Google is taking big risks with AI search

Key Takeaways

• When Google integrates generative AI into search, the problem of false answers arises.
• Google risks reputational damage from the AI ​​Overview, as erroneous results could undermine user trust.
• The competition between ChatGPT Search and Microsoft could impact Google's AI strategy.

It's been a while since Google announced its biggest search changes in years, leveraging new generative AI tools to provide an overview of the topics being searched for, eliminating the need to click through result links to find the information you need. In a way, maybe. The integration of the new AI capabilities hasn't always been smooth sailing, quickly causing all sorts of problems with incorrect answers, ranging from irrational to potentially dangerous.

Google has paid a heavy price for its changes to search, which seems to reflect the pressure the company is now feeling from both Microsoft, ChatGPT and Apple as it strives to improve AI integration into its products.

What is generative AI search?

Announced at Google I/O in May, AI Overview is a previously experimental feature of Google Search that leverages generative AI to answer search terms and questions. Over the past few weeks, it has been rolled out across the U.S. with wildly mixed results. Google has reportedly relied heavily on an agreement it struck with Reddit to use content for its AI-generated datasets, and users have reported that some of the generated answers are solely sourced in individual comments on Reddit, along with a host of other inaccurate, misleading, or just downright ridiculous answers to simple questions.

Google is taking a big risk

Integrating generative AI into search increases reputation risk

Most of the AI ​​features in its core revenue-generating products have been “nice to have” features — add-ons to existing basic services. Google is taking this a step further by prioritizing AI responses in search results, signaling as clearly as possible its intention to put AI front and center, even at the risk of damaging the reputation and integrity of its most fundamental products.

This seems risky at the very least. Gemini has struggled to compete with ChatGPT, arguably the first serious search competitor to reach the masses since Bing. But both ChatGPT and Gemini have had their problems, and until now, the reputational damage from those shortcomings has been limited to those products. Now, the integration of direct AI responses into Google Search, the default search tool in nearly every browser on the planet, brings more eyes, ears, attention, and opportunities for unpleasant hallucinations.

Apple has adopted a similar approach, using Apple Intelligence to integrate both its own models and ChatGPT into iOS, but these are useful additions to an existing OS and unlikely to drive users off the platform to a competitor.

False results are negatively impacting the overall search experience

Search is perhaps the product with the greatest risk to Google's reputation. Previously, Google search results provided links to information at a basic level, directing users to summaries of sites and other users' content. But now, generative AI search is in charge of not just displaying information, but summarizing and sometimes editing it, and we are already seeing a variety of results. After the rollout of AI summaries, Google was forced to make changes to improve results and uncover the causes of horribly incorrect summaries.

There's also the issue of audience. Google receives over 80 billion visits per year and accounts for over 90% of the total search engine market. Gemini and ChatGPT have garnered some traction, but the volume of queries, edge cases, and potential damage is growing exponentially. Even iOS AI integration is likely to be used at a scale several orders of magnitude smaller than Google Search.

The combination of these two factors poses a legitimate risk to the reputation of Google Search: False or misleading results could be quickly made public, dealing a blow to the credibility of Google Search, as well as damaging broader consumer perceptions of AI and raising doubts about the trustworthiness of AI-generated results.

ChatGPT search is rumored to be coming

Microsoft is once again Google's biggest rival

The problem at issue here is the rumored ChatGPT search. As mentioned earlier, ChatGPT has moved beyond the tech realm and into our everyday lives. It's used for everything from programming to data processing to writing cover letters. There are even rumors that it's developing its own search product.

If ChatGPT becomes popular and the company gets widespread coverage of its new search-only product, Google risks getting a stumped as it integrates its products more broadly. It may also need to consider ChatGPT's “brand new” effect: ChatGPT may be perceived as more cutting-edge than equally old Google products by a significant percentage of users.

Integrating search now could be Google's attempt to get ahead of the narrative around other AI-powered search alternatives, even if it has to overcome some initial issues during rollout. It's also a good opportunity to get the broader Google user base (since more people than just tech bloggers use Google) to understand the power of the AI ​​assistant and provide tangible, real improvements to the broader search experience (if it works well).

Whatever the outcome, Google is acting boldly.

It will be hard to judge how effective integrating generative AI into search will be, especially after such a shaky start. But one thing is clear: Google is taking decisive and bold action on AI, not only investing in developing a best-in-class LLM, but also firmly positioning AI at the forefront of its most important products. It's a big risk, but it may be what's needed to thwart the ChatGPT threat. Google hasn't faced the scale of this threat since Bing tried and failed to compete in the early 2010s.