Joe Weisenthal and Tracy Alloway host Liz Reid, VP of Search at Google, who has been with the company for over 20 years and currently leads the product, engineering, design, and data science teams building Google Search.

The conversation explores how Google is navigating the tension between its traditional click-based search business and the rise of AI overviews that provide direct answers without requiring users to click through to websites.

Reid discusses user behavior patterns across Google Search, AI mode, and the Gemini app, explaining how different query types drive users to different interfaces and how AI is expanding rather than cannibalizing search volume.

The discussion covers practical challenges from AI-generated content quality to software engineering hiring practices, while examining Google's strategy for maintaining its search dominance in an AI-transformed landscape.

When Google Shows AI Overviews vs Traditional Results

Google determines AI overview placement based on user signals indicating added value, not simple rules like question marks - 'We shouldn't give you AI for the sake of giving you AI' - Liz

Simple queries like 'Corgi' show traditional links while 'What is a Corgi?' triggers AI overviews because users seek descriptions rather than just navigation to specific sites

The system learns over time when AI overviews provide value versus when users prefer direct access to websites like Wikipedia or specific podcasts

User Behavior Across Google's AI Interfaces

Users choose between Google Search, AI mode, and Gemini based on query complexity - informational queries favor Search/AI mode while creative/productivity tasks go to Gemini

AI mode users typically ask longer, more complex questions expecting follow-ups, while traditional search handles quick lookups and browsing queries

Google observes users fact-checking LLM responses through search - 'We're definitely aware that people use Google as a fact checker for some of their LM use cases' - Liz

The Evolution from Keywords to Natural Language

Users are moving from keyword searches to expressing actual needs - instead of 'restaurants New York' they now specify 'restaurant for five people, not too pricey, with vegan options and kid-friendly'

The falafel example illustrates query ambiguity - some users want definitions, others recipes, nutrition info, or restaurant locations, all using the same single word

Natural language queries enable better personalization and reduce the 'translation' burden on users to convert their real problems into computer-friendly keywords

AI's Expansionary Effect on Search Volume

Google measures success by whether users 'hire Google more often' - returning to search more frequently rather than just using it more per session

AI lowers barriers to asking questions people previously deemed not worth the time investment - 'You actually make a calculation when questions go through your mind of is it worth spending any time to figure out the answer' - Liz

The technology enables curiosity similar to children asking 'why?' repeatedly, but adults previously self-censored due to time constraints and uncertainty about finding answers

Multilingual capabilities unlock content for non-English speakers where web corpus may be limited in their native languages

Monetization Strategy for AI-Enhanced Search

Google shows ads on less than 25% of queries, so many AI overview queries wouldn't have generated ad revenue anyway - like the tankers in Strait of Malacca example

Commercial queries still require transactions - 'The answer doesn't buy the pair of shoes. You actually have to buy the shoes' - Liz, maintaining opportunities for shopping ads

More detailed natural language queries provide better targeting opportunities for advertisers compared to single-keyword searches

New ad formats may emerge as technology evolves, similar to how Instagram created feed-based advertising that wasn't initially obvious

Combating AI-Generated Content and Quality Control

Reid emphasizes that content quality issues predate AI - 'There has always been slop on the web' including human-generated spam and low-quality content

Google's approach focuses on surfacing quality content rather than eliminating all poor content - crawling far more pages than it indexes and indexing more than it surfaces

The company's 'bread and butter' ranking systems have long experience filtering spam, with financial incentives driving continuous evolution of both spam and anti-spam measures

Future Interface Evolution and Multiple Entry Points

Reid expects continued proliferation of form factors rather than convergence - phones supplemented but didn't replace desktops, watches supplemented but didn't replace phones

Different interfaces excel at different tasks - chat interfaces work well for some queries but are inefficient for others like removing specific items from lists

Google already operates multiple search boxes (YouTube, Maps, Chrome, Google app) serving different user preferences and use cases without full convergence

Future development will focus on reducing friction and making interfaces more adaptive, personal, and ambient rather than forcing everything into a single entry point