Google has begun the gradual rollout of its Search Generative Experience, as announced at I/O 2023. It's essentially a collection of AI-powered features that promises to change the way we search online.

The idea is to make Google Search smarter, faster, and better in dealing with complex queries in a way that helps you navigate the sprawling, often chaotic world of online information. Since we’ve all been speculating about it for weeks, we thought it was time we gave it a spin and shared our first impressions.

How to Access Google's Search Generative Experience

After waiting patiently for days, we finally got the email: “It’s your turn to try Search Labs.” For now, Google’s AI Search Generative Experience is only available on your Chrome desktop browser, on the Google mobile app, and only to select users who reside in the United States.

Only US residents who have previously signed up for Google's SGE waitlist are invited to try out the new AI search features. If you want to try out the feature before it rolls out to the public, here's a guide on how to join the SGE waitlist.

If you've previously joined the waitlist, you should make sure you know how to turn on and use Search Generative Experience. We've been using SGE for some days now, and here's what we think about the new AI-powered search features.

How Accurate Are Google's AI Search Results?

Since SGE is all about AI-generated results for search queries, one of our first concerns was how accurate the generated information would be. Remember, AI chatbots are infamous for making up information. To test the accuracy of SGE results, we started with a simple task: googling the definition of "longtermism."

Although Google's AI Search results look similar to responses from its standalone Bard AI, the two are quite different. While Bard, like most AI chatbots, aims to primarily create "original content," Google's SGE does what you'd expect a search engine to do—aggregate content. The majority of search results from Google's Search Generative Experience is a near word-for-word copy of selected sources.

AI snapshot of longtermism definition-1

In our googling of "longtermism" above, the paragraphs labeled one and two above were taken almost word-for-word from Wikipedia and the William MacAskill webpage. Below is an excerpt taken from Wikipedia:

Longtermism - Wikipedia

And here is another excerpt taken from the William MacAskill webpage:

Longtermism — William MacAskill screenshot

Content from both sites (and two others) was stitched together to form the AI-generated snippet. This implies that Google's new AI search feature doesn't face as much of an issue with accuracy, given that it doesn't produce content in the same manner as chatbots like ChatGPT and Google Bard.

Information that is displayed on the AI snapshot panel is sourced from websites that Google already considers as "reliable" and "generally accurate." Doing things this way helps Google avoid the costly mistake of AI hallucination that comes with the bulk of generative AI. It ensures that the text on Google's AI Search result panel is mostly if not completely human-generated and fact-based information.

What's Different in Google's AI Search Generative Experience?

How is the typical Google Search experience different from the AI Search Generative Experience? Well, Google has given a lot of the SGE interface a makeover. One such makeover is the color of the links. While the current Google Search result pages display blue links, SGE uses white links for the dark theme and black links for the light theme.

sample of SGE search (light mode)

However, beyond the aesthetic, there are several other differences in the way SGE results are presented. Key among them is the fact that results are more wholesome and smart. We googled "TV shows similar to Shadow and Bone."

Here's the regular Google result. Google uses the regular featured snippet to highlight the suggestions from the highest-ranking source.

Tv shows like shadow and bone1

And here's the SGE result. Here, Google picked some of the most popular shows across a variety of sources.

TV shows like shadow and bone

We also googled "What are the best used phones to get under $500?" The regular Google search returns a list of websites, most of which do not answer the question. We specifically asked for used phones under $500, with emphasis on "used." However, only one of the ten results talked about used phones.

Best used phones under 500 dollars (Standard Google Search)

We used the same query on Google's AI Search, and the results from the Generative AI panel were more precise and useful. It started with advice on how to choose a used phone, followed by a curated list of used phones under $500 from trusted vendors across the web.

Best used phones under 500 dollars

Google's SGE is not the typical Bard-like AI product many observers expected it to be. It doesn't generate information in the way we expect from generative AI tools. That might sound disappointing at first, but it's a decision to ensure accuracy of generated results.

Although Google has branded this as an entirely new feature, it is similar to Google's featured snippets that we've already been using for years. Let's put it this way; Google's SGE is mostly an upgraded version of featured snippets with the additional ability of maintaining context and aggregating results from multiple web sources rather than one.

It might be subtle for now, but Google's SGE signals a change in Google Search's core function. While Google Search has always been primarily a tool that aggregates links and directs users to resources, SGE means Google itself is becoming both an aggregator and resource.

In other words, Google is making it such that you can complete your search intent without ever clicking through to any website on the result pages by bringing the information you need to the result pages. This could change how we use websites on the internet.