AI agents tripled web traffic in a month. Now sites are fighting back

Agentic web traffic tripled between August and September, and Meta's new AI agent Muse racked up 5 million downloads in its first month. That surge has triggered a backlash: platforms from Pizza Hut to Yelp are blocking the agents, Amazon is waging a court battle against Perplexity, and the internet's old bot-versus-human rules are collapsing.

The conflict is not just a nuisance for early adopters. Wharton professor Ethan Mollick predicts that once people realize agents can negotiate refunds and prices, customer service will be overwhelmed, and "humans will eventually get very annoyed when their agents aren't allowed to do this on their behalf." Meanwhile, Shopify is already giving agents their own checkout lanes, signaling that e-commerce giants see dollar signs in the new traffic.

For years, the web sorted visitors into two buckets: bots and people. Bots were allowed to crawl pages under strict rules, while humans got the full experience, captchas included. Personal AI agents break that binary. They are not human, but they act on behalf of a specific person doing legitimate tasks like booking flights or comparing prices, just at machine speed. Because they don't fit existing bot-detection systems, security-conscious platforms have chosen to ban them outright.

The bans are already visible. TechCrunch reported users struggling to deploy agents on Pizza Hut and Yelp, which refuses to let AI book restaurant reservations. Amazon has blocked Meta's Muse and is appealing a court ruling that favored Perplexity, arguing that AI agents make it harder to distinguish human-viewed ads from bot-served ones, threatening its advertising business. The judge sided with Perplexity, but Amazon's appeal keeps the legal question open.

The stakes go far beyond a few blocked vacation bookings. Mollick, an associate professor at Wharton, argues that once agent-assisted refund requests and price negotiations become common knowledge, customer service departments will be flooded to the point of being unusable for real people. That could force platforms to choose between accommodating agents or watching their human users get locked out of basic services.

Money will likely decide the outcome. The vast majority of agentic traffic is hitting e-commerce sites, and Shopify is already welcoming agents with dedicated checkout lanes that bypass human-designed pages. If agents can shop on behalf of real people, platforms have a clear incentive to let them in. The question is what happens to the human experience in the process.

A web optimized for agents could look bleak for people. Sites might become neglected as destinations for humans, interfaces degrade, and verification systems grow more intrusive - imagine constant human checks or even iris scans. Prices could vary by agent, and you might never know whether a recommendation is driven by ads or platform commissions. That is a far cry from the future Google, Anthropic, and OpenAI have promised, where agents give consumers more bargaining power.

Agents can genuinely help, from fighting bogus medical bills to finding cheaper car insurance. Whether they deliver depends on the rules being written now for how agents interact with websites. Some rules, like securing credit card data, will benefit everyone. The contested ones - whether businesses can block agents or prevent them from negotiating better deals - will determine whether the web becomes a highway for agents working for us or against us. For executives, the message is clear: the architecture of the internet is being redrawn, and the choice is whether to build for humans, agents, or both.