Google's Search VP on Generative AI in Search, User Notes on Search Results, and SEOs — With Cathy Edwards

Episode Summary

Executive Summary: Google Search VP Kathy Edwards said Google’s generative search experience is designed to drive users to the open web, not trap them in Google, and that early lab data shows more links, more diverse clicks, and stronger engagement with long-tail content. She also unveiled two new features: Follow, for tracking topics and queries, and Notes, a human annotation layer aimed at making search more interactive while preserving web health.

Main Topics: Generative Search Experience and the Open Web (Priority: 5/5): Edwards argued SGE is meant as a jumping-off point to other sites, emphasizing more links, diverse clicks, and long-tail discovery, though she acknowledged the data is still early and opt-in. Monetization, Ads, and Product Readiness (Priority: 4/5): The conversation explored whether ads, monetization, and speed will determine rollout. Edwards said ads are separated from the product discussion and that latency remains a major reason SGE is still opt-in. Publisher Control and AI Training Data (Priority: 5/5): A substantial segment focused on the value exchange between publishers and AI companies. Edwards highlighted robots.txt control, Googlebot extended controls, and the principle that publishers should choose how their content is used. New Search Features: Follow and Notes (Priority: 5/5): Google announced Follow, which lets users subscribe to search queries/interests, and Notes, which lets users add contextual annotations directly under results, initially in the US and India/labs. Moderation, Civility, and Community Context (Priority: 4/5): Notes raised moderation questions similar to YouTube comments and Twitter/X community notes. Edwards said Google will combine algorithmic and human moderation and avoid risky queries such as highly volatile news topics. SEO, Spam, and AI-Generated Content (Priority: 4/5): The interview addressed concerns that AI will worsen SEO spam. Edwards said Google already fights made-for-search content aggressively and believes generative AI content can be good if it serves humans rather than search engines. Search vs. Social and Assistive Future (Priority: 3/5): The discussion closed by comparing Notes/Follow/Perspectives to social feeds and assistant-like experiences. Edwards insisted Google is not building a social network, but a more proactive information product.

Key Arguments: SGE is designed to increase web exploration, not replace the open web; Edwards said Google is seeing more links and more diverse clicks. Because SGE is opt-in, Google cannot run a clean large-scale A/B test yet, so conclusions about net click-out behavior remain preliminary. Users are showing higher sentiment and asking more complex queries, suggesting generative search helps people ask questions they previously avoided. Publisher control matters; Google says robots.txt and the new Googlebot extended controls let site owners decide whether their content is used for generative AI training or APIs. Google argues there is not yet a multibillion-dollar generative AI revenue business, so the value-exchange debate is still early. Notes is meant to add a human layer to search rather than become a fact-checking or social network product; it aims to improve context and freshness. Google believes it can moderate Notes using a blend of algorithmic rules and human review, drawing on lessons from YouTube and its spam-fighting teams. AI-generated content is not inherently bad; Google says its systems are designed to reward content for humans and discard spam before indexing. Follow and Perspectives reflect a broader shift toward proactive, interest-based discovery, but Google frames this as information utility rather than social media competition.

Data Points: SGE rollout status: opt-in / labs only - Edwards repeatedly said Google is still testing and has not rolled SGE out broadly because it remains early and slow. Rollout timing for Follow: fully by December - Edwards said Follow would begin rolling out on the day the episode airs and reach all users by December. Launch regions for Notes: U.S. and India - Google said Notes will launch first in labs in these two markets. Assistant update: Bard into the assistant soon - Edwards said Google announced bringing Bard into the Assistant, with more details to come. Open web signals: more links and a new diversity of clicks - Edwards described the early SGE data as encouraging for web traffic diversity. Cricket widgets example: weather, stocks, cricket widgets - Edwards used Discover widgets as evidence that users like staying on top of topics without repeated queries. Instagram engaged hours per day: 1.1 billion - Alex cited Sensor Tower data to argue social platforms still have strong demand. Instagram growth year over year: 29% - Alex used this number to show Instagram’s continued strength. X/Twitter engaged hours per day: down 8% - Alex cited Sensor Tower data to suggest social media may be opening a vacuum for Google to fill. OpenAI annualized run rate: $1.3 billion - Edwards referenced OpenAI’s revenue scale while noting that broad generative AI monetization is still early. Search history: 25 years plus - Edwards cited Googlebot’s long history and Google’s 25-year search legacy when discussing control and moderation.

Pivotal Quotes: "We really want to make sure that as we're bringing generative AI into search, that we're really thinking about the health of the open web." — Kathy Edwards: Explaining Google’s philosophy for Search Generative Experience and its relationship to publishers. "We're really focused on building a really great product for users and making sure that the web ecosystem is really healthy." — Kathy Edwards: A response to questions about ads and monetization in generative search. "It is to organize the world's information and make it universally accessible." — Kathy Edwards: Describing Google Search’s core mission while discussing the future of search and assistive features.

Implications: Google is trying to add AI, community input, and personalization without breaking search’s link-driven web model. For publishers, the key issues are control, moderation, and whether these features increase or dilute traffic and value.