## Background
Published on YouTube by Ai4 on August 22, 2026, this panel discussion features Yahoo News GM Kat Downs Mulder, Yahoo Research SVP & GM Eric Feng, and Fast Company Global Technology Editor Harry McCracken. The session examines Yahoo's evolution from traditional keyword search to conversational AI synthesis while preserving the open web and supporting publisher ecosystems.
## TL;DR
Yahoo leadership discusses the shift from traditional search indexing to conversational AI synthesis, highlighting the rollout of Scout, Yahoo's dedicated AI answer engine. Operating as both a top search engine and a major digital publisher, Yahoo advocates for sustainable media economics through multi-channel monetization, citation transparency, and licensing frameworks like Really Simple Licensing. Drawing on a 30-year legacy, 500 million user profiles, and 300 billion daily behavioral signals, the company balances ambient content feeds with interactive deep-dive queries. Ultimately, the speakers emphasize that high-quality LLM performance inherently depends on a healthy publisher ecosystem, maintaining Yahoo's enduring commitment to trust.
## Atmosphere & style
The conversation carries a measured, authoritative, and collegial tone typical of a high-level industry keynote. Harry McCracken's veteran journalistic moderation pairs naturally with Mulder and Feng's pragmatic, insider perspectives on media sustainability and AI engineering. The discussion moves thoughtfully from technical infrastructure and media economics to philosophical reflections on enduring platform trust, striking a balanced and optimistic vibe without glossing over industry challenges.
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=4s) Yahoo reaches roughly 9 out of 10 Americans across its network, including 175 million monthly active users on Yahoo News alone.
- [00:04](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=4s) User discovery habits are transitioning from fragmented keyword queries and ten blue links toward conversational, AI-synthesized answers powered by engines like Scout.
- [02:05](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=125s) Operating simultaneously as a major search engine and one of the world's largest content publishers gives Yahoo a unique mandate to protect the open web from AI cannibalization.
- [02:05](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=125s) Maintaining publisher traffic and monetization is essential for AI health, as large language models degrade without continuous, high-quality primary reporting.
- [02:05](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=125s) Scout integrates prominent sourcing and direct citations inside its answers rather than relegating original content creators to background footnotes.
- [05:00](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=300s) Adapting lessons from film and streaming distribution, Yahoo is pioneering multi-channel AI monetization by joining Microsoft's publisher content marketplace and adopting the Really Simple Licensing (RSL) standard.
- [07:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=439s) Yahoo's retrieval and personalization models leverage a 30-year technical foundation encompassing 500 million profiled users and 300 billion daily behavioral signals.
- [07:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=439s) Information consumption spans both ambient 'push' discovery across homepages and alerts, as well as intentional 'pull' exploration through conversational answer engines.
- [07:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=439s) Scout maintains search neutrality, querying both open web publishers and Yahoo's native properties strictly on content quality rather than self-preferencing.
- [11:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=679s) Internal telemetry demonstrates that premium publisher reporting is cited by LLMs at roughly twice the rate of generic web content due to reasoning quality demands.
- [17:51](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=1071s) While interface paradigms have evolved from directories to keywords to AI synthesis, Yahoo's core mission remains providing a trusted guide to the open internet.
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=4s) Yahoo's Reach and the Shift from Search to Synthesis
Kat Downs Mulder, GM of Yahoo News, and Eric Feng, SVP and GM of Yahoo Research and Search, join Fast Company's Harry McCracken to discuss the evolution of information discovery. Setting the stage, Kat outlines the sheer scale of the Yahoo ecosystem—spanning Yahoo.com, Yahoo News, Yahoo Games, and Yahoo Weather. In an increasingly chaotic digital media landscape rife with misinformation, Yahoo's primary focus remains delivering high-quality, verified information to an enormous user base.
Eric introduces Yahoo's search footprint and its latest artificial intelligence initiative. While Yahoo Search remains the third-largest consumer search engine in the world, the launch of Scout—Yahoo’s dedicated AI answer engine—marks a pivotal transition in user behavior. Instead of typing fragmented keywords and expecting ten blue links, modern users increasingly ask conversational questions and expect comprehensive, synthesized answers.
```mermaid
flowchart LR
accTitle: Evolution of Search to Synthesis
A["Traditional Search<br/>(Keywords)"] --> B["10 Blue Links<br/>(Manual Navigation)"]
C["AI Answer Engine / Scout<br/>(Full-Sentence Queries)"] --> D["Synthesized Answers<br/>(Contextual Sourcing)"]
```
- **Monthly Reach**: Approximately 175 million monthly active users for Yahoo News alone, and over 250 million across the broader Yahoo network.
- **Audience Saturation**: Reaching roughly 9 out of 10 Americans—which Kat humorously notes is "more than people who put ketchup on their hamburgers."
- **Market Position**: Yahoo Search remains the #3 search engine globally as Yahoo rolls out Scout to address full-question retrieval.
---
[02:05](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=125s) Sustainable AI, Sourcing, and Publisher Relationships
Addressing how Scout distinguishes itself from conversational chatbots like ChatGPT, Gemini, Copilot, or Claude, Eric explains that Yahoo operates from a rare vantage point: it is simultaneously one of the largest web search engines and one of the world’s biggest original content publishers across news, sports, and finance. This dual role creates a unique responsibility to build an AI ecosystem that protects the open web rather than cannibalizing it. If publishers stop seeing value and traffic degrades, the underlying knowledge and reasoning quality of large language models (LLMs) will inevitably collapse as well.
Kat emphasizes that Yahoo's heritage as an aggregator is rooted in mutual financial sustainability. Yahoo maintains syndication and revenue-sharing partnerships with hundreds of global publishers, funding original journalism while bringing high-quality reporting to broader audiences. Scout is engineered to reflect this ethos by prominently featuring original sourcing and primary citations within the product experience, avoiding the trap of treating content creators as mere background footnotes.
- **Dual-Sided Perspective**: Yahoo balances the incentives of a major publisher (Yahoo Finance, Yahoo Sports, Yahoo News) with those of an open search engine.
- **Ecosystem Health**: Protecting publisher traffic ensures the continuous generation of high-quality data necessary to keep LLMs accurate and capable.
- **Attribution Design**: Scout integrates prominent sourcing and direct links directly within generated responses.
---
[05:00](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=300s) Multimodal Discovery and New Licensing Models
Content discovery in the AI era cannot rely on a single interface. As Eric notes, user habits are inherently multimodal: some individuals bookmark and visit destinations directly, others search via traditional navigation queries, and many now seek automated synthesis. Meeting users across these modalities requires pioneering new economic structures so that publishers are fairly compensated when their intellectual property powers AI engines.
Drawing on his background in online video at Hulu, Eric compares the current state of web publishing to the evolution of television and film distribution. Just as video sustainability required a combination of theatrical windows, pay TV, streaming platforms, and digital sales, online publishing must embrace varied monetization models. Yahoo has committed to this multi-pronged approach by joining the Microsoft publisher content marketplace and adopting the Really Simple Licensing (RSL) standard.
```mermaid
flowchart TD
accTitle: Publisher Content Monetization Framework
A["Original Publisher Content"] --> B["Direct Web Traffic & Ads"]
A --> C["Syndication & Revshare (Yahoo)"]
A --> D["AI Marketplace (Microsoft)"]
A --> E["Licensing Standards (RSL)"]
B & C & D & E --> F["Sustainable Open Web"]
```
- **Microsoft Publisher Content Marketplace**: Yahoo's partnership aimed at compensating media companies for AI usage of their editorial work.
- **Really Simple Licensing (RSL)**: Standardized framework supported by Yahoo to enable automated web licensing for machine consumption.
- **The Media Precedent**: Lessons from Hulu illustrate that no single monetization channel can sustain content production on its own.
---
[07:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=439s) Massive User Graph and Push vs. Pull Information Consumption
Building modern AI retrieval products on Yahoo’s infrastructure leverages decades of accumulated data assets. Eric highlights that Yahoo operates over a 30-year technical foundation, featuring a user profile graph of over 500 million people and processing 300 billion behavioral signals daily. When answering queries on Scout—such as questions about his favorite basketball team, the San Antonio Spurs—the engine draws from Yahoo Sports alongside the broader open web based strictly on quality and relevance rather than self-preferencing.
Kat expands on the structural difference between "push" and "pull" information needs. When users are in an active, inquisitive state, conversational answer engines like Scout excel at answering specific questions. Conversely, passive morning routines rely on ambient discovery via the homepage, apps, and push notifications. The goal is using the "Scout intelligence layer" across all Yahoo properties to bridge ambient curiosity into deeper, intentional exploration.
- **Scale of Signals**: Over 500 million profiled users and 300 billion daily interaction signals inform Yahoo’s personalization and retrieval models.
- **Search Neutrality**: Scout balances proprietary editorial properties with open web publishers, surfacing whichever source provides the best content.
- **Discovery Modes**:
- *Ambient / Push*: Homepages, alerts, and curated feeds satisfying casual awareness.
- *Inquisitive / Pull*: Scout conversational sessions allowing deep, tailored investigations.
---
[11:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=679s) Aligned Incentives and Yahoo's Core DNA of Trust
Addressing media industry anxiety regarding LLMs scraping editorial content without attribution or referral traffic, Kat and Eric maintain that technological disruption must not be downplayed, but long-term incentives remain fundamentally aligned. Media outlets strive to produce fresh, accurate reporting, while language models require that exact data to remain functional. In fact, internal Scout telemetry shows that premium "head publisher" content is utilized by LLMs roughly twice as often as generic web text.
Reflecting on Yahoo’s journey from Jerry Yang and David Filo’s 1994 directory to 2026, Kat and Eric assert that while interface paradigms evolve from directories to keywords to synthesized responses, the company’s underlying DNA remains unchanged. Yahoo's lasting value proposition is serving as a trusted guide to the open internet, helping users accomplish their goals across news, email, finance, and sports.
- **Citation Disparity**: Premium publisher reporting is referenced by LLMs at twice the rate of generic web content due to reasoning quality demands.
- **Incremental vs. Transformational**: Balancing immediate utility (summaries, digests) with foundational product bets on conversational synthesis.
- **Foundational Mission**: Preserving the core value of human and algorithmic trust established during Yahoo's founding in 1994.
[17:51](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=1071s) Trust in the AI Era and Yahoo's Enduring Mission
As artificial intelligence continues to play a substantially larger role in shaping digital experiences, the critical necessity for a reliable entity to establish and provide trust for both content and consumer experiences will only grow in importance. Reflecting on the company's legacy and forward trajectory, the speakers emphasize that this fundamental commitment to credibility is an element of their core mission that is poised to "endure for the next 30 years just as it had for the previous 30."
Closing out the session, moderator Harry McCracken thanks the Yahoo team for striving to be a "force for good" in navigating the shifting landscape of digital media and AI synthesis. He wraps up by thanking the panel and attendees, wishing everyone an engaging remainder of their time at the Ai4 conference.
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=4s) more than people who put ketchup on their hamburgers. — Kat Downs Mulder
> [02:05](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=125s) Yahoo operates from a rare vantage point: it is simultaneously one of the largest web search engines and one of the world’s biggest original content publishers across news, sports, and finance. — Eric Feng
> [02:05](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=125s) If publishers stop seeing value and traffic degrades, the underlying knowledge and reasoning quality of large language models (LLMs) will inevitably collapse as well. — Eric Feng
> [02:05](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=125s) Yahoo maintains syndication and revenue-sharing partnerships with hundreds of global publishers, funding original journalism while bringing high-quality reporting to broader audiences. — Kat Downs Mulder
> [05:00](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=300s) Just as video sustainability required a combination of theatrical windows, pay TV, streaming platforms, and digital sales, online publishing must embrace varied monetization models. — Eric Feng
> [07:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=439s) Scout intelligence layer — Kat Downs Mulder
> [07:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=439s) When users are in an active, inquisitive state, conversational answer engines like Scout excel at answering specific questions. — Kat Downs Mulder
> [11:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=679s) Media outlets strive to produce fresh, accurate reporting, while language models require that exact data to remain functional. — Kat Downs Mulder
> [11:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=679s) Yahoo's lasting value proposition is serving as a trusted guide to the open internet, helping users accomplish their goals across news, email, finance, and sports. — Eric Feng
> [17:51](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=1071s) endure for the next 30 years just as it had for the previous 30. — Eric Feng
> [17:51](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=1071s) force for good — Harry McCracken
## Mentioned
Kat Downs Mulder, Eric Feng, Harry McCracken, Yahoo, Yahoo News, Yahoo Games, Yahoo Weather, Yahoo Search, Yahoo Finance, Yahoo Sports, Scout, Fast Company, Ai4, ChatGPT, Gemini, Copilot, Claude, Hulu, Microsoft, Microsoft publisher content marketplace, Really Simple Licensing (RSL), San Antonio Spurs, Jerry Yang, David Filo
## Video description
Kat Downs Mulder - GM of Yahoo News, Yahoo
Eric Feng - SVP & GM, Yahoo Research, Yahoo
Harry McCracken - Global Technology Editor, Fast Company
Yahoo executives Kat Downs Mulder and Eric Feng argue that the evolution from search engines to artificial intelligence synthesis must protect the financial sustainability of the open web, leveraging Yahoo's dual position as a major digital publisher and search provider to pioneer new citation and licensing models.
The Shift from Keywords to AI Synthesis (0:04)
Digital information discovery is shifting from traditional keyword queries that yield ten blue links to conversational interfaces that synthesize direct answers. Yahoo, which operates the world's third-largest search engine alongside a media network reaching over 250 million monthly active users—including 175 million on Yahoo News—launched its dedicated answer engine, Scout, to address this change in user behavior.
Mulder and Feng state that modern consumers increasingly expect comprehensive responses to full-sentence questions rather than manual link navigation. In response, Yahoo is deploying Scout to bridge high-volume consumer reach with verified information retrieval across its product ecosystem.
Protecting Publisher Economics to Sustain AI Quality (2:05)
Feng highlights that Yahoo occupies an unusual position as both an open web search engine and a major publisher of original content across news, finance, and sports. Mulder and Feng contend that if AI retrieval tools cannibalize publisher web traffic without providing value in return, the flow of original reporting will diminish, causing the factual accuracy and reasoning quality of underlying large language models to inevitably degrade.
To address this interdependence, internal Scout telemetry indicates that language models utilize premium head publisher reporting roughly twice as often as generic web text. Mulder notes that Yahoo relies on syndication and revenue-sharing partnerships with hundreds of media outlets, prompting the company to design Scout with prominent primary citations and direct links rather than relegating source creators to background footnotes.
Multi-Pronged Monetization and Licensing Standards (5:00)
Drawing on his background at Hulu, Feng argues that online publishing cannot rely on a single revenue model in the AI era, comparing the transition to the film and television sector's reliance on theatrical windows, pay television, streaming subscriptions, and digital sales. Web publishers similarly require diversified monetization paths to support continuous editorial production.
To develop these alternatives, Yahoo joined the Microsoft publisher content marketplace to help compensate media organizations when their reporting informs generative models. Additionally, the company adopted the Really Simple Licensing (RSL) standard to establish an automated framework for web content licensing by machine systems alongside standard syndication and advertising channels.
Behavioral Signal Scale and Discovery Modes (7:19)
Yahoo supports its retrieval architecture with a 30-year technical infrastructure that tracks over 500 million profiled users and processes 300 billion behavioral signals daily. Addressing concerns over self-preferencing, Feng asserts that Scout operates neutrally, retrieving content from both Yahoo's proprietary outlets and external web publishers based strictly on relevance and content quality.
Mulder categorizes user habits into ambient push discovery and inquisitive pull inquiries. While passive morning routines depend on homepages, curated feeds, and push notifications, targeted inquiries rely on conversational retrieval. Yahoo's strategy focuses on using the Scout intelligence layer across its platforms to move users fluidly from passive awareness into deeper, intentional investigation.
While Yahoo frames its longstanding publisher relationships and emerging frameworks like Really Simple Licensing as viable models for the AI era, an unresolved industry challenge remains whether automated licensing and enhanced citations can generate enough revenue and traffic to sustain independent web journalism as synthesized answers replace direct site visits.
## Background
本内容由 YouTube 频道 Ai4 于 2026 年 8 月 22 日发布。视频记录了《Fast Company》全球技术编辑 Harry McCracken 对 Yahoo News 总经理 Kat Downs Mulder 以及 Yahoo Research 高级副总裁兼总经理 Eric Feng 的深度对谈，围绕雅虎新一代 AI 答案引擎 Scout 展开，探讨 AI 搜索如何赋能并维系开放网络生态。
## TL;DR
面对生成式 AI 对传统搜索与媒体行业的巨大冲击，雅虎凭借兼具大型内容发布商与全球第三大搜索引擎的双重身份，推出了全新 AI 答案引擎 Scout。雅虎强调开放网络（Open Web）是维持大语言模型智力上限的基石，主张通过醒目来源引用、参与微软发布商内容市场及支持 RSL 授权协议等多元商业模式反哺原创媒体。同时，雅虎依托 5 亿用户画像与每日 3000 亿条行为信号，将被动浏览与主动探索深度融合，坚守以‘信任与向导’为核心的三十年品牌底色。
## Atmosphere & style
整场对谈专业深刻且富有前瞻性，三位资深媒体与科技领袖在平实质朴的对话中碰撞出理性的火花。交流氛围既坦诚正视了大模型‘抽吸’内容给传统出版业带来的存亡危机，又充满建设性的乐观态度与战略笃定。现场叙事条理清晰、层层递进，既有宏观生态与商业机制的理性剖析，也洋溢着雅虎三十年来作为互联网守望者与向导的行业情怀。
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=4s) 雅虎拥有庞大的用户基本盘：Yahoo News 月活跃用户达 1.75 亿，全平台月活达 2.5 亿（覆盖全美约 90% 人口），搜索业务仍为全球第三大搜索引擎。
- [00:04](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=4s) 用户搜索行为正经历从“碎片关键词与 10 条蓝色链接”向“完整自然语言问题与综合提炼答案（synthesized answers）”的范式转移。
- [02:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=139s) 雅虎兼具顶级内容发布商与搜索引擎的双重身份，主张 AI 产品的可持续性依赖于开放网络（Open Web）的健康生态，防止公地悲剧。
- [02:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=139s) 若缺乏优质原创媒体输入，开放网络萎缩将直接导致大语言模型（LLM）的智能与知识质量退化。
- [03:30](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=210s) 雅虎通过收益分成（Revshare）反哺合作出版商，并在 Scout 答案中显著突出原创来源，而非仅作为文末微小脚注。
- [03:30](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=210s) 为探索 AI 时代的内容补偿与版权变现，雅虎加入了微软发布商内容市场，并率先支持 RSL（简易版权许可）开放协议。
- [07:26](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=446s) Scout 依托雅虎 5 亿用户图谱及每日 3000 亿条行为信号构建，同时坚守搜索中立性，依据内容质量而非自有归属进行整合展示。
- [09:16](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=556s) 信息消费被划分为“Push 模式”（被动泛浏览，如首页推送）与“Pull 模式”（主动深度探索，如 Scout 问答），Scout 智能层旨在无缝连接二者。
- [10:46](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=646s) LLM 对高时效、高质信息的需求与专业新闻业天然契合，媒体应对变革需兼顾渐进式优化（如摘要）与颠覆式重构（如新型问答引擎）。
- [13:09](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=789s) Scout 实践数据显示，LLM 引用权威头部出版商内容的频率是普通网页内容的约两倍，机器价值判断与人类高度一致。
- [15:46](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=946s) 面对 AI 浪潮，雅虎将延续其 30 年来作为“互联网向导与策展人”的定位，以“信任（Trust）”为核心价值应对信息过载与失真挑战。
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=4s) 雅虎的双重角色：内容帝国与 AI 答案引擎 Scout 的诞生
主持人 Harry McCracken（《Fast Company》全球技术编辑）开场指出，虽然雅虎（Yahoo）家喻户晓，但其当前的业务版图尤其是 Yahoo News 与新推出的 Yahoo Scout 仍需深入探讨。Yahoo News 总经理 Kat Downs Mulder 介绍了雅虎庞大的用户基本盘：Yahoo News 涵盖 Yahoo.com、Yahoo News、Yahoo Games、Yahoo Weather 等一系列产品矩阵，单是新闻业务的月活跃用户数（MAU）就高达 1.75 亿，而整个雅虎全平台的月活跃用户数更是达到了 2.5 亿。这相当于全美约 90% 的人口都在使用雅虎，Mulder 风趣地调侃道：“我们常说，这个比例比吃汉堡加番茄酱的人还要多，信不信由你。”在当今充斥着混乱与虚假信息的舆论环境中，雅虎新闻的核心使命就是帮助海量用户获取高质量的真实信息。
负责雅虎消费者搜索业务与 AI 答案引擎的 Yahoo Research 高级副总裁兼总经理 Eric Feng 随后介绍了搜索业务的演进。目前 `search.yahoo.com` 依然稳居全球第三大搜索引擎，而雅虎在今年推出了首个 AI 答案引擎产品——Scout（访问入口为 `scout.com`）。Feng 指出，当前信息检索与搜索行业正处于由 AI 引发的颠覆性变革初期：用户的搜索习惯正从输入“碎片化关键词”转向提出“完整的自然语言问题”，对结果的预期也从传统的“10 条蓝色链接（10 blue links）”演变为“高度整合与提炼的综合答案（fully synthesized answers）”。
- **Yahoo News 月活跃用户数**：约 1.75 亿。
- **雅虎全平台月活跃用户数**：约 2.5 亿（覆盖约 90% 的美国人）。
- **全球搜索市场地位**：雅虎搜索依然保持全球第三大搜索引擎的位置。
- **全新 AI 引擎**：推出 AI 答案引擎 Yahoo Scout（`scout.com`）。
---
### [02:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=139s) 对抗“公地悲剧”：开放网络生态与 AI 的共生可持续性
针对 ChatGPT、Google Gemini、Microsoft Copilot、Anthropic Claude 等聊天机器人的冲击，Harry 询问 Scout 与这些竞品的本质区别。Eric Feng 指出，雅虎最大的独特性在于它同时身处天平的两端——雅虎既是全球最大的内容发布商之一（拥有 Yahoo News、Yahoo Sports、Yahoo Finance 等庞大媒体矩阵），又拥有全球领先的搜索引擎。这种双重身份赋予了雅虎对整个媒体与搜索行业全局健康发展的责任感与长远视角。
Scout 核心的设计基石是“可持续性（Sustainability）”。Feng 警告称，开放网络（Open Web）的健康直接决定了大语言模型（LLM）的智力上限：如果用户停止访问像《Fast Company》这样的优质原创媒体，开放网络陷入萎缩与退化，LLM 能够学习和吸收的专业知识质量也必然随之退化。因此，保障内容发布商的利益、维护繁荣的开放网络，本质上是在为终端用户打造更优质的 AI 产品。
```mermaid
flowchart LR
accTitle: 开放网络与大语言模型的共生循环生态图
subgraph PublisherEcosystem["优质内容生态"]
direction TB
A["原创发布商 / 媒体<br/>(Fast Company, Yahoo 等)"]
B["高质量专业知识 & 实时信息"]
A --> B
end
subgraph AIScalability["AI 与用户闭环"]
direction TB
C["LLM 训练与检索增强<br/>(高质量输入)"]
D["优质 AI 答案 / Scout 引擎"]
E["回流流量、版权收益与用户激励"]
C --> D
D --> E
end
B --> C
E --> A
```
- **双重业务身份**：兼具“大型内容发布商”与“顶级搜索引擎”双重视角。
- **核心逻辑**：开放网络若退化，AI 大模型的知识与专业度必然退化；维持媒体生态是提升 AI 体验的先决条件。
---
### [03:30](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=210s) 媒体联盟、版权许可（RSL）与多元化变现模式
Kat Downs Mulder 深入阐述了雅虎新闻的生态运作机制。雅虎在本质上是一个聚合平台（Aggregator），不仅产出原创内容，还与美国及全球数百家优质媒体建立了内容辛迪加（Syndication）合作关系。雅虎将这些媒体的内容分发给数以亿计的用户，并将产生的广告及商业收益通过收益分成（Revshare）返还给合作出版商。这种共生关系为新闻行业提供了持续的经济激励和受众扩张渠道。
基于这一理念，Yahoo Scout 在设计上极度注重对原作者与出版商的尊重：不仅在生成的综合答案中显著标明引用来源，而且摒弃了仅在文末放置微小脚注的做法，将原始报道和链接醒目地呈现在答案主体中。
关于 AI 时代媒体内容的商业化变现，Eric Feng 结合自身在视频流媒体平台 Hulu 的职业经历指出，未来的内容变现不会是单一模式，而是多元商业模式的集合（正如影视业由院线、流媒体、传统电视广播和 iTunes 数字销售共同支撑）。雅虎在此领域积极布局：
1. **加入微软内容市场**：在 1 月发布 Scout 时，雅虎宣布加入“微软发布商内容市场（Microsoft Publisher Content Marketplace）”，探索媒体内容被 AI 调用时的补偿机制。
2. **支持 RSL 标准**：雅虎率先支持“简易版权许可（Really Simple Licensing, RSL）”协议，为发布商提供新型的内容授权与变现标准。
- **深度共生机制**：通过收益分成（Revshare）反哺数百家原创出版机构。
- **引用展示**：Scout 在答案中突出显示原始内容来源与链接，而非单纯隐藏于脚注。
- **商业化新基建**：接入 Microsoft Publisher Content Marketplace，并全面支持 RSL（Really Simple Licensing）开放协议。
---
### [07:26](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=446s) 雅虎的技术底座：5 亿用户画像与 3000 亿日活行为信号
谈及雅虎支撑新型 AI 体验的技术实力，Eric Feng 兴奋地表示，雅虎在互联网领域深耕 30 年所积累的资产就像一个巨大的“玩具箱（Toy Box）”。雅虎拥有超过 5 亿用户的图谱与画像信息（User Graph），并且系统每天能捕捉并处理大约 3000 亿条用户行为信号（User Signals）。这些深厚的数据积淀构成了 Scout 推荐与合成算法的坚实基础。
Feng 同时强调，Scout 在信息整合上坚守“搜索中立性”。例如，当用户在 Scout 上查询其最喜欢的 NBA 球队“圣安东尼奥马刺队（San Antonio Spurs）”时，生成的回答虽然会吸纳 Yahoo Sports 的深度报道，但同时也会广泛综合全网其他顶级体育媒体的信息。搜索作为互联网的门户（Front Door），核心原则是保障所有内容的公平访问；优质内容之所以能在 Scout 中获得突出展示，完全取决于内容本身的质量，而非雅虎的内部归属。
- **用户图谱积累**：覆盖超过 5 亿人的用户画像库。
- **实时行为信号**：每日捕获并处理约 3000 亿条用户交互信号。
- **中立展示原则**：坚持内容为王，对自有内容（如 Yahoo Sports）与第三方外部媒体一视同仁。
---
### [09:16](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=556s) 用户消费模式演进：从“被动推送”到“主动深挖”与 Scout 智能层
在讨论 AI 搜索对内容运营的影响时，Kat Downs Mulder 提出了用户消费内容的两大核心模式演变：
1. **主动探索模式（Inquisitive / Pull Mode）**：用户带着具体的问题、明确的学习目的或特定的对话诉求而来，此时以 Scout 为代表的基于自然语言回答的交互界面能够精准“满足用户的即时需求（Meet users where they are）”。
2. **被动接收模式（Ambient / Push Mode）**：用户在清晨醒来时并不带有特定查询目的，只是想浏览当下发生了什么。这种“环境式”的信息获取需求由 Yahoo 首页及 Yahoo News 通过算法推送与通知来高效满足。
Mulder 强调，雅虎的关键战略在于连接这两种模式——当用户通过首页推送获得对某条新闻的泛泛兴趣后，雅虎通过将“Scout 智能层（Scout Intelligence Layer）”无缝嵌入全线产品，提供一整套深度探索工具，帮助用户顺畅地从“被动浏览”过渡到“深入探究”。
- **双模态体验设计**：
- **Push 模式**（Yahoo Homepage / Yahoo News）：满足日常无目的、泛浏览的环境式（Ambient）信息获取。
- **Pull 模式**（Yahoo Scout 答案引擎）：满足针对特定事件或疑问的高互动深度探索。
- **Scout 智能层定位**：作为底层通用能力横向打通雅虎各产品线，承接用户从浅层兴趣到深度洞察的转化。
### [10:46](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=646s) 原创内容与AI合成：媒体行业在颠覆性技术下的生存法则
在探讨信息呈现方式时，Kat Downs Mulder 指出，个性化推荐与主动探索（Push and Pull）这两种体验完全可以优美地共存。然而，如果网络上充斥的都是被 AI 二次合成的信息，而缺乏源头活水般的原创内容，整个信息生态的质量就会不可避免地下滑。出版业当前面临的最大挑战，正是如何在新的技术环境下持续支撑起高产量、高质量的原创内容生产。
当 Fast Company 的 Harry McCracken 询问媒体从业者应如何应对当下局势、在变革中确立生存路径时，Kat 强调首先“绝不能低估这项技术将带来的颠覆程度”。一旦正视这一现实，真正的机遇才会显现。她认为，大语言模型（LLM）需要高相关性、高质量且最新的信息来进行推理与回答，而优秀的新闻工作者恰恰致力于生产这种优质的即时资讯，两者的底层诉求在本质上是高度一致的。目前行业急需解决的核心难题是经济分配机制与商业变现模式。Kat 提出，媒体与科技公司不仅要采取渐进式改良（如开发摘要、提炼简报等优化信息获取体验的功能），更必须勇于下大赌注（Big Bets），从根本上重塑信息的消费方式——Yahoo 正在探索的“Scout”问答引擎就是这种尝试的一部分，旨在借助 AI 让优质信息更具可见度。
```mermaid
flowchart LR
accTitle: 媒体与AI的激励协同闭环
subgraph ContentEcosystem["内容生产与消费生态"]
A["专业记者与出版商<br/>（产出高质量原创内容）"] -->|"提供高新鲜度语料"| B["大语言模型与AI系统<br/>（推理、合成与问答）"]
B -->|"精准分发与价值回流"| C["终端用户<br/>（获取可信理解与服务）"]
C -.->|"建立可持续变现与订阅机制"| A
end
```
- **核心利益契合**：LLM 对高时效、高质信息的需求，与严肃新闻媒体的生产目标天然契合。
- **双轨探索路径**：
- **渐进式改进**：利用 AI 提供文章摘要（Summaries）、新闻速览（Digests），降低阅读门槛。
- **重构式创新**：如 Yahoo Scout，打造新型智能交互产品，让优质原创内容更容易被检索与发现。
---
### [13:09](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=789s) 开放互联网的存亡之辩：大模型对优质出版物内容的深层依赖
Harry McCracken 提出了传媒界普遍存在的深层忧虑：大语言模型正在肆无忌惮地“抽吸”（hoovering up）各大媒体的内容，并直接向用户输出答案；即使运气好能给出一个引用链接，也往往无法带来实际点击。如果受众不再访问媒体网站，出版商失去营收来源并被迫倒闭，LLM 最终将面临无内容可用的荒漠，诞生自上世纪 90 年代的开放互联网（Open Web）生态或将走向枯萎。
面对这一严峻质疑，Eric Feng 直言这种担忧“百分之百是合理的”，但他依然抱有坚定的乐观态度。他的乐观根源于双方深层的利益一致性：“拥有充满活力的开放互联网和强大的出版商生态，完全符合大语言模型的最大利益。”Eric 强调，解决流量与生存危机不会存在“一刀切”（One-size-fits-all）的通用方案，行业目前尚处于初期探索阶段，未来需要组合运用多种技术方案、商业模式以及版权与收益结算机制。
Eric 分享了 Yahoo 在研发旗下问答引擎 Scout 时观察到的第一手实证数据：
- **头部出版商内容的超高调用率**：在 Scout 的实际运行中，来自头部权威出版商的内容被 LLM 引用的频率，大约是普通网页内容（General Web Content）的**两倍（2x）**。
- **机器推理偏好与人类一致**：大语言模型在信息理解与价值判断上的逻辑与人类非常相似，能够明确识别并高度珍视权威机构产出的专业报道。
---
### [15:46](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=946s) 穿越三十载的基石：以“信任与向导”定位迎接 AI 新浪潮
Harry 提到一个有趣的设想：如果一个 1990 年代的人穿越到 2026 年打开 Yahoo，虽然会为诸多现代化技术感到惊叹，但依然能一眼认出它，因为 Yahoo 的使命始终清晰稳固。他由此发问：AI 是否会在未来彻底改变 Yahoo 的本质，还是说其核心价值在未来十年依然如过去三十年一样历久弥新？
Kat Downs Mulder 回应道，Yahoo 从诞生之初就扮演着“互联网向导”（Great Guide）与“网络策展人”（Great Curator）的角色，其根本目的始终是帮助人们达成目标——无论是获取信息、关注赛事、查询财务，还是通过电子邮箱管理日常生活。虽然产品的表现形态在未来可能会发生翻天覆地的变化，甚至诞生出全新维度的产品品类，但这种深植于品牌中的 DNA 不会改变。
Eric Feng 则着重强调了“**信任**（Trust）”这一关键词。他回顾了 1994 年杨致远（Jerry Yang）和大卫·费罗（David Filo）创立“Jerry and David's Guide to the World Wide Web”的最初时刻，指出 Yahoo 的核心价值正是在浩瀚的网络世界中为用户增加一层信任保障。
> “随着 AI 在未来不可避免地扮演愈发重要的角色，为内容和交互体验提供‘信任’背书的需求不仅不会消失，反而会变得前所未有的重要。这项使命在未来的三十年里，将一如既往地延续下去。”
在对谈的尾声，Harry McCracken 感谢了 Yahoo 团队作为行业积极推动力量所做出的探索，并向线上观众致谢，对谈在融洽的气氛中圆满结束。
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=4s) 我们常说，这个比例比吃汉堡加番茄酱的人还要多，信不信由你。 — Kat Downs Mulder
> [00:04](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=4s) 用户的搜索习惯正从输入“碎片化关键词”转向提出“完整的自然语言问题”，对结果的预期也从传统的“10 条蓝色链接”演变为“高度整合与提炼的综合答案”。 — Eric Feng
> [02:19](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=139s) 开放网络（Open Web）的健康直接决定了大语言模型（LLM）的智力上限：如果用户停止访问优质原创媒体，开放网络陷入萎缩与退化，LLM 能够学习和吸收的专业知识质量也必然随之退化。 — Eric Feng
> [07:26](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=446s) 雅虎在互联网领域深耕 30 年所积累的资产就像一个巨大的“玩具箱”。 — Eric Feng
> [10:46](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=646s) 如果网络上充斥的都是被 AI 二次合成的信息，而缺乏源头活水般的原创内容，整个信息生态的质量就会不可避免地下滑。 — Kat Downs Mulder
> [10:46](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=646s) 绝不能低估这项技术将带来的颠覆程度。 — Kat Downs Mulder
> [13:09](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=789s) 拥有充满活力的开放互联网和强大的出版商生态，完全符合大语言模型的最大利益。 — Eric Feng
> [15:46](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=946s) Yahoo 从诞生之初就扮演着“互联网向导”与“网络策展人”的角色，其根本目的始终是帮助人们达成目标……这种深植于品牌中的 DNA 不会改变。 — Kat Downs Mulder
> [15:46](https://www.youtube.com/watch?v=ajBlbvp-gcM&t=946s) 随着 AI 在未来不可避免地扮演愈发重要的角色，为内容和交互体验提供“信任”背书的需求不仅不会消失，反而会变得前所未有的重要。这项使命在未来的三十年里，将一如既往地延续下去。 — Eric Feng
## Mentioned
Yahoo (雅虎), Yahoo News, Yahoo Scout (scout.com), Yahoo Research, Yahoo Sports, Yahoo Finance, Yahoo Games, Yahoo Weather, Fast Company, Harry McCracken, Kat Downs Mulder, Eric Feng, ChatGPT, Google Gemini, Microsoft Copilot, Anthropic Claude, Microsoft Publisher Content Marketplace (微软发布商内容市场), Really Simple Licensing (RSL, 简易版权许可), Hulu, San Antonio Spurs (圣安东尼奥马刺队), Jerry Yang (杨致远), David Filo (大卫·费罗), Jerry and David's Guide to the World Wide Web
## Video description
Kat Downs Mulder - GM of Yahoo News, Yahoo
Eric Feng - SVP & GM, Yahoo Research, Yahoo
Harry McCracken - Global Technology Editor, Fast Company
[Source](https://www.youtube.com/watch?v=ajBlbvp-gcM)
Source quotes
[0:04] Kat Downs Mulder: 我们常说，这个比例比吃汉堡加番茄酱的人还要多，信不信由你。
[0:04] Eric Feng: 用户的搜索习惯正从输入“碎片化关键词”转向提出“完整的自然语言问题”，对结果的预期也从传统的“10 条蓝色链接”演变为“高度整合与提炼的综合答案”。
[2:19] Eric Feng: 开放网络（Open Web）的健康直接决定了大语言模型（LLM）的智力上限：如果用户停止访问优质原创媒体，开放网络陷入萎缩与退化，LLM 能够学习和吸收的专业知识质量也必然随之退化。
[7:26] Eric Feng: 雅虎在互联网领域深耕 30 年所积累的资产就像一个巨大的“玩具箱”。
[10:46] Kat Downs Mulder: 如果网络上充斥的都是被 AI 二次合成的信息，而缺乏源头活水般的原创内容，整个信息生态的质量就会不可避免地下滑。
[10:46] Kat Downs Mulder: 绝不能低估这项技术将带来的颠覆程度。
[13:09] Eric Feng: 拥有充满活力的开放互联网和强大的出版商生态，完全符合大语言模型的最大利益。
[15:46] Kat Downs Mulder: Yahoo 从诞生之初就扮演着“互联网向导”与“网络策展人”的角色，其根本目的始终是帮助人们达成目标……这种深植于品牌中的 DNA 不会改变。
[15:46] Eric Feng: 随着 AI 在未来不可避免地扮演愈发重要的角色，为内容和交互体验提供“信任”背书的需求不仅不会消失，反而会变得前所未有的重要。这项使命在未来的三十年里，将一如既往地延续下去。
Source quotes
[0:04] Kat Downs Mulder: more than people who put ketchup on their hamburgers.
[2:05] Eric Feng: Yahoo operates from a rare vantage point: it is simultaneously one of the largest web search engines and one of the world’s biggest original content publishers across news, sports, and finance.
[2:05] Eric Feng: If publishers stop seeing value and traffic degrades, the underlying knowledge and reasoning quality of large language models (LLMs) will inevitably collapse as well.
[2:05] Kat Downs Mulder: Yahoo maintains syndication and revenue-sharing partnerships with hundreds of global publishers, funding original journalism while bringing high-quality reporting to broader audiences.
[5:00] Eric Feng: Just as video sustainability required a combination of theatrical windows, pay TV, streaming platforms, and digital sales, online publishing must embrace varied monetization models.
[7:19] Kat Downs Mulder: Scout intelligence layer
[7:19] Kat Downs Mulder: When users are in an active, inquisitive state, conversational answer engines like Scout excel at answering specific questions.
[11:19] Kat Downs Mulder: Media outlets strive to produce fresh, accurate reporting, while language models require that exact data to remain functional.
[11:19] Eric Feng: Yahoo's lasting value proposition is serving as a trusted guide to the open internet, helping users accomplish their goals across news, email, finance, and sports.
[17:51] Eric Feng: endure for the next 30 years just as it had for the previous 30.
[17:51] Harry McCracken: force for good
Original transcript
[0:04] Thank you so much for joining us. Thank
you, Cad and Eric, for being here today. Um,
[0:10] everybody knows Yahoo, so I'm not going to ask
you to explain what Yahoo is, but it might be
[0:15] useful to talk a little bit about what Yahoo
News is today and what Yahoo Scout is. So,
[0:21] Kat, could you talk a little bit about your world?
Absolutely. Hi everyone. Um, I'm Kat Dwolder. the
[0:26] GM of Yahoo News, uh, which includes Yahoo.com,
Yahoo News, Yahoo Games, Yahoo Weather. Um,
[0:34] a whole bunch of products that support people
getting getting the news and information they need
[0:38] on a daily basis. Um, we have a large audience.
Uh, we have a lot of users at Yahoo. About 175
[0:46] million monthly users just for news. About 250
million monthly users for all of Yahoo, which is
[0:52] about nine out of 10 Americans, which we like to
say is more than people who put ketchup on their
[0:56] hamburgers, if you can believe it. Uh, it's a
lot of people. Um, and so we're really focused on
[1:02] figuring out how we help them. uh and especially
in an increasingly you know increasingly chaotic
[1:07] information landscape with a lot of misinformation
and it people get quality information. So that's
[1:13] our main focus um and and what we spend most of
our time figuring out. Uh hello hello my name is
[1:20] Eric Fang. Uh I lead up Yahoo's consumer search
efforts and our AI answer engine. Uh if you go to
[1:26] searchi.com that's our consumer search product
which is still the third largest search engine
[1:31] in the world. And um uh this year we were very
excited to launch our first ever AI answer engine
[1:37] called Scout which you can access at scout.com.
And um it is a very exciting time for search and
[1:44] information retrieval as a whole because of the
disruptive changes that are happening with AI with
[1:49] users now. Instead of asking keywords, they're
uh asking full questions and instead of expecting
[1:55] just 10 blue links back, they want to see fully
synthesized answers. So, um, we're just at the
[2:00] early stages of trying to figure out what that
means for the entire industry. But, uh, really
[2:05] fun space to work on. A lot of people these days
instead of using a a search engine, they are going
[2:11] to chat GPT or Gemini or Copilot or Claude and
asking those full-bodied questions. Talk a little
[2:19] bit about how Scout is different from from those
chatbot experiences. Yeah, I one of the really
[2:25] interesting things about building this within
Yahoo and uh in partnership with CAD is that
[2:29] we sit on both sides of the equation as one of
the world's largest content publishers. Uh think
[2:36] about all the content that we have at Yahoo News
and sports and finance. So we are a big content
[2:41] publisher but also a very large search engine. So
because of that we have this unique perspective
[2:45] of trying to figure out what makes sense for the
entire industry and that's a responsibility we
[2:50] take very very seriously. So, one of the core
tenants of what we're trying to build at Scout
[2:54] is something that is sustainable. It's sustainable
not only for um the publishing ecosystem, but it's
[3:00] also sustainable for users because if you think
about it, if the open web degrades, if people stop
[3:07] visiting awesome sites like Fast Company, um the
quality and the knowledge and the expertise of LLM
[3:14] will also degrade. So, by having a healthy open
web, we are actually making a better AI product
[3:20] for end users. So this idea of sustainable is not
just for the publishing industry but it's also for
[3:24] end users as well. So um that that is a really
core tenant of what we uh build one. K talk a
[3:30] little bit about what Yahoo News is made up of
and the fact it reflects Yahoo created content
[3:36] but also all these relationships that Yahoo has.
And I I know you have a lot of experience in these
[3:41] partnerships because maybe 25 years ago I worked
for a media outlet whose relationship with Yahoo
[3:46] was really important. Yeah. Um, so we are an
aggregator at heart. We also produce original
[3:52] content, a lot of it as Eric has suggested, but
we have relationships with hundreds of publishers
[3:58] uh in the US and around the globe. Um, and we, you
know, send we we syndicate their content. We bring
[4:03] it to our audiences. Um, and we pay them back a
revshare to those publishers. So, we help support
[4:09] the journalism that they're producing um by by
giving them financial incentive and and access
[4:14] to new audiences which help them grow. Um so that
relationship between us and publishers is really
[4:20] symbiotic and it's very very important to us um
to continue to have a vibrant publishing ecosystem
[4:27] um on the open web and that's one of the you know
sort of core principles that went into the design
[4:32] of Yahoo Scout um because we have because
we have this interest because we really care
[4:37] about these relationships and the quality of the
information on the web um to create a product in
[4:41] Scout that does lots of things like surface links
and and and citation you know not just footnotes,
[4:47] right? But really present in the answers in the
product, original sourcing um that really brings
[4:53] a lot of um emphasis to to those publishers who
are creating the information to begin with. And
[5:00] it does this offer people a a uh a route into all
this content you have that that's better than the
[5:07] days when we all went to the your homepage and
typed in a keyword or two. Yeah, I think that
[5:13] the future of content discovery is going to be
not one answer, one solution. It's going to be
[5:21] multimodal. Some people will bookmark and go
directly to sites like Yahoo Sports uh or Yahoo
[5:28] Finance for their content. Some will go through
search engines looking for keywords and um trying
[5:33] to directly nav and some will want synthesis
uh and sort of summaries of information. So,
[5:39] I think uh as a content aggregator, as a search
engine, we have to meet users where they're at and
[5:46] um and offer all of those different experiences
to the users because it's not going to be one
[5:50] sizefits-all. But I think what's really important
again is this idea of whatever we build, we want
[5:56] to do in partnership with the publishing industry.
When we launched Scout in January, we actually
[6:02] also announced that we had joined uh the Microsoft
publisher content marketplace, which is a really
[6:07] exciting program that Microsoft's pioneering to
try to find new revenue sources for the media
[6:13] industry where they can actually get compensated
for AI usage of their content. Um so we are very
[6:18] very proud to be part of that initiative. We
also launched um last year um uh our support for
[6:24] a standard called RSL, really simple licensing,
which again is a new way that uh publishers can
[6:31] monetize their content. And um it's it's uh I in
my in my background, I came from um from media
[6:38] and and the and online video. I used to uh work
at Hulu and it's a very similar situation where
[6:44] it's not going to be one size fit all for media
content. You know, just like in the Hulu days,
[6:49] there were theatrical, there were streaming, there
was still TV, there was still uh uh you know,
[6:53] iTunes sales. It was a combination of all these
different business models that allowed media to be
[6:58] on the video side to be sustainable. And I think
on web publishing, it's going to be a similar
[7:03] situation where we're going to have a whole bunch
of different solutions that continue to allow this
[7:08] uh industry to be sustainable. And again, it's
in everyone's best interest. It is in in the uh
[7:13] OpenAI's best interest. It is an end user's best
interest because without this content, these LLMs
[7:19] are just not that smart. Talk a little bit about
Yahoo's technical stack. I mean, search has always
[7:26] been a form of AI. So, you have a lot of history
in that, but um how much of this is stuff you've
[7:32] had to build and how easy is it to put this new
type of interface on top of the content you have?
[7:37] Um, I have been so excited to build these types of
products and experiences inside of Yahoo because
[7:43] of all the toys that we have. Think about it.
Um, Yahoo has been an internet pioneer for three
[7:48] decades now. And through that time, we've amassed
a um a a user graph of of profile information of
[7:56] over half a billion people. Um, every day we are
capturing about 300 billion user signals. Um,
[8:03] so there's so much interesting um, toys that we
get to play with that inform the experience that
[8:09] you get on Yahoo Scout. So if you guys go to Yahoo
Scout, uh, we think that we deliver something that
[8:14] is really grounded in that unique perspective. Um,
it is and but it's not just grounded in Yahoo's
[8:22] original content. As Cat mentioned, we are a
very large aggregator. So, if you were to go
[8:27] to Yahoo Scout today and ask a question about the
San Antonio Spurs, which is my favorite basketball
[8:31] team, um you would probably get um interesting
information that's synthesized, of course,
[8:36] from Yahoo Sports, but not just from Yahoo Sports.
We really want to offer you the best of the open
[8:41] web. And that's the the the the kind of the the
neutral position that we uh sit at while operating
[8:49] a search business. And and that's kind of the
unique perspective that we bring because search
[8:53] is again a front door to the internet and it is
very much based on equal access to all content and
[9:00] that perspective helps inform the answers that you
get on Scout. And of course Yahoo Sports is going
[9:05] to be prominently featured but not because we own
Yahoo Sports but because they have great content.
[9:10] It's really the great content that wins out inside
of the uh the scout experience. Kad, what's it
[9:16] like running a content operation as these types of
interfaces start to become commonplace and maybe
[9:23] over time the predominant way that that people
find what you have to offer? Well, I I just want
[9:30] to um sort of lean into something that Eric talked
about, which is I think there is different modes
[9:36] of consumptions for users. So if a user is in an
inquisitive mode, they have a specific question, a
[9:42] specific they may want to learn out of a specific
conversation that they want to have, um these sort
[9:47] of answersbased sessions are wonderful for that.
And so this, you know, Yahoo Scout meets them
[9:53] where they are with that experience. In a world
where you just you just wake up in the morning
[9:58] and you want to know, you would just want to kind
of know what's going on. That is more of a push.
[10:03] You know, we're going to push that to you rather
than you have to ask. That's where products like
[10:08] news and Yahoo home come into play, right? They're
just serving a different need. It's not a request,
[10:13] it's a push. And so, we're really focused on how
do we figure out what information th those users
[10:19] need in that moment? What's relevant in a push
notification? What's relevant on the homepage? Um,
[10:23] or in any other surface where they're just sort
of looking at this information in an ambient way.
[10:28] Then once they've consumed that, how do they go
deeper? Right? then something like scout or scout
[10:32] intelligence layer which is the way that we use
scout across all Yahoo products becomes really
[10:37] relevant. So if I've started on my journey
wanting to know about a current news event,
[10:41] how do we help you go deeper with this toolkit?
Um and sort of convert that ambient interest into
[10:46] something that's more intentional and tailored
to you. So I think they can coexist really
[10:51] beautifully. Um I think there's there's a need
for both that both that push and pull experience.
[10:58] Um, but to Eric's earlier point, I really
think that we have to continue to have original
[11:03] information versus just sort of synthesized
information, otherwise the quality of the whole
[11:08] thing kind of degrades. So, I think that's more
where the challenge is for the publishing industry
[11:12] is how to continue to support the production of a
high volume of high quality original information.
[11:19] any tips for those of us such as me who are in
the media business about how to figure out how
[11:25] to be viable and and what should change these
days and what kind of needs to stay the same.
[11:31] I mean I think that this moment I don't think that
we can underestimate the amount of disruption that
[11:37] this technology is going to cause. Uh so I think
that's thing number one is we have to we have to
[11:42] understand that and and not minimize its impact
or potential impact. I think once you do that then
[11:48] opportunities start to really open up. I think
one thing that's really good about this is um
[11:53] that the incentives now for information creation
or production are really well aligned between LLMs
[11:59] who really need good quality, you know, highly
relevant fresh information and journalists who
[12:05] want to create really good quality fresh recent
information, right? Those are very well aligned.
[12:10] Um, and what we need to figure out is the
economics of how that gets, you know, how does
[12:14] that get monetized and all of those things, which
which we absolutely have to figure out. But from
[12:18] an incentive standpoint, I think is there. Um, and
I think we just have to be, you know, we we we are
[12:25] certainly going to be incremental in some ways
in terms of, you know, sort of changing things
[12:31] by bit by bit. I would say those are things like
summaries and digests and stuff like that, right?
[12:35] How do we just make the information consumption
experience better? But I think as an industry we
[12:40] also have to take some bigger bets. Um and sort of
rethinking how this technology enables information
[12:46] to be consumed. Um and for us, Scout is part of
that and sort of how do we create products that
[12:52] can better serve users and their their need
to understand um and use this technology to
[12:57] highlight and help make more discoverable quality
information. Um and so I'm I'm excited about it.
[13:04] I mean, I think there's a lot to be figured out,
but it's a it's a moment of tremendous opportunity
[13:09] if approached the right way. Eric, um, you talked
about the fact that a lot of media companies are
[13:15] concerned about the process pro prospect of LLM's
hoovering up all of our content and then answering
[13:22] questions with maybe if we're lucky, a citation,
but maybe no citation. And um, you know, the
[13:28] upshot being nobody comes to our sites. If nobody
comes to our sites, we don't have a business.
[13:34] if we don't have a business and the LLM don't
have any content and and maybe that this whole
[13:38] concept of of the worldwide web that Yahoo has
been part of from the starts to wither away. Is
[13:44] this a reasonable concern or how how pessimistic
or optimistic are you given that Yahoo can do
[13:50] a lot to help but Yahoo is not going to save the
open web all by itself? Um it is 100% a legitimate
[13:57] concern. Uh I am though uh unequivocally an
optimist on things and what gives me optimism
[14:04] is again this core belief that the incentives are
aligned. They are truly aligned. Like it is in the
[14:11] LLM's best interest that there is a vibrant open
web and a great publisher ecosystem. It the it's
[14:17] in the interest of users that they support
publisher sites in order to make sure that
[14:22] there's great content being generated that informs
the LM. So given the incentives are aligned that
[14:28] that is really what fuels my optimism that we will
figure out a solution but I do believe that it is
[14:33] not as simple as one one uh solution that will
one sizefits-all. Um it's going to take multiple
[14:40] different um uh uh uh techniques and business
models and payment mechanisms in order to make
[14:47] this work. Um and we are in the absolute early
early innings of figuring that out. So, um it's
[14:53] going to be bumpy, but given the intentions are
aligned, I do believe that we can um we can find
[14:58] a good path forward and um make no mistake of
the importance of publisher content. It is so
[15:06] incredibly important to these systems working.
Uh we see it every day inside of uh if you go to
[15:13] scout.com and play and experiment with our answer
engine, we see it every day in the data that we're
[15:17] getting back from there. um head publisher content
like the stuff that you write, it gets used by the
[15:23] LM LLM about twice as often as general web
content. Um the LLMs, the language models,
[15:30] they reason in in a lot of the same ways that that
humans do and they value this content just as much
[15:36] as we do. So um the importance of that content is
there. I think everyone knows it and now we just
[15:41] have to work together to figure out a solution
that's sustainable. Yeah. If somebody from the
[15:46] 1990s woke up in 2026 and went to Yahoo, they'd
probably be amazed by a lot of aspects of it,
[15:52] but they would also find it recognizable because
this has been a very sturdy mission that you have.
[15:58] At some point, does AI like completely change what
Yahoo is or or 10 years from now, will what Yahoo
[16:04] is doing still be as useful as it has been for
the last 30 years? I think Yahoo has fundamentally
[16:10] since its origins really served this need of
like curating and organizing the internet to help
[16:15] people understand what's going on. We think of
that as being like, you know, great guide to the
[16:18] internet, great curator of the internet. And what
we're really fundamentally trying to do is help
[16:23] people accomplish their goals, whether those are
information needs or things like managing their
[16:27] life through email. That bedrock has been there
through this whole evolution. You know, like you
[16:32] said, you can kind of see the DNA. It might look
different. It might manifest differently, but that
[16:36] DNA is like still there. I think that there is a
tremendous amount of room to keep experimenting
[16:42] and explore it with the form that that takes. And
I think sure there's the potential that that form
[16:48] will change dramatically over the coming years,
but I think that fundamental mission and that DNA
[16:52] will not change because there's so much need for
it. And I think that's where Yahoo really has the
[16:58] biggest opportunity to continue to deliver value
with all the assets that that Eric talked about
[17:02] before, right? We have relation deep and lasting
relationships uh with so many users who trust us
[17:08] with their mail, with their sports, with their
finances, with you know with with with their
[17:13] daily weather or things like that. Like so we're
going to continue to lean in that. I do think
[17:17] there's a tremendous opportunity to invent new
products of category in the you know categories
[17:22] of products in the future and it's going to be
a very exciting time but I think that DNA will
[17:27] still be there. Yeah. I'll just double down on one
word that Cat mentioned which is the word trust.
[17:32] I think that that has been an enduring part
of Yahoo's value to users since the very first
[17:37] time that Jerry and David launched the, you know,
awesome guide to the internet back in 1994. U this
[17:44] idea that they were adding a layer of trust to
what you would find online. And as AI inevitably
[17:51] continues to play a larger role, the need to have
somebody provide trust uh for the content, for the
[17:59] experiences is going to increase. it's going to
it's going to grow in even more importance. So
[18:03] I think that element of our mission is going to
endure uh uh for the next 30 years just as it had
[18:08] for the previous 30. Well, thank you for being a
force for good and u thank you for taking the time
[18:14] to chat with us today and thank you audience and
enjoy the rest of your AI4. Thanks everyone. Bye.
