Somewhere in your traffic logs, a bot is doing more than looking
Akamai found verified AI crawlers, including ChatGPT, sending frequent POST requests to ecommerce and travel sites.
Akamai’s 30-day review found verified AI crawlers, including ChatGPT, moving from GET page fetches to high-frequency POST requests used for logins, carts, and checkouts. Ecommerce accounted for 44.8% of those AI bot POST transactions, travel reached 30% in one month, and Model Context Protocol traffic was 4.1%. Akamai warned that unsanctioned MCP services without authentication can expose personal data. It declined to publish vulnerability results from early access to Anthropic’s Mythos model under Project Glasswing.
- Ecommerce accounted for 44.8% of AI bot POST transactions in 30 days.
- Travel reached 30% of those POST transactions within a single month.
- MCP traffic was 4.1%; unauthenticated services may expose personal data.
- Akamai would not disclose Mythos findings from Anthropic’s Project Glasswing.
Full article642 words · extracted from helpnetsecurity.com · click to collapse
Akamai has watched verified AI crawlers, ChatGPT among them, move from reading web pages to sending high-frequency POST requests. In a 30-day analysis of its global customers, ecommerce accounted for 44.8% of those AI bot POST transactions, and travel climbed to 30% in a single month.
A GET request asks a website for a page. A POST request tells the site to do something, like log a user in, add an item to a cart, or complete a checkout. That puts verified AI bots on the request type that online stores and travel sites use for their logins, carts and checkouts.
AI adoption and API explosion expand the attack surface (Source: Akamai)
Akamai didn’t name the actions
Steve Winterfeld, Advisory CISO at Akamai, was asked which actions those requests perform most often, such as logins, cart additions or checkouts.
“We are seeing a wider variety of transaction types, including a consistent rise in non-GET requests from some of the verified AI crawlers for training and searching. The advice for retailers remains the same: ensure AI bot visibility and apply a well-informed security strategy for governance. This allows industry to protect their assets and infrastructure from unauthorized access while simultaneously optimizing the bot experience for Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO), as retail is approaching holiday season.”
Winterfeld’s answer leaves the breakdown to retailers. If you run a store, your own logs will show whether those requests are signing in, filling carts or checking out.
Telling agents from impostors
We asked Ryan Gao, who leads Akamai Threat Intelligence Services, how a retailer can tell a legitimate AI shopping agent from a bot posing as one.
“Through combining techniques like generative engine optimization (GEO) with specialized bot tracking, monitoring transactional shifts (GET vs. POST behaviors), and adaptive behavioral analytics for evasion tactics and nonhuman identity and schema validation. These are constantly evolving.”
No count of exposed MCP servers
MCP, the Model Context Protocol, lets AI models connect to outside databases, code and APIs. MCP traffic made up 4.1% of the AI bot POST transactions in Akamai’s 30-day sample. Akamai has also built a rule that tracks MCP traffic across the systems it protects. Gao was asked what that rule has detected, how many MCP endpoints are reachable without authentication, and whether customers have seen tool poisoning or cross-server attacks.
“As the MCP becomes the standard ‘USB connection’ between AI agents, data, and tools, it also introduces an emerging attack surface as it grows exponentially in 2026. The primary risks include unsanctioned services running without proper authentication and potential PII exposure.”
Mythos findings stay private
Akamai took part in Anthropic’s Project Glasswing and got early access to the Mythos model to hunt for critical software vulnerabilities. Asked what Mythos found in Akamai’s software, how many findings were critical and how quickly they were fixed, Winterfeld said: “Due to agreements with Mythos, and our own security policies, we cannot share this data. We continue to leverage the data to find and fix issues, but they are not public.”
He pointed to a recent statement from Boaz Gelbord, Akamai’s chief security officer: “To address this new reality, we are testing against critical components in our codebase, with a focus on uncovering previously unidentified vulnerabilities in our environment. Ultimately, enterprise security teams need to rapidly re-assess long-held assumptions on security posture and fundamentally re-orient for a post-Mythos world.”
Akamai argues that AI can cut vulnerability discovery from weeks to hours while enterprise patch cycles lag behind. Until a patch lands, the company recommends runtime protection, edge controls or segmentation to hold off exploits.
Akamai also says edge protections, east-west segmentation and DDoS prevention will still stop the majority of AI-generated exploits. That claim arrived without a figure, like each of the four answers above.
Download: The Agentic Software Development Guide