Google's AI Mode users write paragraphs, not keywords. Here's what that means for your content.

The data: What Google's own numbers say

Google published its first full year of AI Mode data in May, written by Shivani Mohan, the company's VP of Data Science and UXR for Search. The headline number is scale: AI Mode has surpassed a billion monthly active users worldwide, and query volume has more than doubled every quarter since launch.

The behaviour underneath that growth is the interesting part. The average AI Mode query is three times the length of a standard Search query. More than one in six searches in the US now involve voice or images rather than typed text, and image-based searches are growing over 40% month on month.

Planning queries - the "where should I," "what's the best way to" type of searches - grew 80% faster than AI Mode overall over a six-month stretch. Brainstorming queries grew 30% faster than the average since launch. People are using Search the way they'd use a knowledgeable friend, not a card catalogue.

The shift: From keywords to conversations

For twenty years, ranking well meant matching the words in the query box. Someone typed "best CRM small business," you wrote a page with "best CRM small business" in the title, the H1, and a few times in the body, and if you did it competently, you had a shot.

AI Mode users aren't typing that anymore. They're asking something closer to "I run a five-person agency and keep losing track of client emails - what should I actually use." That's not a keyword. It's a question with context and a specific outcome attached to it.

Google's own examples back this up: the fastest-growing query patterns are "where to," "where should I," and "ideas for" - phrasing built around a goal, not a topic.

The gap: Why most content isn't built for this

Most SEO content still answers the question "what is X," because that's what ranked well under the old model. It rarely answers "should I use X for my situation," or the specific follow-up question a reader would actually ask next.

That's the gap Google's data is exposing. A page can be technically well-optimised - clean headings, decent word count, the right keyword density - and still fail the moment a reader arrives asking a genuinely conversational question the page never anticipated.

We see this constantly with client sites: content that answers the search term instead of the person actually reading it.

The fix: What to actually do about it

Start with your ten best-performing pages and rewrite the core question each one answers as something a real person would actually ask out loud, not the keyword phrase you built the page around. If the gap between the two is wide, that page is optimised for a search pattern that's shrinking.

Treat the follow-up question as part of the content brief, not an afterthought. If someone lands on a pricing page after asking about a service, the next thing on their mind is usually "is this worth it for a business my size" - and that answer should already be on the page, not buried in a separate FAQ nobody finds.

Multimodal search is no longer a niche consideration either. With image searches growing that fast, alt text and image context deserve the same care as your body copy, not a box ticked at the end of a build.

The bottom line: This isn't a ranking tweak

This isn't a new algorithm update to chase. It's a genuine change in how people search, backed by Google's own year of data, and it rewards content written for a person with a specific problem over content written for a search box.

The sites that adapt now get a head start most competitors won't bother taking.

If you want a second opinion on whether your content is built for how people actually search now, get in touch and we'll take a look.

Steve Lavine is a full-stack developer and founder of Lavine Web & AI Solutions, working with SMEs and startups across the UK. lavine.dev