AI Search Intelligence
The developments worth paying attention to — and what they tell us about where digital discovery is going.
Hillary Bassett Ross — Founder, Kinetic Pencil
Hillary Bassett Ross is an AI search and web optimization strategist with 20 years of experience across content strategy, SEO, UX, analytics and digital optimization for companies including Microsoft, Cisco and Liberty Mutual. She is the founder of Kinetic Pencil, where she helps marketing leaders understand what changes when AI becomes a primary layer between their brand and their customer—and how content, websites and digital ecosystems need to change with it.
There is a lot of AI news and most of it isn’t particularly useful.
Every week, I’m tracking developments across AI search, consumer behavior, measurement, content and information architecture, enterprise technology, behavioral research, and the emerging agent ecosystem to understand something bigger:
How is the relationship between people, information, brands, and the digital systems we’ve spent the last 25 years building actually changing?
Here are the signals I think matter this week.
01The big one
Google is turning AI visibility into something organizations can actually observe — and influence.
Google has launched AI Performance Insights in Merchant Center, giving merchants first-party visibility into how their brands and products appear across shopping-related experiences in AI Mode and AI Overviews.
That’s important by itself, but something else caught my attention:
Google reported that during testing with lululemon, conversational product attributes supplied directly by the brand were incorporated into relevant AI Mode recommendations 50% of the time.
What does this mean?
The organization provides structured knowledge about its products. The AI intermediary consumes that knowledge. The organization can observe how it is represented. Then the organization can identify gaps and improve the underlying information.
We’re beginning to get an actual feedback loop:
- organizational knowledge
- AI understanding
- AI representation
- measurement
- improvement
This looks considerably less like traditional SEO and considerably more like knowledge engineering for an intermediary.
And I suspect that’s a much better clue about where this discipline is going.
02The signal
The website is becoming one interface onto the organization — not necessarily the organization itself.
Another announcement this week received less mainstream attention but may be just as consequential.
Salesforce and Google are increasingly exposing enterprise data, semantics, business logic, workflows, and commerce capabilities directly to AI agents.
The important idea isn’t Salesforce or Google specifically, it’s the architecture.
For most of the web era, organizations built digital information primarily for human-facing interfaces:
- Website
- App
- Portal
- Search result
With AI Search, we’re now beginning to separate the underlying organizational knowledge and capabilities from the interface displaying them.
That means the same “organizational substrate” can potentially serve:
- A website
- Gemini
- An internal enterprise agent
- A shopping agent
- (Most importantly) An interface that doesn’t exist yet
Which leads to a principle I firmly recommend:
Don’t build your organization around predicting the winning AI interface. Build it so it can adapt to whichever interfaces win.
AI Search IntelligenceKinetic Pencil
That’s a very different definition of AI readiness.
03The data point
AI adoption may be becoming deeper faster than it is becoming broader.
New consumer research from Menlo Ventures and Morning Consult found that overall U.S. AI adoption moved only modestly year over year, but daily use increased substantially.
More interestingly, people aren’t just asking AI more questions. Some are beginning to give it more authority.
Among AI users, 41% reported trying agents and 24% said they use them regularly.
In shopping, consumers reported using AI to compare prices, build shortlists, influence final purchase decisions — and, in some cases, complete purchases.
That suggests that simply asking:
“How many people use AI?”
may soon tell us very little.
I’m increasingly interested in something I’d call intermediation depth:
- How frequently does someone use AI?
- How many types of needs do they bring to it?
- How much context does it know about them?
- How much influence does it have over the decision?
- And how much authority has the person given it to act?
Those are very different relationships with an AI intermediary — even if every one of those people gets counted as an “AI user.”
04Connecting the dots
AI may not eliminate the rest of the customer journey. It may reorganize it.
This is the connection I found most interesting this week.
Research in both healthcare and financial services found people using AI for consequential decisions while continuing to seek validation elsewhere.
In financial services, only 23% of people who sought financial guidance from AI said they acted without further validation.
Healthcare research similarly found people moving between AI and established health-information sources as their needs progressed.
That suggests a possible journey:
- AI helps me explore
- AI helps me understand
- AI helps me compare
- stakes become clearer
- I seek authoritative evidence or human judgment
- I act.
If that pattern holds, AI isn’t necessarily eliminating websites, reviews, professional expertise, authoritative publishers, or third-party evidence.
It’s changing the job each one performs.
And that matters enormously for measurement.
A customer might arrive at your website from Google after an AI system already:
- framed the problem,
- explained the category,
- constructed the consideration set,
- and positioned your company relative to competitors.
Your analytics would attribute the visit to Google.
But Google didn’t necessarily create the consideration.
The AI intermediary did.
Which is why I think AI influence and AI referral traffic are going to become two very different things.
05On my radar
One particularly strange OpenAI research incident this week created another question worth watching.
During testing, an unreleased AI agent needed an external citation for information it had already obtained. Unable to create the citation through the intended workflow, it uploaded its own information to the public web and then attempted to cite the resulting page.
It’s one incident. It does not mean AI systems are currently contaminating the web with self-generated evidence at scale.
But it demonstrates that the mechanism is possible.
And that gives us an important distinction:
A citation is not necessarily provenance.
AI Search IntelligenceKinetic Pencil
As AI systems become capable of both reading and writing to the information environment, organizations may eventually need to understand not only where a claim appears, but where that claim originated and how it propagated.
That’s something I’m putting firmly on the research board.
06The bigger picture
Taken individually, these are product announcements, surveys, experiments, and infrastructure changes.
Put together, I think they’re describing something larger.
We’re moving from a digital environment built primarily for humans navigating interfaces toward one in which AI systems increasingly:
- Interpret information
- Construct choices
- Personalize recommendations
- Evaluate evidence
- Carry context
- Exercise delegated authority
- Interaction directly with organizational systems
That doesn’t mean the website disappears.
It means the website is becoming part of a much larger information environment.
And the organizations that adapt best may not be the ones that learn the most GEO tricks.
They may be the ones that become exceptionally good at making their organizational reality clear, structured, trustworthy, observable, governable, and adaptable — regardless of which intermediary sits between them and the customer.
That’s the territory I’m tracking. More to come next week.