- TechCrunch reports that personal AI agents such as Meta's Muse are being turned away by retail and travel sites, some on purpose and some by accident.
- Customers cannot tell a deliberate block from a bot-check side effect, so they are frustrated with both the retailer and the agent.
- Leaders should decide their agent policy on purpose and test their own checkout with a personal agent.
A "no" with several possible causes
When an AI agent fails to book a flight or buy groceries, the customer sees one thing: it did not work. The reason behind that failure can be very different from case to case.
The lesson here is that bot defences built to stop spam now meet customers who send software to shop for them. The defence sees automation, not intent. The open question is how many blocks are chosen and how many are accidents. In this report, Walmart is the one company that says its block was unintentional.
What TechCrunch reported
TechCrunch reports that personal agents, including Meta's Muse, Instinct and ChatGPT's Dots, are running into blocked websites. Amazon has begun blocking Muse from its retail site, so the agent can no longer browse or buy from the catalog.
The problem is wider than one retailer. TechCrunch says user complaints on social media describe similar blocks. Some are deliberate, like Amazon's. Others come from ordinary anti-bot measures.
These are complaints and company statements, not a measured count. TechCrunch notes that Cloudflare had no specific data to share on the blocking of personal agents.
How an accidental block happens
Walmart is the clearest case. It is a Muse partner, announced at Meta's Connect conference in September. Even so, TechCrunch saw a handful of complaints that Muse could not finish a purchase there. NBC also reported that Walmart had faced the problem in its own tests.
Walmart told TechCrunch the blocks were not intentional. The issue customers appear to hit, according to TechCrunch, is a button that asks a visitor to prove they are human. If that step is interrupted, the check can fail and the agent is removed.
This is the mechanism to understand. An agent that drives a website like a person meets the same gates meant to stop bots. A gate cannot read intent. It sees automation and acts.
One social media user, quoted by TechCrunch, said they could not even sign in when buying directly. That is a single claim. If accurate, it suggests a defence aimed at agents can also catch people.
Deliberate blocks take different forms
Not every refusal is a defect. Several companies describe positions they chose, and treating them as one problem leads to the wrong fix.
Yelp says it does not permit non-human traffic unless the agent has paid for access through its data licensing program. A Yelp quote, a waitlist spot or a table booking will likely fail without that deal.
Delta says it has no partnership that lets a third-party agent book its flights. Any opening would need to proceed with "security and the customer experience in mind." United pointed to its Terms of Use, which bars robots and automatic devices without written permission.
eBay says it does not prohibit all third-party shopping agents. Its policies restrict unauthorized agents and actions such as automated scraping and model training.
A couple of people said on social media that eBay suspended their accounts over agentic AI use. eBay did not address this in the statement TechCrunch quotes.
Could a vendor default be setting the policy?
A suspicion has formed online that Cloudflare, or similar network services that sit in front of websites, may be interrupting Muse traffic. It is tied to a September 15 change in crawler defaults. The change lets site owners block AI training while allowing other bots.
The suspicion is that sites previously set to block AI bots for security reasons were shifted to block AI agents on pages with ads. TechCrunch does not confirm this. Cloudflare told TechCrunch it had no specific data to share. It pointed to its public Cloudflare Radar site, which shows rising bot activity but does not separate good bots from bad.
The link is unconfirmed. It still raises a fair question: can a vendor's default setting turn into a business policy without a meeting?
Questions to put to your team
Meta, Walmart and Stripe are among the named firms drafting a shared rulebook for how agents deal with merchants. It focuses on online commerce. Meta says it should help tell helpful bots from bad ones. It is a draft effort, so decisions cannot wait for it.
Ask who owns your agent policy. Is it written down, or is it whatever your bot-protection vendor shipped?
Ask whether anyone has run a purchase through a personal agent on your own site. Do human checks fail when the flow is interrupted?
Ask whether each block is a deliberate decision, such as a licensing model or a terms-of-use position. Or is it a side effect you would fix if you saw it?
Ask what you tell customers who arrive with an agent and fail. TechCrunch says they end up frustrated with both the retailer and the agent.
A closed door is a decision. A door that sticks is a defect. Customers cannot tell which they have met, so owners need to know before their customers ask.
Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: TechCrunch.
