Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsSome links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Web Bot Auth lets an automated client prove which signing identity sent an HTTP request by signing it with a private key. A website can discover the corresponding public key and verify the signature, then make its own decision about whether to allow the request. It is an emerging protocol—not a universal feature of AI agents—and a valid signature is not blanket permission or proof of safe behavior.
What Web Bot Auth proves—and what it does not
Web Bot Auth is a way for an automated HTTP client to cryptographically sign outbound requests. A server that supports the mechanism can verify the signature against a public key associated with the claimed agent identity. The protocol is being developed in an IETF Internet-Draft titled “HTTP Message Signatures for automated traffic.” The retrieved revision, draft-ietf-webbotauth-httpsig-protocol-00, is dated September 1, 2026, and says it expires March 5, 2027. It is a work in progress, not a published RFC; draft text can change, be replaced, or be obsoleted.
The key distinction is authentication versus authorization. Authentication answers, “Does this request carry a valid signature from the identity associated with this key?” Authorization is the site’s separate decision about what that identity may do. The signature does not establish that a bot is benign, follows crawl rules, represents a particular human, has consent to access private material, or may bypass a login or paywall.
How a website verifies an agent request
- The operator publishes public-key information. The bot operator controls a signing key and makes its public key discoverable in a JWKS-based key directory. The draft describes a well-known location for that directory.
- The agent identifies the directory. It sends a
Signature-Agentfield with the request. In the draft, that field provides in-band information for finding the key material. - The agent signs the HTTP message. It creates an HTTP Message Signature using its private key. The draft specifies a
web-bot-authtag and describes@authoritytogether with the signedSignature-Agentmember as baseline covered information. - The verifier obtains the public key and checks the signature. The website resolves the signing identity to key material, verifies the signature, and checks the request components covered by it.
- The site applies its own policy. A valid signature can inform bot identification or access decisions, but the origin still decides whether the request is permitted and how it is handled.
The draft permits signatures to cover additional request components. Covering method and path, for example, narrows which request the signature authenticates. This matters because a signature that covers only @authority can be replayed against different methods, paths, or bodies at that same authority until it expires. Expiry limits that replay window; including more relevant components limits the signature’s scope.
#1 Best Overall
Body integrity requires explicit coverage
A signature does not automatically bind every part of an HTTP exchange. The draft says a signer that needs to protect the request body must send and cover Content-Digest. A verifier should not infer body integrity from a valid signature unless the signed components actually provide it.
Provider instructions are not the whole protocol
Cloudflare’s implementation documentation gives provider-specific setup guidance, including HTTPS and handling of directory responses. Those are Cloudflare implementation requirements, not a reason to treat every provider’s deployment as identical. A site should follow the requirements of the verifier it operates and the current protocol revision it implements.
How Web Bot Auth differs from familiar bot checks
Before signed-agent verification is available, websites commonly rely on network and request metadata. Google’s surfaced experimental guidance describes IP address and user-agent verification as the current de facto standard; Cloudflare also lists IP validation and reverse DNS among verification methods. These checks can be useful, but they are different kinds of evidence from a cryptographic signature.
| Signal | What it relies on | Operational consideration |
|---|---|---|
| Web Bot Auth | A signature associated with discoverable public-key material and the request components it covers. | The client must sign, the verifier must support the mechanism, and key discovery and signature validation must work. Expiry and component coverage affect replay risk. |
| IP allowlist or IP validation | The source network address and an operator-maintained address policy. | Addresses can change, so operators need an update and validation process. Cloudflare lists IP validation as a verification method. |
| Reverse DNS | A DNS relationship used as part of identifying a host. | It is a network- and DNS-based check rather than a signature over the HTTP request. Cloudflare lists it among verification methods. |
| User-agent checks | A client-supplied request header and the site’s rules for interpreting it. | It is metadata, not cryptographic proof by itself. Google’s surfaced experimental guidance says IP and user-agent checks remain the de facto standard. |
These signals need not be treated as mutually exclusive. A site can use the checks it supports, assess the strength and freshness of each signal, and apply its own risk and access policy. A signed identity does not require Cloudflare: the protocol’s key-directory model is not owned by one platform.
What adoption looks like today
Cloudflare
Cloudflare documents Web Bot Auth as a bot and agent verification method. Its documentation, last updated July 1, 2026, says signed agents are represented in verified-bot metadata and describes an application process for requesting directory inclusion. Cloudflare’s verified-bot criteria separately include honest self-identification and non-abusive behavior, including respecting robots.txt and crawl directives. That separation is instructive: signature validation addresses identity, while behavior and policy remain separate checks.
OpenAI ChatGPT Work Cloud browser
OpenAI documents that its ChatGPT Work Cloud browser signs outbound requests with Web Bot Auth and makes public verification keys available through a well-known directory. Its guidance names Akamai, Cloudflare, HUMAN, and Vercel configuration or support examples. This is a documented platform example, not evidence that all AI agents sign requests. OpenAI also states that, at launch, the Cloud browser cannot sign in to websites or complete payments; that is a time-sensitive product limitation and may change.
Google experimental guidance
Google’s surfaced experimental guidance describes verifying signed requests according to RFC 9421 while noting that IP and user-agent checks remain the de facto standard. It does not establish broad deployment or a rollout size. Treat experimental guidance as platform-specific and check the current documentation before relying on it for an integration.
What a site operator should decide before accepting signed agents
- Which identities are eligible? Decide whether a valid signature is merely logged, receives a distinct bot classification, or is eligible for specific routes or rate limits.
- Which request parts must be protected? Review the signature’s covered components. If method, path, or body affects the decision, ensure the signature covers the relevant information; body protection requires the draft’s
Content-Digestcoverage. - How will keys be discovered and refreshed? Identify the directory source, retrieval behavior, caching and failure policy, and how key updates are handled. Follow the verifier implementation’s own requirements.
- What happens when verification fails? Choose deliberately among treating the request as unidentified, applying existing bot checks, or denying selected operations. A lookup or verification failure is not the same as proof that a request is malicious.
- How is behavior governed after identification? Keep authorization, rate limiting, abuse detection, crawl directives, and privacy controls independent from identity verification.
- What is the fallback for unsigned clients? Not all agents implement Web Bot Auth. Preserve the site’s existing controls for ordinary browsers and bots unless a route has a deliberate policy requiring a signed identity.
Implementation and troubleshooting considerations
The protocol draft describes the message fields, signature coverage, and key discovery model, but a deployment needs a compatible signer and verifier. Do not hand-build signature serialization from a brief overview or assume that adding a header alone makes a request verifiable. Use an implementation that follows the current protocol draft and the HTTP Message Signatures format it references, then test it against the actual verifier.
Signature is missing or not recognized
Confirm that the client is using a Web Bot Auth-capable implementation, that the request contains the expected signature metadata, and that the server or intermediary supports the mechanism. An AI label in a user-agent string is not a substitute for a signature.
Rank #4
Key lookup fails
Check that the advertised directory reference resolves to the expected JWKS-based key directory and that the verifier can retrieve and process the response. For Cloudflare, apply Cloudflare’s HTTPS and directory-response requirements; do not assume those exact operational rules apply unchanged to a different verifier.
Signature validation fails
Check that the public key corresponds to the signing key, that the signed components match the received request, and that the signature has not expired. Proxies or gateways that rewrite request details can matter when those details are covered: validate what the verifier sees against what the signer signed.
Free tools Windows power users keep installed
One-click scans. No signup required.
Request verifies but access is denied
This can be correct behavior. Verification authenticates the signed identity; the origin still evaluates permissions, route rules, rate limits, and behavior policy. Check the site’s authorization decision rather than weakening signature checks to grant access.
Best Value
Body or replay protection is weaker than expected
Inspect the covered components. The draft warns that authority-only coverage can allow reuse across methods, paths, and bodies at the same authority until expiry. Ensure the signature covers the components relevant to the operation, use an appropriate expiry, and include and cover Content-Digest when body integrity is required.
Where ScreenshotNeo fits—and where it does not
ScreenshotNeo is a website screenshot API and MCP server for developers, not a Web Bot Auth signer, verifier, or key directory. It is relevant when an AI agent or developer workflow needs to capture a page visually; taking a screenshot does not authenticate the request to the site or grant access. Its MCP server provides take_screenshot, get_page_info, and capture_pdf tools for Claude, Cursor, and other MCP clients.
Or skip the browser setup
For a screenshot capture, one GET request can return an image or PDF. The example below saves a WebP response; see the ScreenshotNeo API documentation for parameters and response details.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpScreenshotNeo accepts cookie and consent banners before capture and removes 60+ known consent platforms, newsletter popups, and chat widgets; each of these steps can be turned off. Bot checks and CAPTCHAs, blank pages, timeouts, failed loads, and cache hits are not billed, with response headers indicating the page verdict and billing status. The MCP server lets AI agents request screenshots through an MCP client. The Free plan includes 1,000 shots a month with no card; paid plans start at $5 for 3,000 shots. Sign up for 1,000 free screenshots a month with no card.
FAQ
Does Web Bot Auth mean a website trusts an AI agent?
No. It can authenticate a signing identity. The website decides separately whether that identity is authorized and whether its behavior is acceptable.
Do all AI agents use Web Bot Auth?
No. OpenAI documents support for its ChatGPT Work Cloud browser, but that example does not establish universal support across AI agents.
Is Web Bot Auth a published IETF standard?
No. The cited protocol revision is an Internet-Draft. Its status and contents can change before any final standard is published.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
