About a year and a half ago, I started stress-testing ChatGPT search functionality and Google’s AI Overviews. Not as curiosities. As systems that might change how my clients get found.

I asked real buyer questions. The kind a homeowner types at 11pm when their furnace stops working. The kind a parent searches when their kid got hurt on the soccer field.

The answers came back different.

Not ranked. Summarized. Not neutral. Contextualized. The system wasn’t saying “here are ten things.” It was saying “here’s what’s good, and here’s why.”

The moment a system explains why it chose one option over another, it stops acting like an index. It starts acting like an assistant.

That shift, from retrieval to recommendation, isn’t an algorithm update. It’s a change in role.

The Assistant Does the Searching Now

For twenty years, our job was clear: get clients into Google’s top ten results. We knew the interface. We knew the rules.

And most businesses did exactly what they were told to do.

That model assumed humans did the work. The searching. The comparing. The pruning of bad fits.

Now the system does that work on the user’s behalf.

And it doesn’t search the way people did.

It pulls from websites, business listings, reviews, LinkedIn posts, Reddit threads, press mentions, and whatever else it can treat as evidence. Even Google’s AI isn’t just “looking at Google” anymore.

According to research tracking over 300,000 keywords, AI Overviews now appear in roughly half of all Google searches.

Half of consumers polled in a McKinsey survey now intentionally seek out AI-powered search engines. They say it’s the top digital source they use to make buying decisions.

This isn’t coming. It’s here.

What Compression Means for Local Businesses

Traditional search presented ten options. The user chose.

AI search compresses those ten down to three recommendations. Sometimes one.

Traditional SEO still feeds the system. It’s just no longer the last layer before the user.

When that happens, what gets evaluated?

The assistant leans on other systems—Google, Bing, directories—to surface credible sources. Then it does something different on top of that.

It asks: Do multiple independent sources agree this business matters? Do they agree on why?

That’s where compression happens.

A strong page can win in a directory. A coherent story is what survives synthesis.

Recent data shows AI systems place heavy weight on non-branded sources. Reddit is the single most-cited domain across ChatGPT, Copilot, and Perplexity. YouTube and Facebook follow close behind.

Your website still matters. But it’s treated as self-reported. Useful, but not sufficient on its own.

The Gap Between Good and Visible

Most local businesses I’ve worked with do good work. They have loyal customers. They’re trusted in their communities.

But that trust is often trapped in one-to-one interactions.

It lives in conversations, relationships, and repeat business. It doesn’t automatically become visible at scale.

The gap isn’t between good businesses and bad businesses. It’s between lived reputation and legible reputation.

When assistants do the searching, they ask: “Is there enough consistent evidence across independent sources to justify recommending this business to someone who doesn’t know them yet?”

If everyone loves working with you but nobody talks about it publicly, that love is invisible.

If your story changes depending on where someone encounters you, the assistant hesitates.

Authority Gets Triangulated Now

Authority used to be calculated inside a single system. Link graphs. Page-level scoring. Domain strength.

Now it’s triangulated across many.

The assistant finds the same business showing up across different contexts, saying roughly the same thing. The system gains confidence. It doesn’t have to trust any single source completely.

The agreement between sources does the work.

Research shows that nearly 60% of ChatGPT Search results cite business websites. But those websites are evaluated as part of a larger ecosystem, not in isolation.

The assistant checks multiple sources. Your website is one input among many.

What Changes When You Can’t Control the Page

For a long time, SEO rewarded direct control. You owned the website. You controlled the copy. You could optimize, test, refine, and predict outcomes.

Visibility felt engineerable.

Control isn’t centralized anymore.

The story is still influenceable. Just less directly and usually over longer time horizons.

When the assistant looks at listicles, press coverage, reviews, or community discussion, those aren’t random. They’re shaped by how a business shows up in the real world, how it communicates, what it prioritizes, and what it gives people reasons to talk about.

The mistake businesses make isn’t trying to game the system. It’s trying to apply page-level optimization instincts to a story-level evaluation process.

They ask: “How do I optimize this page?”

The better question: “What story are independent sources likely to tell about us?”

You don’t control what a journalist writes, but you influence whether you’re newsworthy. You don’t control what shows up in a Reddit thread, but you influence whether customers advocate for you.

The shift is from controlling outputs to shaping inputs.

Making Excellence Hard to Miss

When I use the word “system,” I’m contrasting it with the ad hoc advice most businesses hear. “Post more content.” “Be more active on social.”

More activity isn’t a system.

A system produces consistent outcomes even when attention and effort fluctuate.

In practice, a visibility system has a few defining characteristics:

A clear and stable message. Before anything gets published, there has to be clarity about what the business does, who it’s for, and what differentiates it. Without that, pushing content into more places only multiplies confusion.

Repeatable expression across contexts. The same core story needs to show up in different forms and on different surfaces. A website page, a short article, a local mention, a social post, a third-party write-up. Not copied and pasted, but recognizably aligned.

Intentional distribution beyond owned channels. A system creates reasons for the business to appear in places it doesn’t own: industry write-ups, community discussions, local publications, directories, and social platforms. Not because “being everywhere” is the goal, but because assistants look for corroboration across independent sources.

Reinforcement over time. A system pays attention to which messages get echoed, referenced, or picked up elsewhere and then reinforces those themes instead of constantly reinventing itself.

That’s the difference between posting content and building a system. One creates noise. The other creates continuity.

The Selection Pressure Is Visible

What This Changes…and What It Doesn’t

We don’t yet have perfect tracking to prove clean cause and effect. The data is messy. We’re early.

But the selection pressure is visible. It’s pointing in one direction.

Businesses that are strong in one context but misrepresented, outdated, or fragmented in others fail to survive compression. When the system reduces ten options to a handful it can confidently recommend, coherence wins.

Visibility is not guaranteed.

The change isn’t that search is disappearing. It’s that who is doing the searching is changing.

And assistants don’t search the way humans did.

They triangulate. They synthesize. They compress options.

A coherent story, told by others, beats a strong page you wrote about yourself when options are compressed.

That’s not gaming anything. That’s making sure the truth about your business actually travels.