories.
Anyone expecting an overnight search revolution from Google I/O might be disappointed by the lack of immediacy. However, we’re starting to see Google’s intention with AI take shape.
Some have accused the company of falling behind and reacting too slowly to the adoption of ChatGPT and generative AI tools. But Google insists it plans to take things slowly with AI development and invest the time in building responsibly. Its announcements at Google I/O 2023 are in line with this.
You can watch the full keynote presentation from Google I/O 2023 on Google’s YouTube page. Alternatively, you can watch Cathy Edwards’ (VP Engineering of Google) session discussing the future of AI search. Google also uploaded a written summary to The Keyword blog.
The two main points Google highlights are the use of generative AI in search and using the technology to help shoppers choose and buy products online. Having watched the keynote in full, though, a couple of other details caught our eyes.
For us, the key takeaways are:
Rather than completely overhaul the search experience or follow OpenAI down the chatbot route, Google is carefully implementing AI technology to improve search. Now, let’s take a closer look at how it plans to do this.
The most obvious implementation of AI into Google’s new search experience is the use of generative AI for informational searches. For relevant queries, Google will generate an in-depth response from content around the web. This provides a more comprehensive answer to user questions without them needing to visit multiple web pages. In fact, they don’t need to visit any pages at all.
“With new generative AI capabilities in Search, we’re now taking more of the work out of searching, so you’ll be able to understand a topic faster, uncover new viewpoints and insights, and get things done more easily.” – Supercharging Search with generative AI; The Keyword
Generative AI responses won’t show for all queries – only in cases where Google’s algorithm calculates they’ll add value. This will be most useful for complex queries that can only be answered with information from multiple sources.
Cathy Edwards used this example during their keynote session: “what’s better for a family with kids under 3 and a dog, bryce canyon or arches”. That’s a complex query for any search engine to answer with a list of web pages. It’s also too niche a request for any single piece of content to realistically cater for. But generative AI can pull information from multiple sources and compile it into a single response.
Crucially, Google attributes information with links to the website it’s pulling info from. This is important because any generative AI response can potentially take traffic away from websites, which is something Google will have to address as it develops its AI integration.
Google will never satisfy everyone with its implementation but we’re glad it’s taking obvious steps to address the issue of zero-click searches at this early stage of experimentation.
At the top-right of the generative AI box, users can click on the “bear claw” icon to a complete list of where Google got the information from for its response. Users can also click through to pages shown in the carousel to the right of the response itself.
At the bottom of the response box, users can also click on “Ask a follow up” to apply another query to the session. This doesn’t start a new search but adds the follow-up to the existing query – a key detail we think Google could have made a bigger deal of during the keynote. We’ll explain why we think this is a big deal in a moment.
First, though, let’s take a moment to discuss something else Google breezed over during its keynote. Going back to the example query in the previous section, there’s no way Google’s current search system could handle such a complex query. A quick Google search provides all the proof we need.
If we copy and paste the same query into Google, it matches the two national parks in question but completely misses the intent and conditions of the query. It also matches the query word for word with articles referencing the query and Google’s keynote presentation.
The point is, Google’s existing search experience isn’t designed to handle such complex queries. Even if it accurately interprets the intent of the query – and all of the conditions included in it – the search experience can only return a list of results. It can’t provide a comprehensive answer by its own accord.
In the current search experience, users would have to type in multiple separate queries, essentially starting a new session every time.
“Normally, you might break this one question down into smaller ones, sort through the vast information available, and start to piece things together yourself.” – Supercharging Search with generative AI; The Keyword
Then, users would have to piece the information together from multiple searches themselves to get the answer they’re looking for.
Implementing generative AI into search changes all of this, allowing Google to pull information together from multiple sources and provide a complete answer. For users, this means they’ll get complete answers to complex queries more often – without having to perform multiple searches.
In cases where Google can’t provide everything they’re looking for in a single response, users can add follow-up queries to the session. So, instead of starting a new search, they’re adding more conditions to the existing search and Google will apply these to the results, increasing relevance every step of the way.
Google’s new AI search remembers your queries and follow-ups, allowing you to explore topics and refine results. So, instead of starting with a new search every time you want something more specific, you can simply ask a follow-up and Google will apply this to your results.
“Context will be carried over from question to question, to help you more naturally continue your exploration. You’ll also find helpful jumping-off points to web content and a range of perspectives that you can dig into.” – Supercharging Search with generative AI; The Keyword
Let’s say you’re looking to buy some home gym equipment and you start with a query like “best home gym setup for a small room”. Google might generate some advice from different websites for this query, including specific product recommendations and things to consider. But what if the results focus more on cardio and you’re more interested in weight lifting and building core muscles? Well, all you have to do is type in a follow-up like “show me setups for weight lifting and building core muscle strength”. Google will take this follow-up and apply it to the previous search, refining its results and showing more relevant content, products, things to consider, etc.
With traditional search, you would have to start a new session (with a new query) and hope Google can deliver relevant results. With the new AI search experience, you can explore topics in a linear process with Google remembering what you’ve already searched for and refining results as you provide more context. Obviously, this creates a more immersive experience but it also helps Google understand complex queries more effectively.
With a relatively simple starter query like “best home gym setup for a small room,” Google can interpret this accurately and bank this in its session memory. Next, when the user adds a follow-up like “show me setups for weight lifting and building core muscle strength,” Google can interpret this and apply it to the information it already has for the session.
This is a much easier task for Google’s AI system than getting one shot with a complex query like “show me the best home gym setups for weightlifting and building core muscle strength in a small room”.
Google has also teased how generative AI can improve the search experience outside of informational queries. We’ll see more examples of this as its SGE experiment progresses but, for now, it’s teasing the potential to help people shop online.
“With generative AI in Search, we can help you understand the full picture when you’re shopping, making even the most considered and complex purchase decisions faster and much easier.” – Supercharging Search with generative AI; The Keyword
If a user is searching for a specific product, such as an electric bike, Google’s generative AI technology can help them choose the best product for their needs. The AI response might include a product description, key things to consider when buying the product inquisition, product recommendations, reviews and other relevant info. For example, the response for eBikes could include advice about battery duration, motor power and other key things to consider.
Users will be able to put more complex queries to Google and relevant results. For example, someone might specify that they’re looking for an eBike capable of handling five-mile commutes with hills.
From this, Google can pick out the parameters for distance and inclination, pulling information from multiple sources to generate a helpful response. It can also tap into its vast Shopping Graph to match these requirements with product specifications.
With more than 35 billion product listings, this is the biggest product data graph of its kind in the world. Google says it’s also seeing 1.8 billion updates every hour, making this the most reliable and up-to-date source of product information around. This will be an important differentiator for Google as it faces competition from other AI innovators.
With ChatGPT dominating the AI conversation and Bing quickly bringing an AI-powered search engine to market, some have accused Google of falling behind in the biggest technology race of our time.
Google certainly hasn’t rushed its response to generative AI going mainstream – at least, not as much as some. In fact, the biggest name in search is taking a relatively measured approach to implementing AI into the search experience. It’s also not blindly following the likes of OpenAI or the companies implementing ChatGPT into their products as soon as they can.
Google explicitly says it doesn’t want the search experience to feel like a conversation with a human. It wants the experience to provide relevant, accurate information, not simply present information in a convincing way.
It’s hard to imagine Google wasn’t shaken by the recent ChatGPT noise, at least on some level. Its rushed Bard release showed all the signs of a tech giant spooked by the meteoric rise of a new rival. Yet the Google I/O 2023 keynote brings back the calm, composed tone we’re used to seeing from them.
As we’ve explained before, ChatGPT is not a search engine nor a real threat to Google. That is, unless Google gets distracted by the hype surrounding OpenAI’s products and forgets its place in the market: search. The Google I/O 2023 keynote is the best indication so far that it’s not losing sight of this.
Google has shown glimpses of what its generative AI technology could bring to the future of search. At the same time, it keeps reminding us that the Search Generative Experience is an experiment. Obviously, Google has put plenty of development into the system already but a lot could change before any of this technology is implemented into everyday search.
It also leaves a lot of questions unanswered and we’ll almost certainly have a long before getting any answers:
We’ll have to wait until users in the US have had more time to experiment with SGE before we can start to answer any of these questions. Hopefully, Google will open the experiment to users outside of the US soon and we can start to test the system for ourselves, but we’ll have to wait and see.
It seems Google is in no rush to implement AI features that don’t benefit the search experience. It also needs to find the right balance of providing users with information while still making the platform attractive to publishers and advertisers.
Google has already seen how much damage rushed implementations can cause with its Bard demonstration. It’s also seen that – as the leader in its space – it can’t afford to make the same mistakes smaller competitors might get away with. Slow, calculated moves are the search giant’s best strategy and it seems to know this.
As soon as Google expands the Search Generative Experience to users in the UK, we’ll be signing up to test the system for ourselves. We recommend every marketer does the same to start getting first-hand experience of what the future of AI search might look like.
In the meantime, if you’re struggling to optimise for the latest Google Search innovations, our SEO services team can help. Call us on 02392 830 281 or send us your details and we’ll get right back to you.
Dave is head of SEO at Vertical Leap. He joined in 2010 as an SEO specialist and prior to that worked with international companies delivering successful search marketing campaigns. Dave works with many of our largest customers spanning many household names and global brands such as P&O Cruises and Harvester. Outside of work, Dave previously spent many years providing charity work as a Sergeant under the Royal Air Force Reserves in the Air Cadets sharing his passion for aviation with young minds. He can often be found in the skies above the south coast enjoying his private pilot licence.
Looking for evidence-led search marketing expertise?
Categories: Content Marketing, SEO