Liz Reid, VP and head of Google Search, told the AI Inside podcast on June 26, 2026 that paywalls predictably reduce traffic, that personalizing search results helps - rather than harms - specialist and small publishers, and that users are adapting to conversational search faster than even she had anticipated.

The interview and its context

The episode, number 135 of the AI Inside podcast hosted by Jason Howell and Jeff Jarvis, ran for roughly 75 minutes and drew on Reid's 23 years at Google. She spent 17 of those years building Google Maps before moving into a leadership role within Search. The conversation ranged from query behavior and filter bubbles to the "ignore" bug that briefly disrupted AI Overviews weeks before recording, and to the question of what search is becoming as agents begin to act on users' behalf.

The timing matters. AI Mode surpassed one billion monthly users globally as of May 2026, a milestone Google announced at I/O on May 19. Queries in AI Mode now average three times the length of traditional searches, and total query volume has more than doubled every quarter since the feature's US launch in May 2025. Against that backdrop, Reid's appearance on a podcast aimed squarely at media and technology observers carried more than routine significance.

How users are changing - faster than expected

Reid described a shift in how people phrase their searches that she said has moved more quickly than she expected. According to Reid, keyword searches reflected a translation task: users had to convert their thoughts into terms a machine could parse. Conversational queries work the opposite way - the system does the interpretation, and users speak as they would to another person.

"People are adapting faster than I might have guessed a few years ago," Reid said, noting that users are not just asking more questions but asking longer ones that include the word "I" and describe their situation in first-person terms. "I am trying to do the following things. Please help me figure this out," she said, illustrating the shift.

She argued this dynamic creates an additional layer of query growth. People make rapid mental calculations about whether a question is worth pursuing, and reducing friction - making search easier to engage with - causes questions that would otherwise be dropped to surface instead.

Search versus Gemini: different north stars

Jarvis asked directly whether Gemini and Search would converge into a single product. Reid distinguished them by purpose. According to Reid, Gemini is built around being a personal, productive, powerful assistant, while Search is focused on information journeys. Both run on the same underlying models and share tooling, but the curation at the product level differs. Search, she said, puts more weight on information needs; Gemini handles more productivity and creativity tasks.

She used AI Mode as the clearest example of this distinction. AI Mode and AI Overviews, in her framing, are specifically for users who want the web alongside AI - a synthesis of both rather than a replacement of one with the other. The product tries to make it easy to start with AI assistance and then jump off to a specific source, whether for a World Cup result or a restaurant booking.

Personalization and what it means for publishers

A substantial portion of the interview turned on the question publishers have been asking since AI Overviews began absorbing traffic: what does personalization do to their visibility? Reid made an argument that has not often been articulated publicly in this direct a form.

According to Reid, unpersonalized keyword search makes everything look the same. A generic query surfaces generic results, and the content that ranks is the content that wins at scale. Personalization, she argued, pushes results "into the tail" - surfacing the eco-friendly reviewer, the specialist merchant, the local journalist, rather than defaulting to the same handful of dominant outlets. "There's a lot of opportunity with personalization for creators and journalists who specialize in something," Reid said.

She connected this to the Preferred Sources feature, which Google extended into AI Overviews and AI Mode on May 27, 2026, allowing users to designate sites whose links receive a visible badge inside AI-generated responses. According to Google's data, users are twice as likely to click a link carrying the Preferred label. The feature now covers more than 345,000 unique sources selected by users. Reid described the mechanism as a way to let publishers build and reinforce a direct relationship with their audience, so that a reader who trusts a particular brand can make that preference legible to Search rather than relying solely on algorithmic inference.

For subscriptions, Reid said Google is working on ways to surface content from outlets where a user already holds a paid subscription - highlighting those links in AI Overviews and AI Mode specifically. The goal, as she framed it, is to strengthen the relationship between a publisher and the reader who already has an account, not to replace it.

The paywall question

Reid was candid about paywalls in a way that some publishers are unlikely to welcome. Asked about declining traffic, she described a pattern she sees repeatedly: a publisher adds a paywall and then observes that traffic falls. "Yes, yes, that is what will happen if you charge," she said. Her point was not that paywalls are wrong but that the traffic effect is the expected outcome of restricting access - and that framing it as something AI search has caused misdiagnoses the situation.

Her broader argument for small publishers was that the current environment is, in her view, an opportunity. According to Reid, the thing that makes a small publisher valuable - a local focus, a specialist perspective, knowledge that is genuinely hard to find elsewhere - is precisely what distinguishes it from content that AI can summarize adequately. She cited the example of town meeting transcripts: information that is not on the web in any accessible form, that local residents actually need, and that a local outlet is uniquely positioned to provide.

Jarvis, drawing on his own background in journalism and media, pushed the point further. He argued that small publishers should think of themselves as data repositories rather than article factories - scanning restaurant menus, transcribing local government meetings, becoming the canonical source for a defined geography. Reid did not dismiss this framing. According to Reid, when content is otherwise inaccessible, people come. Brand visibility that results from that kind of distinctive sourcing is also, she suggested, something Google is trying to surface more explicitly in its products.

Filter bubbles and what personalization actually does

Jarvis raised the filter bubble question, referencing Eli Pariser's 2012 book and subsequent research. Reid offered a counterargument. According to Reid, the product tries to show a user both their preferred sources and the default top results, not one or the other. Opting into a preferred source does not close off access to other information - it layers on top of it.

More directly, she argued that refusing to personalize actually creates filter bubbles, not prevents them. If Search does not connect a user to the source they trust, they simply navigate directly to that source, bypassing any exposure to alternatives. Showing trusted sources within a result that also includes other perspectives gives people a wider view, not a narrower one. According to Reid, AI Overviews are also designed to present multiple points of view on contested questions, with links to different positions, rather than collapsing to a single answer.

She acknowledged the tension but did not frame it as resolved. "People at the end of the day have to decide for themselves what they trust," she said.

Content quality and what publishers can do

Howell asked what publishers can do to remain visible in a search environment increasingly mediated by AI. Reid answered in two parts. First, make content accessible: blocking crawlers eliminates the possibility of surfacing in AI features. Google's Search Console provides controls publishers can use, but restricting access is the first and most consequential choice. Second, produce content that people actually want to read. According to Reid, the more that content brings expertise, freshness, and genuine specificity - rather than being engineered for search engines rather than readers - the more it will work. She referenced updated guidelines Google has published for website owners and publishers.

On the distinction between AI responses and deeper engagement, Reid drew a line that practitioners in search have been trying to locate. The user who would previously have read only a headline before navigating away may now get what they wanted from an AI Overview. But a user preparing to spend ten minutes on a long-form article will not regard an AI summary as a substitute. "If you're going to go spend 10 minutes reading the article, you're probably not going to decide that an AI overview is a substitute for the 10-minute article," she said. The click that AI Overviews absorb, in her framing, is the shallow engagement click - not the deep-read click. Independent research on publisher traffic presents a more contested picture, with Chartbeat data showing small publishers lost 60% of search traffic in two years while AI chatbot referrals remain below 1% of total page views.

Metrics and AI impressions

The interview addressed how publishers can verify that personalization is working for them. Reid noted that Google has introduced AI impressions in Search Console - a metric that shows how often content appears in AI-generated features. She said the team is actively working on what additional metrics would be useful. She was direct about how early the product is: AI Mode is one year old, AI Overviews two years. "We're still learning," she said.

On the publisher side, she encouraged tracking engagement quality rather than raw clicks. According to Reid, AI-mediated clicks tend to have lower bounce rates than the clicks they partially replace, because users arrive knowing more about why they are clicking. Whether a visitor converts to a subscriber, returns repeatedly, or engages with the content are the measurements she said publishers should be building on their own side.

The "ignore" bug and how Google responds

Howell asked Reid where she was when news broke that typing the word "ignore" was disrupting AI Overviews. The incident, which occurred in the weeks before the June 26 recording, had caused AI Overviews to behave as if the word were a command rather than a dictionary term. Reid described the standard response process: a team is assembled, the nature of the problem is diagnosed - bug or systemic quality issue - and work begins on a fix.

According to Reid, the underlying context for the ignore incident was actually an expansion: Google had been opening definitions and explanations of a much longer list of words in many more languages than had previously been supported. The system encountered a conflict between a new word it had been taught to explain and a term that the AI architecture treated as an instruction. "The system thought it was a command instead of a definition," she said. The bug was found and fixed.

She was clear that higher-stakes queries receive additional scrutiny before launch - the category she called "your money or your life" queries, covering medical and financial topics. The ignore case was low-stakes in consequence, even if widely noticed, because the harm from a disrupted dictionary definition is limited compared to a malfunction in health or financial guidance.

Agents and what search is becoming

Toward the end of the interview, the hosts raised the question of agentic search: what happens to the concept of "searching" when an agent calls a salon and books an appointment without a user typing anything? Google has been expanding search agents, making them available across all AI Mode languages for Google AI Ultra subscribers as of June 12, 2026, with a broader rollout described as coming later in the summer.

Reid's answer resisted the framing of search-versus-agents as a clean break. According to Reid, the hair salon example still begins with a search: the user specifies what they need, and the system gathers highly dynamic information - not necessarily indexed on the web - to help them select and confirm. The distinction between searching, acting, and verifying has always been blurry from a user's point of view, she argued. The product is trying to reduce the friction between those substeps rather than treating them as separate products. "For users, it's all blurry," she said.

What this means for search marketing

The interview adds texture to a picture PPC Land has tracked throughout 2026: Google's head of search is consistently more optimistic about the publisher and small creator opportunity in AI search than third-party data tends to support. Reid's argument - that personalization and Preferred Sources together can surface specialists who were previously invisible - is structurally coherent. SISTRIX data from March 2026 found that AI Overviews cut the click-through rate at position one from 27% to 11% in Germany, costing the market 265 million organic clicks per month across 100 million keywords. Index Exchange data from 69% of publishers on its platform showed year-over-year declines in ad opportunities throughout 2025, with an average decline of 14%.

The Preferred Sources mechanism Reid described is real and measurable - a 2x click-through lift for labeled links is a meaningful signal. But it is a mechanism that rewards publishers who already have an audience willing to take a deliberate action inside Google settings. For publishers who do not yet have that audience, or who are trying to build one, the tool presupposes the relationship it is designed to strengthen. That is the gap at the center of the current publisher-platform dynamic, and the interview did not resolve it.

Timeline

  • June 26, 2025 - Google announces Preferred Sources as a Search Labs experiment in the United States and India
  • August 12, 2025 - Google launches Preferred Sources in Top Stories with formal technical documentation
  • October 2025 - AI Mode reaches 75 million daily active users; AI Overviews drive 10%+ additional queries globally
  • December 10, 2025 - Google expands Preferred Sources globally across all supported languages
  • February 4, 2026 - Ahrefs publishes research finding AI Overviews correlate with a 58% reduction in click-through rates for top-ranking pages, based on 300,000 keywords
  • March 1, 2026 - SISTRIX data shows AI Overviews cost Germany 265 million organic clicks per month
  • March 31, 2026 - Index Exchange data shows 69% of publishers experienced year-over-year ad opportunity declines in 2025, with an average drop of 14%
  • May 19, 2026 - Google announces at I/O that AI Mode has surpassed one billion monthly active users globally; queries have more than doubled every quarter since launch
  • May 27, 2026 - Google extends Preferred Sources into AI Overviews and AI Mode, introducing visible badges for user-designated sources and a "Highly Cited" label for article links
  • June 12, 2026 - Google expands Search information agents to all AI Mode languages for Google AI Ultra subscribers
  • June 26, 2026 - Liz Reid appears on AI Inside podcast episode 135, hosted by Jason Howell and Jeff Jarvis

Summary

Who: Liz Reid, VP and head of Google Search, speaking with hosts Jason Howell and Jeff Jarvis on episode 135 of the AI Inside podcast. Reid has been at Google for 23 years, spent 17 years leading Google Maps, and now runs the Search product used by billions of people daily.

What: A 75-minute interview covering how users are shifting toward conversational queries faster than Google anticipated, why Reid argues that personalization helps small and specialist publishers rather than marginalizing them, what publishers can do to improve their content's visibility in AI-mediated search, the "ignore" bug that disrupted AI Overviews, the distinction between Gemini and Search, and the emerging role of agents in the search experience.

When: The episode was published on June 26, 2026, two days before this article. The interview was recorded in the weeks after the word "ignore" temporarily disrupted AI Overviews, and after AI Mode crossed one billion monthly users at Google I/O on May 19, 2026.

Where: The AI Inside podcast, available at aiinside.show and on YouTube. The conversation applies to Google Search and its AI features globally, including AI Overviews (more than 2.5 billion monthly users) and AI Mode (more than one billion monthly users).

Why: The interview is significant for the marketing and publishing community because it is one of the few times the head of Google Search has addressed the publisher traffic decline question directly and at length, offered a structural argument for why personalization is an opportunity for niche and specialist content, and acknowledged openly that both the metrics and the opt-out infrastructure for AI search features are still being built.