SGE Evolves

While OpenAI garners much of the spotlight, Google has been active in the generative AI arena long before the startup emerged.

Initially, Google approached generative AI cautiously, developing robust systems like DeepMind without making them publicly available.

The rise of OpenAI and ChatGPT, however, compelled Google to respond. To maintain its relevance, the tech giant introduced Gemini (formerly known as Bard), a chatbot powered by its own Large Language Model (LLM).

Additionally, Google began experimenting with a more advanced offering.

Initially called the Search Generative Experience (CGE), this new tool utilized generative AI to provide summaries of web pages and deliver information directly at the top of search results.

Rather than the traditional “10 blue links”—or other features like visual snippets or YouTube videos—Search Generative Experience aimed to answer users’ questions outright, eliminating the need for extensive searching.

During the early testing phase of Search Generative Experience, I found that despite initial criticisms about its rough edges, the potential for transformative change was evident.

The Concept of "Satisficing"

The new feature exemplifies the psychological concept of “satisficing,” which refers to the inclination to seek a satisfactory solution rather than the optimal one.

When users input queries into Google, the SGE responses tend to be average at best.

However, many users may find these responses sufficient enough to avoid the 2 to 3 minutes of additional research typically required to find the best answer through multiple traditional websites.

It appears my assessment holds true, as Google reports that SGE has been utilized “billions of times” even during its beta testing. While users may not universally embrace it, their behavior indicates a preference for its convenience.

Moreover, SGE ultimately benefits Google financially. By retaining users on its platform rather than directing them to external sites, Google can continue to display profitable search advertisements.

SGE for All Users

Throughout much of last year and into early 2024, Google has been refining SGE through beta testing. At their recent I/O event, the feature was announced for a wider audience under the name AI Overviews.

Now, when users perform a Google search, they will receive an AI-generated summary for numerous queries, whether they ask for it or not.

This represents a monumental shift in how people navigate the Internet. With Google commanding over 90% of the search market, this means they facilitate at least 1.5 trillion searches annually.

Thus, for trillions of inquiries made online, Google becomes the gatekeeper of the information provided.

With the introduction of AI Overviews, that gate is now managed by a generative AI Large Language Model.

If Google successfully rolls out AI Overviews and maintains them, a significant portion of the information accessed by users—let's be honest, this encompasses most of the information people encounter—will either be filtered through or generated by AI.

Google anticipates that “hundreds of millions of users will gain access to AI Overviews this week, with a goal of reaching over a billion by year’s end.”

This extensive reach positions AI Overviews as the largest and most significant application of generative AI to date.

While ChatGPT has made waves and amassed a vast user base, it still reaches fewer than 200 million people.

In contrast, AI Overviews will expose the global population to the outputs of Google's generative AI systems, regardless of their preferences.

The Backlash

Not everyone is pleased with this development.

Leading up to the launch of AI Overviews, Google implemented several algorithm updates that have severely impacted numerous independent blogs.

Many travel, food, and local news blogs have seen traffic declines of 80 to 90% since last September.

Although Google claims that AI Overviews will link prominently to publishers’ websites, it remains uncertain how many users will actually click through to these links instead of obtaining answers directly from the AI.

Concerns about the accuracy of AI Overviews have also been raised, especially regarding sensitive medical inquiries or other critical issues, where inaccuracies could have serious repercussions.

For instance, when a user inquired about cleaning their washing machine, AI Overviews reportedly suggested mixing bleach and vinegar—a combination that can release toxic chlorine gas.

At least the AI acknowledged that this mixture could be hazardous!

There are also ongoing discussions about the sources of data used to train AI Overviews.

Many in the publishing sector are questioning how Google can legally utilize information from sites that the AI might potentially replace.

Addressing these concerns and understanding the broader impact of AI Overviews will require time.

What is certain, however, is that with the launch of such a widely accessible AI feature, Google is unleashing a potent technology that could disrupt business models, reshape the way information is consumed, and spark intense debates regarding the benefits and drawbacks of generative AI.

Will It Be Accepted?

Many creators might still hold onto the hope that this trend is merely temporary.

Perhaps users will reject the new AI Overviews, demand their removal, and push Google back to the traditional format of “10 blue links.”

As a web publisher, I would welcome that! Yet, I also recognize that such a reversal is improbable.

While AI Overviews may not be flawless and experts can certainly critique them, the legality of generative AI training will be hashed out through lawsuits and private licensing agreements in the coming years.

In the meantime, users will undoubtedly engage with AI Overviews because people have limited time, often prioritize speed over accuracy, and are inclined to accept quick solutions, even if they are not perfect.

For more intricate topics, such as planning a substantial trip or addressing a medical concern, users may be more inclined to explore multiple blogs for better information than what AI Overviews can offer.

However, for simpler inquiries, they likely will not.

What Should Creators Do?

As content creators, it’s crucial to acknowledge this new reality.

If your content is susceptible to being replaced by an AI Overview response, it ultimately will be.

Only content that serves a distinct purpose— or that cannot be easily replicated by generative AI—will thrive in this new environment.

As consumers, we must also recognize this shift and the potential implications for how we acquire information.

AI Overviews remain a largely experimental feature. While they may be reaching a vast audience, the road to maturity will be complex and lengthy.

Google might temporarily retract AI Overviews if the system encounters issues, similar to their previous actions with parts of Bard’s image generator.

They could face regulatory or legal challenges as well, leading to public backlash.

Throughout this process, it’s essential to remember that these AI Overviews are ultimately machine-generated—despite their remarkably human-like qualities.

For straightforward queries, it may be acceptable to trust the AI’s output. For instance, if I want to know the capital of Wyoming, I could likely rely on Google’s AI response.

However, for medical inquiries or anything related to health and safety, I should absolutely not rely on information from any AI system.

I should investigate further, seek authoritative sources, consider various perspectives, and look for well-referenced information to address critical concerns.

In summary, the launch of a powerful feature like AI Overviews necessitates that we all improve our media literacy skills.

With the prevalence of deep fakes and biased news, we’ve become adept at identifying human attempts to deceive or mislead us.

Now, we must enhance our ability to recognize the benign inaccuracies produced by machines.

If you use Google (and most of us do), educate yourself about how AI Overviews function. Understand the technology's capabilities and limitations, and determine for yourself when it's acceptable to trust an AI response and when further exploration is necessary.

Additionally, instill these lessons in younger generations. It's never too early to discuss media literacy and teach children to identify misleading and inaccurate online information.

Currently, there is no straightforward way to disable AI Overviews. Even if you’ve managed to avoid generative AI thus far, this technology is rapidly becoming ubiquitous.

It’s time to acknowledge its presence, grasp its implications, and decide what we, as consumers and creators, want from this technology—and what we will not accept.

I’ve experimented with thousands of ChatGPT prompts over the past year. As a dedicated creator, there are a select few I return to daily that align with the ethical applications discussed in this article. I’ve compiled them into a free guide, *7 Enormously Useful ChatGPT Prompts For Creators*. Download your copy today!