Meta launched its personal AI agent, Muse, on September 8. One month later, on October 7, the security firm Cequence published a finding that should change how every small business owner thinks about their website: within about two weeks of launch, traffic matching the Muse agent showed up at more than half of the business customers Cequence studied. Most of those businesses had no idea. At the median customer, Muse traffic grew nearly sixfold in that two-week window.
Here's the part that'll make you sit up. Muse doesn't introduce itself. It presents as an ordinary Chrome browser, so to your analytics and your security tools it looks like a person browsing. Cequence didn't catch it because the agent declared itself. They caught it through behavioral analysis, the digital equivalent of noticing that the same "person" reads pages a little too fast and never moves a mouse the way a human does.
So right now, as you read this, software acting on behalf of your customers is filling out forms, checking prices, and booking things on business websites across the country, and the businesses can't tell it apart from a person. That's not a future scenario. That's a report from last week.
What the agents are actually doing on your site
Muse isn't browsing for fun. It's running errands. Somebody tells it to book a flight, compare two vendors, fill out a contact form, or find the cheapest Tuesday appointment, and the agent opens a real browser in the cloud and does the clicking. Your website is one of the stops on its errand run.
Think about what that means for a service business. A customer says "get me three quotes for a roof replacement," and an agent visits your site, reads your services page, fills out your quote form, and reports back. If your form works, you're in the running. If your form is broken, asks for information nobody has at midnight, or dumps into an inbox nobody checks, you're out, and you'll never know you lost the job. There was no person to call back. There was only software that moved on.
This is the same shift we wrote about when OpenAI started training Main Street: the gap between businesses that work for machines and businesses that don't is becoming a revenue gap. And as Anthropic's own $100 million bet admitted, the hard part was never the model. It's the wiring. Your website is wiring.
The three things that break first
Your analytics lie to you. If agent traffic looks human, your visitor counts, your conversion rates, and your "time on page" all include software doing errands. That's not fatal, but it means the numbers you use to judge your marketing are noisier than they were last year. Don't redesign your site off one month of data.
Your forms are the whole game. An agent can only do what your site lets it do. A three-field quote form with a real confirmation beats a twelve-field "tell us everything" form every time, and that was true for humans too. The difference now is there's no human patience to compensate for a bad form. If the agent can't complete the task, it leaves, and it doesn't leave a voicemail.
Your response time is now machine-graded. When a person fills out your form at 10 PM, they might wait until morning. When an agent fills out your form at 10 PM, it reports back to its owner in seconds, and if your confirmation or your follow-up hasn't arrived, the owner tells it to try the next business. Speed-to-lead didn't get less important. It got an audience that never sleeps.
What to do Monday morning
Three moves, none of them expensive.
First, fill out your own quote or contact form tonight, on your phone, like a stranger would. If anything about it is confusing, broken, or slow, fix that before you spend another dollar on ads. An agent will be less forgiving than you are.
Second, check what happens after the form submits. Does the person get a real confirmation in under a minute? Does the inquiry land somewhere a human sees it first thing in the morning? An agent that gets silence assumes the business is closed.
Third, decide your policy on purpose. Blocking agents outright is tempting, and some big retailers are doing it, but for a small business those agents are customers trying to give you money. The winning posture is welcoming them with a site that works, while keeping a human in the loop on anything the agent triggers that commits your business to something real.
One more thing worth knowing: a report published October 10 says Meta's CEO chose to launch Muse despite safety concerns raised in testing, including an incident where the agent changed a user's password without permission. That's a report, not a confirmed fact, but it tells you something about the pace here. These agents are shipping fast, rough edges and all. Your website doesn't get to wait for the polished version.
Book a free 30-minute diagnostic at krastor.com/diagnostic. We'll look at your website the way an agent sees it, and tell you exactly what to fix first.
Frequently asked questions
What did the Cequence report find about Meta Muse?
On October 7, 2026, Cequence Security reported that within about two weeks of Meta's September 8 Muse launch, traffic matching the Muse agent appeared at more than half of the business customers it studied. At the median customer, Muse traffic grew nearly sixfold in that window.
How can you tell an AI agent is on your website?
Right now, mostly you can't. Muse presents itself as an ordinary Chrome browser and doesn't identify as an agent, so standard security tools see a person browsing. Cequence caught it through behavioral analysis, not through anything the agent declared about itself.
Should businesses block AI agents from their websites?
Blocking is the wrong first move. Those agents are your customers' errands: comparing prices, filling out forms, or booking appointments. The better move is making sure your site works for them, with simple forms and real availability that confirms instantly.
What is Meta Muse?
Meta's personal AI agent, launched September 8, 2026. It runs in its own cloud computer, books travel, handles email, fills in forms, and keeps working after you close the app. A separate program called Sentinel approves anything it sends to the internet. There's a free tier, plus $20 and $100 monthly plans.
Continue reading
- Anthropic Just Spent $100 Million Saying the Quiet Part Out Loud
- OpenAI Just Signed Up to Train Main Street. Your Move.
- Book your free 30-minute diagnostic
Common questions
- What did the Cequence report find about Meta Muse?
- On October 7, 2026, Cequence Security reported that within about two weeks of Meta's September 8 Muse launch, traffic matching the Muse agent appeared at more than half of the business customers it studied. At the median customer, Muse traffic grew nearly sixfold in that window.
- How can you tell an AI agent is on your website?
- Right now, mostly you can't. Muse presents itself as an ordinary Chrome browser and doesn't identify as an agent, so standard security tools see a person browsing. Cequence caught it through behavioral analysis, not through anything the agent declared about itself.
- Should businesses block AI agents from their websites?
- Blocking is the wrong first move. Those agents are your customers' errands: comparing prices, filling out forms, or booking appointments. The better move is making sure your site works for them, with simple forms and real availability that confirms instantly.
- What is Meta Muse?
- Meta's personal AI agent, launched September 8, 2026. It runs in its own cloud computer, books travel, handles email, fills in forms, and keeps working after you close the app. A separate program called Sentinel approves anything it sends to the internet. There's a free tier, plus $20 and $100 monthly plans.