01
What a client name establishes
Applies to marketplace onlyown storefronthybrid
An automated client identifies itself with a header it selects. Operators publish the names they use: OpenAI’s documentation lists its crawlers, the purpose stated for each, and the IP ranges they originate from.5 That publication tells a site owner which strings are intended to be that operator’s, and it does not prevent any other party from sending the same string.
The check applied on top has historically been address-based. The operator publishes ranges, the site verifies the request came from one, and the list is fetched, cached, and expired. The cost of that approach scales with the number of operators, and it addresses clients running from a crawler fleet rather than from a buyer’s own infrastructure.
02
What a signature covers
Applies to marketplace onlyown storefronthybrid
RFC 9421 is a published Standards Track document specifying how a client produces a cryptographic signature over selected components of an HTTP message and how a server verifies it.1 The word selected is load-bearing: the signer chooses which components are covered — method, path, particular headers — and verification applies to those and not to the rest.
The specification establishes how a request can be signed and verified rather than who is entitled to access anything.1 A verified signature is therefore an input to a policy, and the policy is written separately.
| Question | Answered by | Not answered by |
|---|---|---|
| Was this request altered in transit? | The signature over covered components | An allowlist of any kind |
| Which key sent it? | Signature verification with key discovery | A self-declared client name |
| Should it be served? | The site owner’s policy | Any cryptographic result |
03
Web Bot Auth and its current status
Applies to marketplace onlyown storefronthybrid
Web Bot Auth is the work applying message signatures to automated clients specifically. The current document covers signing and verifying automated client identity, key discovery, and deployment models, which is the material between a signature format existing and a site being able to act on one.2
Its status is stated in the document. It is an active Internet-Draft, which is work in progress and is cited only as such. It is an individual submission rather than working-group output, it replaced an earlier architecture draft, and the current revision carries an expiry date in December 2026.2 A description of it as a standard is not accurate as of this lesson’s review date.
04
Verification as it is deployed today
Applies to marketplace onlyown storefronthybrid
Verification is running in production ahead of a finished standard. Cloudflare’s verified-bot criteria admit two routes, one through Web Bot Auth and one through IP validation.3 A storefront behind that infrastructure has this setting whether or not anyone in the business has looked at it.
Two limits apply to how far it carries. It describes one provider’s list and criteria, so verification there is verification by that provider. And the same documentation states that inclusion on the list creates no requirement for a site owner to serve the client.3 The identity result and the access decision remain separate records.
The decision side is a separate set of controls: per-crawler allow and block rules, visibility into crawler activity, and a pay-per-crawl path.4 The last of these is described as a private beta, which places it among things that exist rather than among things a seller can currently plan around.
05
Writing the access posture
Applies to marketplace onlyown storefronthybrid
A reseller storefront consumes these mechanisms rather than implementing them. The available work is writing the policy in advance, so that it exists before an incident rather than being decided during one.
- Write the three questions separately: which client is this, may it fetch this path, and at what rate.
- State what an unverified client receives, which for a public product page is ordinarily the page.
- State what no client receives regardless of verification: checkout endpoints, account pages, bulk export paths.
- Check whether the host or CDN already exposes verified-bot handling, and record its current setting.
- Record the one condition that would change the policy, so the review has a stated trigger.
| Path | Verified client | Unverified client | Live setting |
|---|---|---|---|
| Product pages, including the camera’s | Served | Served — it is a public page | Matches, confirmed in the dashboard |
| Search and filter endpoints | Served, rate-limited | Rate-limited more tightly | Matches |
| Checkout and account paths | Refused | Refused | Matches |
| Bulk export path left from a migration | Refused | Refused | Unknown — its configuration was not located |
Three rows confirmed what the seller already believed and took a few minutes each. The fourth did not resolve, and it is the row that records something new: there is a path on this site whose handling of automated traffic nobody in the business can currently state.
A posture written this way does not depend on the identity question being settled. It holds if signatures become common, and it holds if the draft expires and is replaced, because each row states a decision about a path rather than a decision about a mechanism.
06
Checking a claim in this area
Applies to marketplace onlyown storefronthybrid
The vocabulary in this area is ahead of the deployment, and three questions can be checked against the primary documents.
- Which document is being implemented, and what is its current status: published, active draft, or replaced?
- Does the verification being offered establish identity only, or is a permission also being asserted?
- What happens by default when a client fails verification, and is that setting visible to the site owner?
The first question has a documented answer. RFC 9421 is published and Standards Track, and Web Bot Auth is an active individual-submission draft, so a claim describing the second as a standard can be checked against the draft’s own status.
07
The identity documents and their status
Applies to marketplace onlyown storefronthybrid
RFC 9421 specifies signing and verifying selected components of an HTTP message and does not address entitlement.1 Web Bot Auth applies that to automated clients and is an active Internet-Draft, an individual submission with an expiry date in December 2026.2 OpenAI publishes its client names, their purposes, and their address ranges.5 Cloudflare publishes verification criteria with two routes, and states that inclusion imposes no requirement on a site owner.3 Its separate controls cover per-crawler rules, crawler-activity visibility, and a pay-per-crawl path in private beta.4
Between them these establish what a signature covers, what the draft currently is, and what one provider verifies. None states which clients a site owner should serve, which is the entry the seller’s own posture fills.
08
Practice
Exercise
Write the access posture for one storefront
- List your paths in three groups: public catalog, gated, and never served to an automated client.
- For each group, state what a verified client receives and what an unverified one receives.
- Check what your CDN or host currently does with verified bots, and record whether the live setting matches what you wrote.
Check yourself
A client sends a valid RFC 9421 signature. What does that establish?
That the covered components of the request were not altered and that the holder of the corresponding key sent it. It establishes nothing about whether the client may fetch what it requested.
A product is described as implementing the Web Bot Auth standard. What is inaccurate?
The status. It is an active Internet-Draft, an individual submission not adopted by a working group, carrying an expiry date, and an earlier draft in the same line was replaced.
A CDN verifies a shopping agent. Does the site have to serve it?
The provider’s documentation states that inclusion on the verified list creates no requirement to serve the client. The identity result and the access decision stay separate.
Progress is saved in this browser only. No account, nothing sent anywhere.
09
Common questions
Does a reseller need to implement signature verification?
For a storefront this arrives through a host or CDN rather than through the seller’s own code. The available work is writing the policy and checking the live setting against it.
Is refusing unverified clients a reasonable default?
For public product pages it removes traffic that carries no verification, including user-initiated fetches. The paths where refusal is unambiguous are checkout, account, and bulk export.
Does verification stop scraping?
It raises the cost of impersonating a named operator. A client willing to be anonymous is outside its scope, which is where rate limits and gated paths apply instead.
When will the draft be settled?
The sources do not state a date. The draft carries an expiry in December 2026 and no working-group adoption, so a policy written against paths rather than against the document survives its revision.
10
Research and sources
Rules and platform policies change. These primary sources were reviewed on ; confirm the current position for your jurisdiction and account before acting.
Claim evidence
- RFC 9421 specifies cryptographic signatures over selected components of an HTTP message; it establishes how a request is signed and verified, not who is entitled to access anything.
- technical requirement. Supported by RFC 9421: HTTP Message Signatures .
- Web Bot Auth is an active Internet-Draft and an individual submission not adopted by an IETF working group; it replaced an earlier architecture draft and carries an expiry date, so it is work in progress rather than a settled standard.
- current external fact. Supported by Web Bot Auth: HTTP Message Signatures protocol (Internet-Draft) .
- Cloudflare’s verified-bot criteria describe one provider’s allowlist, and verification confers no obligation on a site owner to admit a verified client.
- current external fact. Supported by Verified bots , AI Crawl Control .
- OpenAI publishes its crawler names, stated purposes, and IP ranges, which documents its own clients only and does not prevent another party sending the same user-agent string.
- current external fact. Supported by Overview of OpenAI crawlers .
- RFC 9421: HTTP Message Signatures Internet Engineering Task Force · Tier D · published Standards Track RFC · RFC 9421
Cryptographic signatures over selected components of an HTTP message. Limit: A signature mechanism only. It establishes how a request can be signed and verified, not who is entitled to access anything.
- Web Bot Auth: HTTP Message Signatures protocol (Internet-Draft) Internet Engineering Task Force · Tier D · active Internet-Draft; individual submission, not adopted by an IETF working group · draft-meunier-webbotauth-httpsig-protocol-00, expires 2026-12-28
Signing and verifying automated HTTP client identity, key discovery, and deployment models. Limit: An Internet-Draft is a work in progress and must be cited only as such. It replaces draft-meunier-web-bot-auth-architecture, is an individual submission rather than working-group output, and expires 2026-12-28; nothing in it is a settled standard or an access entitlement.
- Verified bots Cloudflare · Tier A · current documentation · unversioned live documentation
Cloudflare’s verification criteria for bots and agents, and its Web Bot Auth and IP-validation paths. Limit: Describes one infrastructure provider’s allowlist and its criteria. Verification by Cloudflare is not verification anywhere else, and it confers no obligation on a site owner to admit a verified client.
- AI Crawl Control Cloudflare · Tier A · current documentation · unversioned live documentation
Per-crawler allow and block rules, pay-per-crawl monetization, and crawler-activity visibility on Cloudflare zones. Limit: One provider’s control surface; pay-per-crawl is described as private beta. It establishes what the product can do, not what any given zone is currently configured to do.
- Overview of OpenAI crawlers OpenAI · Tier A · current documentation · unversioned live documentation
Named OpenAI crawlers, their stated purposes, robots.txt behavior, and published IP ranges. Limit: Documents OpenAI’s stated crawler behavior only. It does not establish what any other operator’s client does, and it states that robots.txt rules may not apply to user-initiated ChatGPT-User fetches.