Agentic AI Mode is just a $250 restaurant finder—for now
What happens when you take an experimental feature, add a few bells and whistles, slap the word agentic on it, and hide it behind a $250 early-access paywall?
You get Google Ultra’s agentic AI Mode, a pricey chatbot-style search upgrade that’s all hype and no bite.
I tested it with a few simple, real-world tasks. The results showed obvious limits, though they did hint at Google’s long-term plans.
I found out the hard way that ‘Agentic’ doesn’t mean ‘agent’
In my first test, I asked AI Mode to find a brunch spot for six people: Saturday at noon, cheesecake required, and reservations a must.
It gave me some options. But when I asked it to make the reservation?
Let’s just say honesty was the only thing I got out of it:
AI Mode deserves credit for not hallucinating. But accuracy alone doesn’t make an AI agentic. Autonomy does.
OpenAI’s ChatGPT Agent showed what real progress looks like when you give AI a virtual computer interface. Now, slapping the label “agentic” on anything less feels hollow.
Test take two: The silent update
A few days later, I reran the test and noticed something different.
AI Mode now checked availability, organized results into categories, and even launched an asynchronous “task” that pinged me with later options:
It took a few minutes, but new Mode eventually produced restaurants with actual reservation slots.
That’s progress, but it still fell short of action. New Mode can surface reservation slots, but handed the booking back to me.
Not quite ChatGPT Agent, but not nothing either. The changes were real. They just didn’t change the fundamentals.
Old Mode versus agentic AI Mode
When I tested AI Mode back in April, it was just as underwhelming to me then as it is now. But it was already shying away from old-school search and inching toward something more agentic. Today, that shift is more visible. Here’s what’s different:
| Feature / Behavior | Old AI Mode | Agentic AI Mode (Ultra, $250) |
| Output style | Straight list of restaurants | Structured categories, browses real-time availability |
| Reservation handling | No direct reservation info | Surfaces available time slots from integrated systems (e.g., P.F. Chang’s, with multiple time options) |
| Task flow | Immediate results, but static | Async “Task started” workflow, sends notification when results are ready |
| Reasoning | Minimal, just names and links | Walks through constraints like reservations, cheesecake availability, hours |
| Tone | Flat, list-dumping | More cautious, hedging (“options may vary,” “call ahead”) |
| User experience | Quick but shallow | Feels more systematic, but still requires user action to book |
| Overall impression | Basic restaurant guide | Fancier restaurant guide with reservation data. Still no real autonomy |
The core difference? Old Mode was more Yelp Digest than agent. The agentic version at least acts like it’s solving a problem.
But Google still hasn’t crossed from retrieval to facilitation. These “agentic” upgrades feel cautious, especially next to semi-autonomous systems like Manus or GPT-5’s Agent Mode. With all its resources, Google could have gone bolder, but it chose restraint.
Why Google’s slow push to a truly agentic AI Mode matters
Because it’s inching toward autonomy.
As Sundar Pichai put it at Google I/O 2025: “We think of agents as systems that combine the intelligence of advanced AI models with access to tools, so they can take actions on your behalf and under your control.”
Right now, finding restaurant reservations is as deep as new Mode goes, but Google plans to expand to local service appointments and event tickets.
So yes, the search giant is shifting its focus from answering questions to guiding actions. It’s just moving at a snail’s pace.
AI Mode’s iterative agentic progress comes down to the numbers
It’s an experiment with an agenda.
Sundar Pichai told investors in July 2025 that AI Mode has “over 100 million monthly active users” in the U.S. and India. By comparison, AI Overviews has over 2 billion monthly users worldwide—largely because it’s not optional.
SimilarWeb data shows AI Mode usage hovers around 1-2%, up 25% from earlier 2025. But growth has stalled since May.
So when you connect this stalling graph to the fact that over 50% of users try AI mode once and never return, what you have is an experiment with high initial buzz and uncertain long-term adoption. It’s obviously meant for curious testers like me—marketers and tech pros.
Google isn’t denying it either. Leadership has emphasized a slow, cautious rollout of AI Mode.
VP Liz Reid said it clearly: “AI Mode is where we’ll first bring Gemini’s frontier capabilities, and it’s also a glimpse of what’s to come.” As Google gathers user feedback, it “will graduate many features and capabilities from AI Mode right into the core Search experience.”
This tracks, because from where I’m standing, it’s still not front and center, even with the update. It sits on the side of my Google Ultra desktop account as an optionable feature:
And on mobile too:
It’s also light on ads, a feature Google supposedly integrated into AI Mode in May. I haven’t encountered any yet, which tells me something.
Google is protecting its ad model.
By keeping AI Mode visible but tucked away, Google can balance two priorities: innovation and revenue.
Garret Sussman from iPullRank said it best: “If too many users shift from traditional search to AI Mode, and it doesn’t drive the same revenue, that’s a problem.”
So sure, Google will bust out more Ads in AI Mode, but it’s cautious. Google needs to keep dolling out more agentic “upgrades”, watch how people use them, and then overflow the successful ones into AI Overviews and eventually core search.
The bottom line
Agentic AI Mode isn’t worth $250 for most users. It’s not a true agent. It’s a more polished restaurant guide with experimental workflows baked in.
But writing it off, especially if you work in SEO, is a bad idea. Each of AI Mode’s iterations is a warning shot. One that says, “The future of search is less about blue links and more about action flows.”
So even though Google is still playing it safe, that doesn’t mean you should. Once the economics line up, expect “agentic” to become mainstream in search engines. Make sure your content and data are ready for that shift.
Be visible in the agentic era
Get ahead of the curve by tracking your AI Mode visibility with SE Ranking.
