OpenAI
- ChatGPT AgentAI AgentHigh
- ChatGPT-UserAI AssistantMid
- GPTBotAI ScraperLow

See which AI agents and bots reach your site, check that each one is who it claims to be, then allow it, block it or make it read-only.
AI AGENT TRAFFIC
Review the paths each agent requests and the traffic you allow or deny. Check requests that claim an agent’s identity but fail verification.
AGENT ACCESS CONTROLS
Kasada shows how each agent’s identity is checked and how strong that verification is. You choose which endpoints agents can use, with policies by endpoint category and HTTP method. Keep login, checkout and payments human-only while allowing agents to browse other pages.
VERIFIED BOTS DIRECTORY
OpenAI and Anthropic each run several bots with different jobs. Kasada lists each bot separately, with its purpose and verification strength, so you can set access for each one.
AI BOT MANAGEMENT BY INDUSTRY
Choose where AI bots can browse and where actions like login, checkout and payments stay human-only.
How can we welcome shopping assistants without giving scrapers unrestricted access?
Shoppers send agents to find products and compare options. Scrapers use the same pages to pull prices, inventory and promotions at scale. See which is which, then let shopping agents browse and block the rest.
Keep products open to the assistants your customers use.
Check an agent’s identity instead of trusting its name.
Block bulk extraction of prices, inventory and promotions.
Proof
Forrester named Kasada a Leader in The Forrester Wave™: Bot And Agent Trust Management, Q2 2026, with the top score in 9 criteria.
Sources: The Forrester Wave™: Bot And Agent Trust Management, Q2 2026; The Total Economic Impact™ Of Kasada (Forrester Consulting, 2026). The ROI figure covers the whole platform, not AI Agent Trust alone.
FAQ
Not sure what to allow yet? Read the guide.
Yes. Dashboards show request volume and the paths each agent requests, with crawlers and agents shown separately. You can export the request logs to your own tools.
By what the agent can prove. A cryptographic Web Bot Auth signature is strong evidence of the service behind a request. So is a request from an IP range the operator publishes, as Google does for its crawlers. A user-agent string is easy to copy, and signals such as fingerprints are weaker evidence on their own. Kasada grades each agent by that evidence, and a request that claims an agent’s name but fails its verification is flagged as impersonation.
Yes. Set an access policy for each agent, by endpoint category and HTTP method. Allow requests, restrict access to read-only or block the agent. Verification and permission are separate, so you can recognize an agent and still deny it.
Yes. You can keep those endpoints human-only while allowing AI agents to use other parts of your site. Set policies by endpoint category and HTTP method.
No. Verified bots skip browser interrogation. Kasada checks their identity and applies your access policy in milliseconds, with the same focus on keeping pages responsive for people.
Not yet, but we’re looking into it. In the meantime, Kasada lets you allow an agent, block it or make it read-only. To slow a useful crawler down, use your CDN’s rate limits.
With Account Intelligence, you can start to see it. When an agent signs in to a customer’s account or uses authenticated checkout, Account Intelligence links that activity to the account. We’re working to make that link more complete.
Kasada sees it as browser automation and treats it that way. It isn’t recognized as that customer’s personal agent.
Kasada shows how strongly each agent proves its identity, flags failed verification and lets you set access by endpoint category and HTTP method. You can see what each agent requests and keep your CDN’s rate limiting alongside those controls. Book a demo to compare this with your current setup.
All the major ones, including Meta Muse, and we keep adding more. The directory lists each supported bot with its purpose and verification strength. OpenAI and Anthropic each run several bots, so you can set access for the specific bot rather than treating the whole provider as one.
Not with the safe defaults. Unless you choose to block a search crawler such as Googlebot, which we don’t recommend, the defaults leave your SEO and AI search visibility as they are. On top of them, set controls for your own business rules: for example, let agents search and browse your inventory, but keep checkout for people.
AI Agent Trust is part of Kasada’s bot detection and is included with it. Agent traffic does count toward the usage in your contract. Talk to our sales team for the details.
Not at this time. Kasada covers agents that reach your application through a browser. It doesn’t protect machine-to-machine APIs or MCP endpoints that agents call directly, without a browser.