Meta's new AI agent, Muse, is all over the news. I wanted to try it. But Muse is not open in India yet. So I did the next best thing. I woke up Kabeelan from our US team and asked him to try it for me, live, on a call.
We gave it one simple job: buy a pair of running shoes.
What happened in the next 20 minutes changed how I think about websites, SEO and marketing. Let me walk you through it.
Watch Kabeelan and me test Meta’s Muse live, from finding a running shoe to reaching the payment step.
First, why Muse matters
Muse is an AI agent. That word "agent" is the key. A chatbot answers your question. An agent does the task for you. It opens a browser, looks at stores, compares options and buys.
We have seen this idea before. Perplexity has its Comet browser. OpenAI has Atlas. So why do I care about Muse?
Because of reach. Meta owns WhatsApp, Instagram and Facebook. Billions of people already use its apps. And in the same week Muse took off, two big things happened. Amazon blocked Muse from its store. The very next day, Shopify opened every Shopify store to Muse, with Shop Pay for checkout.
Think about that. Shopify is plugged into a huge part of the D2C world. So Meta now has the users, the shops and the payment. This could be the "app store moment" for AI agents.
What we did
Kabeelan signed in and asked Muse to buy him a running shoe.
The first thing Muse did was ask two questions. What is your shoe size? Men's or women's? Fair enough. Kabeelan said men's size 11, mostly for trail running.
Then Muse opened its own browser and started looking.
What we saw
1. The first answer looked like ChatGPT or Google AI Mode
Muse came back with a short list of shoes and stores. REI showed up a lot. So did Road Runner Sports. It looked a lot like what you get from ChatGPT or Google's AI Mode today.
When we clicked a product, there were two buttons: "Buy with Muse" and "Visit website."
Read that again. Visiting the website is now the backup option.
2. It picked a color on its own
Kabeelan clicked "Buy with Muse." Muse started getting the checkout ready. But it never asked which color he wanted. It just picked one.
We asked why. It said "good call" and went back to check the colors. So the agent will fill in gaps with its own guess if you don't tell it. For shoppers, that means you need to give it clear details. For brands, it means your product details need to be complete, or the agent will guess.
3. Stock changed everything
Then came the surprise. The shoe Muse picked was out of stock at REI. Not online. Not for store pickup. Muse told us no order was placed and asked what we wanted to do.
So we pushed it. We said we need this exact Hoka model, and we need it by Tuesday. Figure it out.
4. A new store showed up that was never in the first list
This is the part that got me.
Muse started checking more places. The Hoka website. Zappos. Other stores. It asked for Kabeelan's zip code. Then it came back and said Dick's Sporting Goods has the shoe and it can arrive by the weekend. It also found another store, Public Lands, but that one would only deliver by October 1, which misses Tuesday.
Dick's was not in the first list at all. It only showed up once we gave a real deadline and a location. Stock, delivery date and zip code decided the winner. Not the banner. Not the homepage. Not the ad copy.
5. It went straight to payment
Muse then asked if Kabeelan wanted to pay with Stripe Link. That's the last step. The whole job, from "buy me a shoe" to "ready to pay," happened inside Muse.
What this means for marketers
For 30 years, we have built websites for people. We worked on the design, the homepage, the banner, the call to action, the copy. The whole funnel was built on one idea: get attention, then persuade a human on your site, then show a button, then close the sale.
An agent does not care about most of that. It does not get excited by your banner. It wants facts. What do you sell? What does it cost? Is it in stock? Can it reach me by Tuesday? What is your return policy?
We need to start thinking about "AX"
We have spent years on UX, user experience. I think we now need AX, agent experience.
Here are the questions I would ask about any website today:
- Can an AI agent tell exactly what I sell?
- Can it read my prices clearly?
- Can it see every size, color and product detail?
- Can it see if the item is in stock right now?
- Can it find my shipping time, my return policy and my cancel policy?
- Can it tell if I deliver to this PIN code or zip code?
In our test, the store that won was the store that could answer "yes, in stock, here by the weekend." That is the new first page.
CRO is going to change too
This part gives me mixed feelings as a marketer.
Say you and I both sell the same shoe at the same price. Today, we fight with design, copy, user experience and CRO tricks. Some of that still matters. Offers still matter. If you buy from me and I throw in free socks or extra laces, an agent can pass that on.
But a lot of classic CRO is about guiding a human's eyes and clicks. An agent does not have eyes. So I think we will need something like "agent CRO." Track how the agent moves through your page. Find where it gave up. Ask why it did not add your product to the cart. Maybe your product page was missing the one detail it needed.
This is also where prompt research becomes key. In our test, the shopper gave size, use (trail running), deadline and zip code. Those are the details the agent then looked for. If your product page doesn't cover the things people tell their agents, you won't be picked.
What we still don't know
I want to be honest here. This was one test, on one product, on one account. It's too early to say how Muse ranks stores.
Here's what I want to find out next:
- Where does Muse get its first list? ChatGPT has its own index. Other AI tools lean on search engines. We don't know yet what Muse uses.
- Does paying Meta help? Right now the results looked organic. But Meta has said it plans to make money by taking a small fee when a sale happens. Most of these stores already run ads on Meta. Does that give them a boost? We don't know.
- Can stores track it? When Muse buys, does it add a tag or merchant ID to the link? If it does, stores can measure agent sales. If it doesn't, those sales may just show up as odd traffic in your reports.
Once Muse opens in India, our team will dig into this properly. We will watch every site it visits and compare the results.
Will this last?
I hope this does not become the next NFT. Big buzz, then nothing.
But I don't think the idea of agents is going away, whether Muse wins or not. Muse matters because of Meta's reach. If Meta can't make agents work for regular people with all its users, the agent future may get pushed back a few years. If it does work, the change will come fast. Muse is already coming to Meta's glasses and a small new device called Muse Charm.
How much time did it save?
At the end, I asked Kabeelan a simple question. If you had to do this without Muse, how long would it take?
Think about it. Your favorite store is out of stock. Now you open ten tabs. You check which sites are real and which are scams. You check who can deliver on time. That could easily eat an hour.
Muse did it in minutes. It found who had the shoe, who could deliver by Tuesday, and got right to payment. And the final "yes, buy it" was still ours to say.
The question every business should ask now
The next version of the internet won't be used only by people. More and more, people will give AI agents permission to act for them.
When that happens, your website, your SEO, your CRO, your analytics and even your marketing funnel will look very different.
So stop asking only "how do I convert this visitor?" Start asking "how do I do business with an agent?"
Your next customer may not be a human. Is your website ready to sell to one?
Do you want more traffic?
