Google says AI Mode queries run three times longer, and discovery now starts before buyers know your product's name

Google says the average AI Mode search query now runs three times longer than a typical search, and the company's own data shows people reaching for AI-powered search earlier in their buying process, often before they know what product category they're even looking for.

The claim comes from Ginny Marvin, Google Ads Product Liaison, who laid out the shift in a LinkedIn post tied to the October 28 episode of Google's Ads Decoded podcast, a Google Ads series. On the episode, Marvin interviewed Shivani Mohan, Google's vice president of data science and user experience research for search, about how search habits are changing alongside AI Overviews and AI Mode. Google published the podcast and a companion blog post the same day, telling advertisers to "adapt your ad strategy to meet the way people search today."

What changed

Marvin's post, as reported by Search Engine Journal, puts a number on something software marketers have mostly been guessing at: how much longer and more exploratory AI-assisted search has become. Two concrete claims stand out.

Search Engine Journal's Brooke Osmundson, who covered the post, flagged that Marvin's wording compares average query length only, and says nothing about intent or conversion, since Google hasn't published the measurement window behind either figure. A narrower Google report from May 2026 had already put multimodal search at "more than one in six" specifically within U.S. AI Mode traffic, not search overall, so Marvin's broader framing should be read with that caveat attached.

The more useful part of Marvin's post is about timing, not just length: people are starting discovery earlier, using conversational search to think out loud about a problem well before they have settled on what to call the solution. Mohan, on the same episode, said users appreciate AI summaries but still want the texture found in forums, podcast conversations, and firsthand video reviews, and that the open web still gets, in her words, "billions of clicks," though she gave no timeframe or comparison point for that figure.

"Appreciate the streamlined clarity of AI summaries, yet they actively crave the rich detail" — Shivani Mohan, VP of data science and UXR for search, Google

Who it affects

This lands hardest on software marketers who built their funnel around people already knowing what they are shopping for. B2B SaaS search campaigns are typically keyword-matched to category or product names, such as "CRM software," "project management tool," or a competitor's name, on the assumption that by the time someone searches, they have already framed the problem in product terms. If AI Mode is pulling discovery earlier, that assumption gets weaker every quarter: the first query might describe a symptom, like why a sales team keeps missing follow-ups, rather than a category.

It also raises the stakes on content software marketers do not fully control. If Mohan is right that buyers still want forums, demos, and firsthand accounts alongside AI summaries, then a SaaS company's presence on review sites, community forums, and video channels is doing real work earlier in the funnel than most attribution models give it credit for, work that a keyword-only paid strategy will not capture or replace.

What to do differently

Osmundson's write-up translates Marvin and Mohan's comments into a short set of changes worth adopting before 2027 planning locks in.

• Mine support tickets, sales call notes, and existing search terms for the symptom language people use before they know your product category, then test ad and landing-page copy that names the problem first and the category second.
• Audit comparison and pricing pages for completeness, including compatibility, setup requirements, and limitations, not just the upside, since AI Mode users appear to arrive already comparison-shopping.
• Tie owned proof, such as reviews, demos, and customer stories, explicitly to the concerns prospects raise earliest, and give each piece a clear next step instead of treating it as pure brand content.
• Stop reporting impressions, click-through rate, or engagement as if they were outcomes. Track whether visitors to comparison or educational pages actually convert over your normal sales cycle, not just whether they clicked.

None of this is Google handing out a new ad format or a reporting feature. It is a data point from the people who build search, delivered on a company podcast rather than a press release, and reported with caveats attached by Search Engine Journal. That combination is worth taking seriously without over-indexing on it: a three-times multiplier on query length is a real signal about how people are using the product, but Google has given no benchmark for what it means for conversion, and no breakdown by industry or buying stage. For a B2B software category with a multi-month sales cycle, earlier discovery is directionally useful and operationally underspecified, which is itself the argument for testing symptom-first messaging in a contained campaign before rewriting the whole funnel around it.