
What happened
Users of Meta's Muse, xAI's Grok, OpenAI's Dots and Instinct report blocks on top sites. Yelp limits agents to data-licensing partners, Delta Air Lines bars ticket booking, and JPMorgan Chase and Charles Schwab cite security concerns.
Why it matters
If sites keep turning agents away, the agents can do fewer useful things — undercutting the pitch that they will handle everything people do online. Companies that pay for access or partner with retailers still let agents in.
WHO IT HITSPeople who rely on AI agents to book travel, shop or manage accounts are finding those tasks blocked or falling back to manual steps. Site operators and anti-bot providers are being forced to decide whether to admit agents, partner with them, or shut them out.
Summaries like this, in your inbox every morning.
The tension has been building since Amazon banned Perplexity's agentic shopping feature and sued the startup. Amazon argued that third parties buying on behalf of customers should operate openly and respect whether a service provider participates; Perplexity countered that Amazon wants human eyeballs on its ads and likened its agent to an extension of the user. A judge initially ordered Perplexity to stop, but an appeals court overturned that ruling in August, and Amazon has filed an amended complaint.
Even where sites have not banned agents outright, agents are running into anti-bot systems. Reports on Reddit suggest Google is blocking some accounts that use Muse for 'suspicious activity,' and others report checkout problems on Walmart despite its Meta partnership. Cloudflare, meanwhile, introduced a one-click block for AI crawlers in 2025 and began blocking crawlers that scrape for both search indexing and AI training.
The practical result is a patchwork: some sites shut agents out, some partner with them, and some simply trigger security systems never designed for them. Chris Canal of EquiStamp says that if companies actively work to keep agents out, there are fewer useful things agents can do — and he expects more human-verification hoops to appear as a result.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
In 15 in-depth interviews with existing ServiceNow customers, 14 said AI Control Tower's runtime enforcement w…
Anthropic's Claude solved a 25-year-old cryptography mystery in under 30 minutes, finding the unexplained seed…

Anthropic CEO Dario Amodei said he spends a third, maybe 40%, of his time making sure the company's culture is…

Meta removed language from its commercial terms that experts said allowed scraping news sites to train its AI…

Bank of America strategist Sebastian Raedler said 12-month forward hyperscaler capex expectations rose from $3…

Deepseek shipped Deepseek-V4.1-Flash-1017 via its website, app, and API, and published the weights on Hugging…
