On the night of Sunday, September 20, 2026, Amazon started blocking Meta's new AI shopping agent, Muse, from its retail site. If you run a small service business, this might look like a fight between two giants that has nothing to do with you. For a small service business, this is a useful prompt to check how automated tools access your website — what happens when your customer isn't a person, but an AI acting on one's behalf?
What Actually Happened
Meta launched Muse, an AI agent that can shop and handle tasks like bookings on a user's behalf, earlier in September 2026. Amazon started blocking it on the night of Sunday, September 20, showing pop-ups telling Muse users that its use violates Amazon's Conditions of Use, as TechCrunch reported. According to GeekWire, Amazon's stated objections come down to three things: Meta didn't tell Amazon that Muse would be browsing its store, the agent doesn't identify itself as an AI when it browses, and — Amazon says — it appears to capture and store customer login credentials, which Amazon frames as a privacy and security risk. Meta's own description of Muse cuts against that last point: GeekWire notes Meta had said Muse "has no visibility into people's passwords or payment methods," and that credentials a user shares go into secure storage so Muse can use them without seeing them.
Whatever you think of either company, the underlying fight is really about three things: disclosure (should an agent say it's an agent?), consent (did the site agree to be accessed this way?), and data handling (what does the agent do with a customer's credentials?). Those three questions apply just as much to a five-person service business as they do to Amazon.
A bilingual website Google and ChatGPT understand
Built for people and for AI search. I build it directly with you.
Why It Matters If You're Small
You're not going to end up in a standoff with Meta. But your website is already being visited by things that aren't people — search crawlers, AI answer engines, and increasingly, AI agents acting on a customer's behalf to research a service or even fill out a form for someone. The Amazon–Meta fight is forcing a decision that used to be optional into the open: do you want AI agents interacting with your site, and on what terms?
I've made a version of this decision on my own site, and it's a useful, plain-English way to see the choice. Cloudflare's managed robots.txt uses a "Content Signals" scheme with three categories, expressed as preferences in your robots.txt file — not an enforcement mechanism, and not something every crawler honors:
- search — a signal that content can be used to build a search index. Note: this category excludes AI-generated summaries; it does not mean "AI can summarize this."
- ai-input — a signal about feeding content into AI systems for real-time answers (e.g., an assistant citing your page in a live response).
- ai-train — a signal about using content to train or fine-tune AI models.
These are stated preferences recorded in robots.txt, not switches that guarantee an outcome: some bots ignore robots.txt entirely. Cloudflare's own documentation is explicit that this is signaling, not blocking — for actual blocking, it points to a separate feature, AI Crawl Control.
On my own site, I went one step further than a robots.txt preference: crawlers for search and for AI agents acting on someone's behalf are allowed, and known AI-training crawlers are blocked at Cloudflare's edge. Visible to what can recommend me, closed to what trains on my work. Bots can misidentify themselves, so it's not a guarantee — and it's not the only reasonable choice for a small business to make.
What to Actually Do About It
You don't need a legal team to start. Three concrete steps:
- Check what your site's
robots.txtcurrently says, or ask whoever manages your site to check it. If you're on Cloudflare, look at both the managed robots.txt / Content Signals settings (preferences) and AI Crawl Control (actual blocking). If you don't know your current setting, that's the finding — and remember it's a preference, not a lock. - Make sure the steps from visiting your site to contacting you make sense to something reading them literally. An AI agent filling out your contact form won't infer that "call us" means "click here" if there's no phone number and no link doing the same job.
- Ask a few AI assistants what they say about your business — whether or not any of them are shopping on it.
None of this requires picking a side in the Amazon–Meta fight. It just means treating "an AI might interact with my site on a customer's behalf" as a current fact instead of a future one.
Not sure what your site tells bots? Ask me.
Sources
- TechCrunch, "Meta's AI agent has been blocked from using Amazon.com" (Sept 21, 2026) — https://techcrunch.com/2026/09/21/metas-ai-agent-has-been-blocked-from-using-amazon-com/
- GeekWire, "Amazon blocks Meta's Muse AI assistant in new standoff over agentic shopping" — https://www.geekwire.com/2026/amazon-blocks-metas-muse-ai-assistant-in-new-standoff-over-agentic-shopping/
- Cloudflare Developer Docs — "Managed robots.txt" (Content Signals: search / ai-input / ai-train) — https://developers.cloudflare.com/bots/additional-configurations/managed-robots-txt/
Tell me the one task that eats your week. I’ll tell you honestly if AI can take it, and what it would take.
Tell me what you need
