Google's Ads Decoded podcast on October 28 published a 25-minute episode in which Shivani Mohan, Search Vice President of Data Science and User Experience Research at Google, said that query length "has increased 3x" and that "one in every six searches is multimodal." Host Ginny Marvin, Google's Ads Product Liaison, closed the episode by stating that ads in Google's AI search surfaces are matched to "both the query and the content of the AI response," and pointed retailers to six optional product data fields. Neither figure came with a market, a time period or a baseline, and both match numbers Google published in May with narrower qualifiers attached.

In Short

A Google executive who leads research on how people use Google Search went on Google's advertising podcast and said people now type much longer questions and often search with pictures, video or voice. That matters to businesses that buy ads, because Google says its ads in AI answers are picked using the whole AI answer, not just the words someone typed. The two big numbers in the episode were stated more broadly than when Google first published them in May, so they describe a narrower slice of searches than the episode suggests.

The episode

The episode, titled "How AI is changing Google Search behavior and what it means for advertisers," appeared on the Google Ads YouTube channel on October 28, 2026. The channel lists 925,000 subscribers, and the video had 3,753 views by October 31. The Ads Decoded by Google Ads playlist now lists 26 episodes.

Marvin opened by flagging a change of subject. "We are doing something a little different today and getting out of the advertising world and into the organic search world," Marvin said. The stated focus was how search behavior is shifting with the rise of AI Overviews and the rollout of AI Mode, which Marvin told listeners would be referred to as AIO and AIM in the conversation.

The YouTube description frames the subject in behavioral terms. "AI is expanding what people bring to Google Search - whether that's asking 3x longer questions, going back and forth on follow-ups, or searching with voice and images," according to the page. It adds that "people are working through complex decisions right as they search," and describes the final segment as Marvin considering "how to translate these insights for your Google Ads strategy."

Google launched Ads Decoded on August 29, 2025, with Marvin as host. Recent editions have been about products: on October 1 a Demand Gen product manager called $100 a day the floor for a Demand Gen campaign, and on October 15 a Google Ads product manager said target or budget changes cause no learning period. This edition carried no product news at all. It consisted of claims about how people search, followed by Marvin's reading of what those claims mean for an ad account.

Who Shivani Mohan is

Mohan described the remit directly. "I lead the data science and UXR teams in Google search on the consumer side," Mohan said. The route into the job was unconventional. Mohan began in industrial design, as a bicycle designer, a field in which "you get one shot to really nail the user understanding because once you build the product, you can't really iterate on it."

Mohan split the work into two halves. User experience research relies on one-on-one interviews in which researchers "ask them a lot of questions to be able to unearth why they have certain ideas, why do they actually want to ask questions in a certain way." Researchers also watch how people "interact with, let's say, with a coach or with a mentor," and try to carry what people value in those exchanges into Search. Data science covers the other half. "Data science is about actual observed behavior on the platform," Mohan said. "So you're looking at logs of how people are behaving, and then you're able to quantify it to be able to generate a truthful picture of how users are behaving today."

The name is not new to PPC Land readers. Google's May 19 report marking a year of AI Mode was published by Mohan, with figures on query length, follow-ups and multimodal use. That report is the most likely origin of both numbers in this episode.

Three times longer, one in six multimodal

Marvin asked what Google means when it calls the current period an "expansionary moment" for search. "We're absolutely seeing this as an expansionary moment," Mohan answered. The explanation rested on friction. People have always had questions, according to Mohan, but "a very small fraction of those have historically been able to be easily translated and verbalized to be able to ask in a search engine." The step of turning a curiosity into a keyword is, in Mohan's account, gone.

"So now we're seeing that people are asking much longer, much more detailed questions," Mohan said, adding that "we can see that the query length has increased 3x." Those questions are not simply replacing keyword searches, according to Mohan: "They're asking net new types of queries that they would have never thought to ask before."

Two further shifts followed. People are now asking "a question about an image or a video," which Mohan said was harder to do before. And the conversational format lets them follow a topic over several turns, "in a stateful conversational way." The outcome, Mohan said, is that "people are just asking more questions because it is much more natural and frictionless." The YouTube description also mentions voice; Google made Search Live, its voice and camera mode, available to all US users in English on September 24, 2025.

The second number came near the end, in answer to Marvin's question about what an agency lead should change in a search strategy the next morning. "One in every six searches is multimodal," Mohan said.

Marvin repeated both figures in the closing summary: "with queries in conversational search running three times longer on average and one in six searches being multimodal, users are relying on Google for discovery well before they are reaching their final purchase decisions."

What the May data said

The episode did not say what the 3x figure is compared against, which market it covers or over what period it was measured. It did not define multimodal either. Google's May 19 report was more specific on each point. As PPC Land reported at the time, the report put the average AI Mode query at three times the length of a traditional search query, said that more than one in six AI Mode searches in the United States were multimodal, and said that follow-up queries in AI Mode had grown by more than 40% a month on average in the US. The behavioral data in that report covered US users.

In May, then, both figures described AI Mode, and the multimodal share was limited to the United States. In October, on a podcast made for advertisers, Mohan said "one in every six searches" and Marvin said "one in six searches." Mohan's sentence on query length opens with a garbled clause in the transcript ("Like even than just search"), which may have been an attempt at a comparison point; the words as published do not make it clear. Marvin placed the 3x figure in "conversational search." Neither speaker said AI Mode or the United States when giving the numbers.

Why does the distinction matter? A share of AI Mode searches in one country describes a single surface that users choose to enter. Read without the qualifiers, the episode's phrasing describes Google Search as a whole. Nothing in the episode supports that broader reading, and the May report did not make it.

Other measurements point in the same direction, if not at the same scale. Robby Stein, VP of Product for Google Search, said in July 2025 that visual searching on Google had grown 65% year over year. The State of Search report from Datos and SparkToro for the fourth quarter of 2025 found the fastest US growth in searches of six to nine words. Since September 24, 2026, Search Console has split web results into text-based and multimodal search types, but the queries behind multimodal searches are not shown to site owners.

How Google says it checks its numbers

Marvin asked how Google derives the statistics it publishes. Mohan described two tests. The first is repeatability: "if we were to run the analysis several times, we would get the same result." The second is durability. "So we don't want to be able to put out any number that is at a peak point or is something that is liable to change," Mohan said, adding that published figures are "tested very strongly for durability."

Those tests govern how a number is produced. They say nothing about how it is restated afterwards, and the episode described no sample, date range or market for either of its two figures.

More searches, or failed searches?

Marvin raised a question that comes up often: do people return to Search more because they are going deeper, or because the first answer fell short?

Mohan said it was the former, describing people "coming to ask more and more new and novel questions, not so much that they're coming back because they didn't get the original question answered." An internal metric was offered as evidence of how Google treats the alternative. "Actually, in our data science practice, we penalize queries, which we call man refs, manual refinements, because we believe that we want to be very attuned to the original user intent," Mohan said. If a user has to type extra words to get the answer the first query should have produced, "we don't see that as a good signal."

The illustration was a weather query. Someone asks about the weather in San Francisco, then follows up with "What about Austin?", a query that would make little sense alone but carries context from the first. Marvin offered an anecdote of a different kind: a friend outside the industry who could not remember the last time they typed a three-word search into Google, and who said most searches were now voice.

The line between a refinement and a fresh follow-up carries most of the weight in Mohan's answer, and the episode did not quantify it. There was no share of follow-ups classified as manual refinements, no trend for that metric and no explanation of how a follow-up is put in one category or the other. As reported in May, the 40% monthly growth in AI Mode follow-ups measured volume; it did not separate refinements from new questions.

Mohan's account echoes one given by Liz Reid, Google's head of Search, in April. On Bloomberg's Odd Lots, Reid described people stating their real problem instead of translating it into what a computer understands.

Authoritative sources and human voices

Marvin then turned to the open web, noting "a lot of discourse in the industry about how AI and transformation is impacting the open web." Mohan's answer was about what users want. "I think a lot of people are still very clear that they want to hear authoritative voices and authentic human voices," Mohan said. People see a summarized theme and then want to dig deeper: "What are the voices on podcasts? What are the voices on YouTube?" The conclusion was that "users want both," the summary and the sources behind it.

Mohan described two modes of use. In one, people ask follow-up questions and, through that deep dive, receive links "much more tuned to exactly what they're looking for." In the other, they want a "lean back experience," with content presented "in thematic ways" as "tranches of interesting content." In that mode, according to Mohan, "there's an enormous opportunity for the ecosystem to continue to play."

Asked which resources people gravitate toward, Mohan described four patterns.

  • One right answer. Users "want to be able to see the actual original source of the content to be able to gain confidence and trust."
  • Comparisons. When choosing between two products, users "really don't want to just see the summary," and look instead to podcasts, review sites and the experiences of people "who are exactly in their circumstance."
  • How-to questions. Users "really appreciate videos which kind of walk them through the how-to scenarios."
  • Specialized content. "We're seeing a move away from the more generic set of content to much more specialized detailed content."

What the click data shows

Mohan's account describes demand for sources. It does not describe how often people click through to them, and the episode gave no figure for traffic sent to websites from AI Overviews or AI Mode.

Measurements published by others have pointed to fewer clicks. A randomized study by Saharsh Agarwal of the Indian School of Business and Ananya Sen of Carnegie Mellon University found that AI Overviews cut outbound organic clicks by 39.8% among 1,065 desktop Chrome users, while zero-click searches rose 34.5%. Ahrefs reported in February 2026 a 58% reduction in click-through rates for top-ranking pages where AI Overviews appear; in July, Google's Nick Fox said AI features in Search were sending billions of clicks to websites every week, without a comparison period. Chartbeat's data put Search at 12% of North American publisher pageviews in the third quarter of 2026, down from 21% a year earlier.

Google has adjusted how links appear in the meantime. On October 15 it moved inline links higher in AI responses and began testing a section for official sites. None of this came up in the episode.

Up the funnel

Mohan said AI is also changing where queries sit in a purchase. "What we're seeing more and more of with AI is that people's queries are kind of moving up the funnel," Mohan said. People are "asking a lot more brainstorming questions, a lot more questions on how to think about a space altogether," and Google, in Mohan's description, now operates "from discovery to evaluation to purchase."

Marvin drew the advertising conclusion: advertisers should recalibrate how they think about search "outside of last click action," toward the discovery, research, awareness and consideration phases.

That framing sits awkwardly beside third-party data on where AI search ads appear. Adthena's April 2026 report, covering 29.1 million queries, found that more than 60% of ad appearances in AI Overviews were on three- to four-word searches. Short queries are exactly what Mohan described as giving way to longer ones. Whether ad placements have followed the shift is a question the episode did not take up.

How the ads are matched

The most direct advertising statement came in Marvin's closing segment. "Remember that ads in AI Mode and AI Overviews are matched to both the query and the content of the AI response. So the entire context as a whole," Marvin said. "Your ad should position your business as the natural next step in that user journey."

Google has described the mechanism before. In a backstage interview recorded after Google Marketing Live in May, Chris Monkman described Gemini using both the query and the full AI Mode response to decide which ads are relevant, and argued that manually chosen keywords would rarely match such queries. The same coverage noted that the new AI Mode ad formats require AI Max for Search or Performance Max with text customization.

The placements have accumulated over two years. Ads in AI Overviews went live for mobile users in the United States in October 2024 and reached 11 more countries on December 19, 2025, through a help page update. Stein confirmed experiments with ads in AI Mode in an interview recorded on October 30, 2025. Google formally announced shopping ads in AI Mode on February 11, 2026, and at Google Marketing Live on May 20 it introduced Conversational Discovery ads and Highlighted Answers for testing in AI Mode in the United States.

Independent measurement of AI Mode's commercial layout is sparse. A Productrise study of more than 100,000 US and UK queries in July found that AI Mode showed roughly 95% fewer product listings than standard search, with products on about 23% of shopping queries against about 88% in standard results.

The episode offered no performance data for ads in either AI surface: no click-through rates, no conversion rates and no cost comparison with ads on the standard results page.

Six optional fields

For retailers, Marvin pointed to Merchant Center. "If you have a Merchant Center account, don't forget to explore the new conversational attributes," Marvin said, describing them as "six optional product data fields that are there to help you provide more nuance that's needed to answer highly conversational shopper queries." Marvin added: "We've discussed that in our holiday ROI episodes."

The attributes have a short history in Google's own communications. In an Ads Decoded retail episode on April 8, 2026, conversational attributes were described as a new category of product data fields in pilot, and Marvin said dozens of new attributes were coming to Merchant Center. Google Marketing Live on May 20 introduced conversational attributes as a new Merchant Center data type, and Vidhya Srinivasan, Google's VP and GM of Ads and Commerce, named using them as one of the actions for advertisers. A week later, on May 27, Google introduced AI performance insights in Merchant Center, a report on brand presence in AI Mode, AI Overviews and the Gemini app.

The October episode did not name the six fields, and it did not say whether they affect ad eligibility, ad ranking or only organic product answers. The reference to "holiday ROI episodes" points listeners to earlier editions for that detail.

Marvin's three takeaways

Marvin closed with three points that "stood out" from the conversation.

Discovery. Search is playing a bigger role in discovery, which Marvin said "requires a mindset shift for a lot of us."

Real perspectives. Users want podcasts, video reviews and nuanced insights. "So generic surface level content no longer cuts it, not that it ever really did, but even more so now," Marvin said. For advertisers, this "points to new opportunities for landing pages." Retailers were told to pay extra attention to product feed details, including the conversational attributes, and lead generation advertisers to build out "rich specific content."

Friction. Borrowing from Mohan's discipline, Marvin said customer journeys should reflect "how real people are thinking," because users "naturally gravitate towards the path of least friction." Marvin called it "a really good time" to audit landing pages and the asset mix, moving away from "generic product blurbs" and "high level corporate copy." The closing list asked advertisers to "test expanding your creative assets to capture that upper funnel discovery moment" and to "focus on reducing post-click friction." Much of that work, Marvin acknowledged, sits outside a paid search team's remit, which is why the episode pointed to search, content and conversion rate optimization teams.

"No matter your business model, if you aren't offering this level of depth, you're bound to miss the moments where searchers are discovering, researching, and making up their minds," Marvin said.

Mohan's version, given earlier, was shorter. Asked what marketers could take from user experience research, Mohan said the aim is to align a product "as closely to a very human mental model as possible," and named three tendencies: "Usually, people find the path of least friction," people gravitate toward simpler things, and they gravitate toward things that "provide delight."

Fundamentals, again

Asked how marketers should adapt landing pages and content for deeper questions, Mohan said the fundamentals "thankfully remain the same." "You have to continue to develop really relevant content because our ability to put a lot of these websites into our ranking systems remains very steady," Mohan said. The change described was one of degree. Users are gravitating toward "more specific, more detailed, more unique answers to their specific questions," while generic answers "in this moment in time are not hitting the mark quite as much."

Mohan's summary: "focusing on authenticity, factuality, and uniqueness is going to always be evergreen."

The message matches what Google's ads organization has told marketers this year. In June, Brendon Kraham, Google's VP of Search and Commerce for Global Ads Solutions, argued in a Think with Google piece that foundational SEO remained the basis for AI search visibility.

Why it matters

The episode is a statement of position rather than an announcement. Google's case, put by an executive whose teams measure user behavior rather than ad outcomes, is that AI search is adding questions rather than replacing them, that those questions are longer and more specific, and that ads matched to whole AI responses can meet them. For advertisers, that case underpins the direction Google has pushed all year: away from manually chosen keywords and toward AI Max, broad automation and product feed data read by Google's models.

The evidence offered in the episode is thin for an argument of that weight. The two figures were stated without the limits Google attached to them in May. The claim that follow-ups reflect curiosity rather than failure rests on an internal metric whose values were not shared. And the question of whether people still reach the sources they say they want, which matters to publishers and to advertisers whose landing pages compete for the same attention, went unanswered.

For PPC Land readers who manage Google Ads budgets, the practical gap is visibility. Search Console withholds queries for multimodal searches, Google has not published performance benchmarks for ads in AI Mode, and the six conversational attributes Marvin promoted were not named in the episode. Three times longer than what, and one in six of which searches? The May report answered both questions. The October episode did not.

Timeline

Summary

Who: Shivani Mohan, Search Vice President of Data Science and User Experience Research at Google, and Ginny Marvin, Google's Ads Product Liaison, on Google's Ads Decoded podcast. The content concerns advertisers running Search, Shopping and Performance Max campaigns, retailers using Merchant Center, and publishers who depend on Search traffic.

What: A conversation about how people search with AI features. Mohan said query length has increased 3x and that one in every six searches is multimodal, described an internal penalty on manual refinements, and said users want authoritative and human sources. Marvin said ads in AI Mode and AI Overviews are matched to both the query and the AI response, and pointed retailers to six optional conversational attributes. Google's May data had limited the figures to AI Mode, and the multimodal share to the United States.

When: The episode was published on October 28, 2026. The figures it repeats first appeared in Google's AI Mode report of May 19, 2026.

Where: On the Google Ads YouTube channel, in the Ads Decoded by Google Ads playlist. The behavioral data behind the original figures covered US users.

Why: Google is making the case to advertisers that AI search adds longer, earlier-stage questions and that ads matched to whole AI responses can reach them, a case that supports its push toward automated matching and richer product data, while offering no performance or click data for those placements.