The average AI Mode query in the United States now runs three times longer than a traditional search query. That single number explains why pages built around a keyword and a slow narrative windup are losing ground: when a system has to assemble an answer from your page, burying the actual answer under three paragraphs of preamble makes your content harder to use, not more persuasive.

Google’s own data on how AI Mode is changing search in the U.S., published by Shivani Mohan, the company’s vice president of Data Science and UXR, gives a detailed picture of exactly how search behavior has shifted — and it points to a specific, practical change every content team should be making this quarter.

The query itself has changed shape

AI Mode has passed 1 billion monthly active users globally, and query volume has more than doubled every quarter since launch. The volume is the least interesting part of the story, though. What matters more is that people aren’t just searching more often — they’re asking differently. More than one in six AI Mode searches now includes something other than plain text: an image, voice, or a back-and-forth conversation. Image-based queries are growing more than 40% month over month, and follow-up questions are climbing at a similar pace, with people refining and narrowing a search in a way that looks nothing like the one-shot keyword query that SEO strategy has been built around for two decades.

Google’s own list of the most common first words in AI Mode queries confirms the shift directly: “what,” “how,” “I,” “is,” and “can” now lead the list, not short-head nouns. People are typing questions the way they’d ask another person, not the way they’d type into a search box. That behavioral shift is part of why AI Mode sends a different kind of visitor than your site was built for — the query pattern, the intent, and the expectations all changed at once.

Five verbs replaced the keyword

Mohan’s report groups the new query behavior into five modes, each mapping to a distinct content need and growing at its own rate:

• Explore — open-ended, brainstorming queries, growing 30% faster than AI Mode traffic overall.
• Decide — comparison-driven queries built around phrases like “which of” and “which one,” growing 40% faster than overall traffic.
• Learn — queries aimed at understanding a new concept or exploring a skill or career path.
• Do — planning-oriented queries, from workout routines to travel itineraries to household budgets, growing 80% faster than AI Mode queries overall over the past six months.
• Create — including image-generation requests, which have more than tripled since the start of the year.

None of these five modes maps cleanly onto a keyword. They map onto a task. That’s the real shift, and it’s a bigger one than a ranking-factor update — it’s a change in what a search actually is, which is the same underlying pattern behind why so much existing content no longer performs the way it used to.

Why lead-first structure wins in AI search

This isn’t a new problem, even if the trigger is new. Journalists wrote stories chronologically for most of the 19th century, until the telegraph forced a change. The Associated Press paid telegraph operators by the word, which made a poetic scene-setting opener an unnecessary tax on the newsroom’s budget. In the physical back shop, editors setting metal type needed a reliable way to cut a story from the bottom up to fit a page without losing the facts that actually mattered. The lead-first structure — the answer first, the context after — became the industry standard between 1880 and 1890 for purely economic and mechanical reasons.

Generative search engines today are working under a similar kind of constraint. A system assembling an AI Mode answer is pulling passages, not reading a page start to finish the way a human might. If the core fact, the specific number, or the direct answer is buried three paragraphs deep behind brand storytelling, the system either has to work harder to find it or skips the page for one that made the answer easy to extract. Protecting a vague brand voice by hiding the actual facts inside a soft introduction is no longer a stylistic choice — it’s a fast route to being left out of the summary entirely. None of this requires writing in flat, robotic prose. It requires aligning how a page is structured with how algorithmic synthesis actually reads it.

Five practical steps for AI Mode optimization

1. Lead with entity-dense opening sentences

Put the primary definition, key metric, or main conclusion directly in the first sentence of every core section. Anchor sentences with specific names, clear dates, exact locations, and verified numbers rather than vague generalizations. Specificity makes a passage easier to extract cleanly and reduces the odds of an AI system paraphrasing your point into something inaccurate.

2. Build scannable, hybrid layouts

Keep paragraphs to two or three sentences on average. Follow major headers with a short summary, an ordered list, or a table where the content supports it. That structure lets an AI system extract a clean, self-contained segment for a generated overview while also giving a human reader a page that’s easy to scan.

3. Write for the follow-up question, not just the first click

With multi-turn conversations growing at over 40% month over month, content needs to anticipate the second and third question a reader would ask after the one the headline answered. Structure long-form pieces so that individual sections can stand on their own as the answer to a likely follow-up — that’s the chunk an AI system is most likely to surface next.

4. Brief content around the five verbs, not five keywords

Before commissioning a piece, decide whether it’s meant to help someone explore, decide, learn, do, or create something. A comparison page built for “which” queries needs a genuinely different structure than a how-to guide built for “how do I start” queries, even when the topic and target keyword look nearly identical on paper. This is the same content-strategy discipline behind avoiding pages that accidentally send AI Overviews to recommend a competitor instead.

5. Treat images and multimodal content as a ranking input

With image-based queries growing 40% month over month and image-creation queries in AI Mode more than tripling since January, alt text, image context, and visual quality are no longer a secondary concern. They’re part of what AI Mode is actually reading when it assembles an answer.

What this means for content strategy

The underlying goal hasn’t changed: getting cited, getting read, and getting trusted. What’s changed is the mechanics of how a system decides which page earns that citation. A content brief built around a head term and a handful of related keywords is aging out in real time, not in some future AI-search scenario — it’s happening in the query data right now. The practical response is structural clarity: state the answer plainly near the top of every section, organize the rest to support a follow-up question, and build content around the task a reader is trying to complete rather than the term they typed to start. That approach lines up with the broader shift toward earning citations in AI answers rather than chasing a ranking position that means less than it used to.

Frequently asked questions

Why are AI Mode queries longer than traditional search queries?

People type search queries the way they’d type a keyword into a box, but they phrase AI Mode queries closer to how they’d ask another person a question — with more context, more specifics, and often a follow-up. Google’s data shows the average AI Mode query in the U.S. runs about three times the length of a traditional search query.

What are the “five verbs” in Google’s AI Mode research?

Google’s research groups AI Mode query intent into five modes: Explore, Decide, Learn, Do, and Create. Each maps to a different kind of content need and is growing at a different rate, with Do queries (planning-oriented) growing fastest at roughly 80% faster than AI Mode traffic overall.

Does lead-first writing hurt storytelling or brand voice?

Not necessarily. Leading with the answer or key fact doesn’t require flat, robotic prose — it means putting the most useful information first and using the rest of the section to add context, nuance, or narrative. The goal is structural clarity, not the removal of voice.

How should image content factor into an AI Mode content strategy?

Treat it as a ranking input rather than decoration. With image-based AI Mode queries growing more than 40% month over month, alt text, surrounding context, and visual quality all contribute to whether a page’s imagery gets pulled into a generated answer.