The Data Tells a More Interesting (and Harder) Truth About AI Search
There’s a lot of loud talk right now about “GEO,” “LLM optimization,” and how AI is supposedly about to replace Google entirely.
That story sounds exciting. It also sounds profitable if you’re selling services.
But when you step away from Twitter threads, conference stages, and sales decks and look at actual usage data, a very different picture emerges.
Not collapse. Not replacement. But compression.
And compression is far more dangerous for small businesses than disruption.
Google Isn’t Dying. It’s Being Used Less Per Person.
One of the most important reports to come out recently wasn’t from an AI company. It came from Datos and SparkToro, based on clickstream data from tens of millions of U.S. users.
The headline finding should make every business owner pause:
Google desktop searches per U.S. user dropped nearly 20% year over year.
That’s not a rounding error. That’s a behavioral shift.
And here’s the nuance most people miss: Google isn’t losing users. It’s losing repeat searches.
People are still going to Google. They’re just asking fewer follow-up questions because more queries are being resolved faster.
In Europe, that same metric only fell 2% to 3%, which tells us this isn’t about search being obsolete. It’s about how answers are being delivered, especially in the U.S.
AI Didn’t Replace Search. It Short-Circuited It.
This is where the AI conversation usually goes off the rails.
Despite all the hype, AI tools still account for less than 1% of total U.S. desktop activity (roughly 0.77%), even after strong year-over-year growth.
Google’s own AI Mode? About 0.06% of U.S. desktop events by the end of the year.
That’s not takeover territory.
What is happening is more subtle and far more impactful.
AI is being layered into search, not pulling users away from it.
Users are:
- Asking longer, more specific questions
- Getting answers faster
- Stopping sooner when satisfied
The data shows that mid-length queries (six to nine words) are growing the fastest. Very long queries (fifteen words or more) remain rare but volatile, a signal that people are experimenting with more natural language.
Translation: users are getting better at expressing real intent, and they are doing fewer retries when the system answers them well.
Zero Click Didn’t Explode. It Settled.
Another myth floating around is that zero click searches are exploding out of control.
They aren’t.
Zero click searches remain high but stable, leveling off in the low twenty percent range.
What changed isn’t how often people click. It’s how many chances you get to earn a click.
Fewer searches per user means:
- Fewer impressions
- Fewer long tail opportunities
- Fewer second and third chances to be discovered
This is where small and mid-sized businesses feel the squeeze first.
Discovery Is Concentrating, Not Expanding.
If AI were truly democratizing discovery, we would see new sites consistently breaking through.
We don’t.
Post-search destinations are largely unchanged:
- YouTube
- Amazon
- Wikipedia
One notable mover is ChatGPT, which climbed to No. 7 among U.S. search destinations.
But here’s the uncomfortable truth:
AI-driven traffic overwhelmingly flows to established platforms, not to independent publishers or small business websites.
This is not a gold rush. It’s a tightening funnel.
The Part No One Wants to Admit: Classic SEO Still Matters
Here’s where the conversation needs to be more honest.
The things that have always been bedrock principles of SEO are the same things that help AI retrieval systems today.
Clean crawlable pages. Clear language. Exact matching between a question and an answer. Strong topical focus.
Especially long tail, high-intent queries.
When someone asks a very specific question, and your page answers that question directly, clearly, and without fluff, you are helping both traditional search engines and AI retrieval systems do their job.
This is why exact match content has quietly become more valuable, not less.
Not because it “optimizes an LLM,” but because it reduces ambiguity for retrieval systems that must decide what to fetch, summarize, and trust.
But Here’s the Hard Business Question No One Asks
Just because something can work doesn’t mean it’s worth doing.
Every business now has to ask a harder ROI question:
Is it worth creating content in my industry to compete for shrinking discovery real estate?
In some industries, the answer is yes. In others, the cost of producing truly authoritative, defensible content may outweigh the returns.
AI has not eliminated content competition. It has intensified it.
Fewer opportunities. Fewer clicks. Higher expectations for quality.
This means content is no longer a volume game. It’s an investment decision.
Some businesses will win by owning very narrow, very specific problems. Others may be better served investing elsewhere.
So Why “LLM Optimization” Is the Wrong Mental Model
LLMs are not search engines. They do not crawl your site. They do not update in real time.
They are answer generators fed by retrieval systems.
Those systems reward:
- Eligibility
- Clarity
- Accuracy
- Association with real problems
There is no black box trick to game this.
Why I’m Saying This Out Loud
I’m Brian Galvan. I’ve spent my career operating at the intersection of marketing, technology, and business operations. I see where traffic actually goes, where it stops, and why businesses feel invisible even when reports say everything is “fine.”
Right now, too many business owners are being sold fear, acronyms, and expensive solutions that don’t match reality.
You deserve data, not dogma.
Have Questions About Whether This Is Worth It for Your Business?
If you’re asking:
- Whether long tail content is still worth investing in
- How AI search affects your actual revenue, not vanity metrics
- If your site is even eligible for modern retrieval systems
- Whether content creation in your industry has a positive ROI
Call me, Brian Galvan.
No hype. No buzzwords. Just a clear conversation about what matters, what doesn’t, and whether the effort is worth the return for your specific business.
