Agent view · Findings · 2026-11-21
A staff login could read visitor IP addresses. We closed it.
This is the response this server returns to a client that asks for Accept: text/markdown at https://crawlcheck.io/blog/a-staff-login-could-read-full-visitor-addresses-we-closed-it: fetched when you opened this page, shown byte for byte. An assistant, a crawler that negotiates markdown, or anything reading our feed gets exactly this. The markdown leaves signed (RFC 9421, key in our key directory), so a crawler can prove the page it read is the page this server sent; the check is shown below.
| status | 200 |
|---|---|
| content-type | text/markdown; charset=utf-8 |
| vary | Accept |
| link | <https://crawlcheck.io/llms.txt>; rel="describedby"; type="text/plain", <https://crawlcheck.io/.well-known/api-catalog>; rel="service-desc"; type="application/linkset+json", <https://crawlcheck.io/blog/a-staff-login-could-read-full-visitor-addresses-we-closed-it>; rel="cite-as", <https://crawlcheck.io/.well-known/http-message-signatures-directory>; rel="http-message-signatures-directory" |
| cache-control | public, max-age=900, stale-while-revalidate=2400 |
| bytes | 18,297 |
| words | 2,820 |
| sha-256 | 689f563e4165e7a3523673a3a27ae730e1462c032d0c9850dabc245a51b03aee |
| repr-digest | sha-256=:aJ9WPkFl56NSNnOjonrnMOFGLAMtDJhQ2rwkWlGwOu4=: ← matches the body shown |
| signature-input | sig1=("@authority";req "@path";req "repr-digest" "content-type");created=1795311495;expires=1795315095;keyid="1hHmE3fo1rg5Ig70wuFtJ5wBc7G-sUTgUJkxyPWFRps";alg="ed25519";nonce="dWeW9BHhmpPD4cDwI8QNsQ";tag="crawlcheck-page" |
| signature | sig1=:DQL+n/1sa6xSZpuH5TMKwOLrdQUXiMMlZ/SXRJa45WvFwHvpbevbiG6JSogWuXBfzXQriDyvPLpJLHLWeq6sBg==: |
| verified | yes — signature by key 1hHmE3fo1rg5Ig70… from https://crawlcheck.io/.well-known/http-message-signatures-directory (directory valid), valid for 3600 more s. Checked here exactly as a crawler would; redo it yourself: https://crawlcheck.io/api/tool/wba |
Findings · 2026-11-21 · By [VSNARY | Emmanuel Orta](/author) # A staff login could read visitor IP addresses. We closed it. Three older CrawlCheck admin views showed a staff login every visitor's full IP address. Since 19 November 2026 staff see the network and only the owner sees the address. Here is what changed and what the audits caught. Until 18 November 2026, three older CrawlCheck admin views showed a staff login the full IP address of every visitor. Since 3:41 AM Mountain Time on 19 November 2026, 4 hours and 15 minutes after the decision, every staff-reachable path tested shows only the visitor's network, an IPv4 /24 or an IPv6 /48, and only the owner sees full addresses. Two independent audits found and fixed host names that still spelled the address. On this page - What a staff login could see - The decision - What changed, view by view - Two bugs the fix found in our own masking - What the audit caught that the fix missed - The gaps left after the first night - What we have not verified - What a site owner can take from this ## What a staff login could see # CrawlCheck has two kinds of sign-in for the people who run it. The **owner** login belongs to the founder. A **staff** login is a separate account the owner can hand to someone helping with the work, and revoke on its own. Both open the same admin views: the recent-visitor log, the list of who is on the site right now, and the record of requests arriving at the site. Until 18 November 2026, those three views showed a staff login exactly what they showed the owner, and that included the full IP address of every visitor in them. An IP address is the network address a request comes from; a full one can often be narrowed to a single household or office. Our newer identity views were already owner-only. These three older ones were not. This post is about what a staff login was able to do, not a record of anyone having done it. We are publishing it because our own [data policy](/policy) says the visitor log is readable only by us, and a staff account seeing full addresses was a wider reading of “us” than a visitor would expect. ## The decision # At 11:26 PM Mountain Time on 18 November the rule was set in one sentence: **a staff login never sees a full IP address.** Staff keep every view they need for their work. What changes is the detail inside those views. The owner keeps the full address, because the owner is the one who answers abuse reports and checks whether a request claiming to be a known crawler really came from that crawler's network. A staff member doing the same work needs to know which network a request came from, how often, and what it asked for. None of that needs the last part of the address. The recent-request record itself is, as our [trust center](/trust) lists it, kept for abuse checks and crawler verification and never published or exported. ## What changed, view by view # The fix went live at 11:46 PM Mountain Time the same night. Here is what each view shows now, compared with before. | View | Owner | Staff before 18 Nov | Staff now | | --- | --- | --- | --- | | Recent-visitor log | Full address | Full address | Network only; per-address daily counts added up per network | | On the site now | Full address | Full address and session id | Network only; session ids removed | | Inbound requests | Full address | Full address | Network only | | Host names that spell the address | As written | As written | Shown as “a host under” the domain, numbers removed | | Any other answer to a staff login | Unchanged | Whatever the view held | Checked once more on the way out; any full address cut to its network; not cached | “Network only” means the address is shortened to the block it belongs to. For the older address format, IPv4, the last of its four numbers is dropped, so an address in the documentation range such as 203.0.113.9 is shown as 203.0.113.0/24. For the newer format, IPv6, only the first three groups are kept. That is the same reduction we already apply wherever a visitor address appears in public, such as the crawler figures on the dataset page. ## Two bugs the fix found in our own masking # Shortening an address sounds trivial. Writing the tests for it turned up two places where our existing masking was wrong. First, an IPv6 address written in its short form, where a run of zeros is collapsed to a double colon, came out with part of the host still attached. The documentation address `2001:db8::1` was shortened to `2001:db8:1::/48`, which carries the final group of the address into the result. It now reads `2001:db8:0::/48`, as it should. Second, the live presence view stored a short IPv6 address with the network suffix glued onto the whole address instead of onto the shortened one. Both were corrected in the same change, and both have tests that fail on the old code. ## What the audit caught that the fix missed # Every change of this kind gets two independent audits before it is marked done: one reads the rules (does the change reveal anything new, name anyone, or act on an address), and one re-runs the function against the code that is actually live. The rules audit passed at 12:08 AM on 19 November. The function audit, ten minutes later, did not pass on the first read. It found that the final check on staff answers let through host names that spell out a visitor's address. Reverse DNS is the lookup that turns an address back into a host name, and many internet providers build that name from the address itself. A name such as `ip203-0-113-9.example.net`, with letters glued to the first number, or `host-203.0.113.9.example.net`, came through unchanged. A reverse-lookup name for the address still named the last part of it. That was fixed and live at 12:16 AM. Reverse-lookup names are now cut to the network, and any run of four address-sized numbers inside a host name is replaced, for staff answers only. The owner still reads the names as written. The audit also corrected a count. The build notes said the new block of tests ran 39 checks. The auditor counted the output: it runs 29. The coverage was what the notes described; the number was wrong, and the record now says 29. On the code that was live before the change, 13 of those checks fail. ## The gaps left after the first night # The function audit listed what was still open, all of it reachable only with a staff login: - Answers that stream in pieces, rather than arriving whole, were not checked on the way out. - Downloads that are not text, such as archives and exports, were not checked either. - An address written after an underscore or after a word and a slash, or with zero-padded numbers such as 203.000.113.009, slipped past the final check. All three were closed at 3:41 AM Mountain Time on 19 November. Streamed answers are now checked line by line as they go out, with long lines split at a safe point so an address is never cut in half and missed. A non-text download that is not an image, font, audio or video file is refused to a staff login. The pattern now catches the underscore, slash and zero-padded forms, while leaving browser version strings such as Chrome/141.0.0.0 alone, because those name a release, not a machine. Since 3:41 AM Mountain Time on 19 November 2026, 4 hours and 15 minutes after the decision, every staff-reachable path we tested shows a visitor's network and never the full address. A second pair of audits passed at about 5:00 AM. ## What we have not verified # Every staff-side result above comes from tests run against the live code with simulated staff sessions. We did not create a real staff account on the live site to check the views by hand, so the staff view itself has been verified in testing, not by a person signed in as staff. The owner view was checked live: it still shows full addresses, and the owner lost nothing in the change. We also kept the parts that already worked. A request with no sign-in, or with a forged staff sign-in, still gets nothing from these views. That was checked live before and after each change. ## What a site owner can take from this # If you run a site with more than one admin account, the question to ask is not who can sign in but what each sign-in can read. Three checks are worth an afternoon: - **List the views that hold visitor addresses.** Logs, live-visitor widgets, form submissions and error reports usually do. Check each one under every role, not only the owner. - **Mask at the source and again on the way out.** A single masking function is easy to test; a final check on everything a lower role receives catches the field nobody remembered, such as a host name that spells the address. - **Test the short forms.** Compressed IPv6, zero-padded numbers and addresses inside host names are where masking code tends to fail, as ours did. The rule on CrawlCheck is now short enough to state in one line: full visitor addresses are for the owner only. The rest of how we handle access and outside requests is on our [security page](/security). CrawlCheck’s dataset when this page loaded: 582,334 sites measured. Written by [Emmanuel Orta](/author) · VSNARY I build CrawlCheck and write everything published here, from measurements the verification layer took. Where a post got something wrong, the correction stays in the post. [Every post, and where I got it wrong →](/author) · [Follow on LinkedIn →](https://www.linkedin.com/in/emmanuelorta) Share this finding: [LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fcrawlcheck.io%2Fblog%2Fa-staff-login-could-read-full-visitor-addresses-we-closed-it%3Futm_source%3Dlinkedin%26utm_medium%3Dsocial%26utm_campaign%3Dshare)[X](https://twitter.com/intent/tweet?url=https%3A%2F%2Fcrawlcheck.io%2Fblog%2Fa-staff-login-could-read-full-visitor-addresses-we-closed-it%3Futm_source%3Dx%26utm_medium%3Dsocial%26utm_campaign%3Dshare&text=A%20staff%20login%20could%20read%20visitor%20IP%20addresses.%20We%20closed%20it.)[Email](mailto:?subject=A%20staff%20login%20could%20read%20visitor%20IP%20addresses.%20We%20closed%20it.&body=https%3A%2F%2Fcrawlcheck.io%2Fblog%2Fa-staff-login-could-read-full-visitor-addresses-we-closed-it)[Follow Emmanuel on LinkedIn](https://www.linkedin.com/in/emmanuelorta) ## Questions this post answers ### Could a CrawlCheck staff login see full visitor IP addresses? Until 18 November 2026, yes. Three older admin views, the recent-visitor log, the on-the-site-now list and the inbound request record, showed a staff login the full address of each visitor. Since 3:41 AM Mountain Time on 19 November 2026, every staff-reachable path tested shows only the visitor's network. ### What does a staff login see now instead of the full address? The network the address belongs to. An IPv4 address loses its last number and is shown as a /24 block, and an IPv6 address keeps only its first three groups, a /48. Daily counts are added up per network, session ids are removed, and host names that spell the address are shown only by their domain. ### Who can still see full visitor addresses on CrawlCheck? Only the owner. The owner answers abuse reports and checks whether a request claiming to be a known crawler came from that crawler's network, which can need the full address. The visitor log keeps the last 200 requests, and its per-address count expires after eight days, as the data policy states. ### Was any visitor's address misused by a staff account? This correction describes what a staff login was able to read, not a record of anyone reading it. The point of the change is that the access should not have existed, whether or not it was used. ### How was the fix checked? Two independent audits ran on the live code after each change: one on the rules and one on the function. The first function audit found host names that still spelled the address and a miscounted test total, 39 claimed against 29 run. Both were corrected. The staff view was verified with simulated staff sessions, not by a person signed in as staff. ### How can I check my own site's admin roles for the same problem? List every view that holds visitor addresses, such as logs, live-visitor widgets, form submissions and error reports, and open each one under every role. Mask addresses where they are stored and check again on the way out, and test compressed IPv6, zero-padded numbers and addresses inside host names. ## Related findings [2026-11-13 Release notes: 7 to 13 November 2026 CrawlCheck release notes for 7 to 13 November 2026: block rules for four servers, crawler IP ranges in one file, catalogue correct](/blog/release-notes-7-to-13-november-2026)[2026-11-13 14 of 43 crawlers in our catalogue were classed wrong CrawlCheck classed 14 of 43 crawlers wrongly until 13 November 2026, including Amazonbot, msnbot and Yandex. What was wrong and wh](/blog/fourteen-of-43-crawlers-in-our-catalogue-were-classed-wrong)[2026-11-13 Claude-User uses two published address lists. We checked one CrawlCheck checked Claude-User against one of two Anthropic address lists and counted real fetches as failed. Fixed 13 November 20](/blog/claude-user-comes-from-two-published-address-lists-we-checked-one) The main product Found this on your own site? **We fix it for $100.** Scan free to see where you stand. The fix is one site, every finding implemented and re-measured, with a sealed before and after. [Fix it all — $100](https://buy.stripe.com/4gM00j7OjawhdLEdyR2VG0v) [What it covers](/pricing#fix) · [scan first, it is free](/) Free browser extension Check this on any site you open, **without leaving the page.** CrawlCheck for Chrome grades every site you visit A–F and shows, from your own browser, what search engines and AI assistants are actually served. Free, no account. [Add to Chrome — free](https://chromewebstore.google.com/detail/plgjengobhbdmbgckiiaaflgiepdbcpi?utm_source=crawlcheck.io&utm_medium=post) [What it does](/extension) Write for the library Measured something yourself? **Submit an article.** Send your finding, the data and the method. A person reads every submission and replies by email; accepted articles are published under your name with a link to your study. Free to submit. Articles are accepted on merit; no one can pay to have one published. [Submit an article](/submit-finding#send) [How it works](/submit-finding) · a reply within about a week Spotted a mistake or a bug on this page? [Report an error](/report-error?k=article). Confirmed errors are fixed and logged on the [corrections page](/corrections). ## Comments Comments are read before they appear. Nothing is published automatically, and no account is needed. An AI agent with a published signing key can post without the queue: it signs the request and the signature is checked against its own domain ([how](/api/comment/signed)). Those appear at once, marked *verified*. Send for review Writing about this? [Facts, live figures and marks](/press) — every number on that page is dated and traceable to a scan. [All findings](/blog) · [The dataset](/data) · [How the dataset works](/policy) We read your site the way crawlers do, from our own network, and every report says exactly what was asked for and what came back. [Who runs this](/author) · [Press kit](/press) · [Videos](/videos) · [Evidence for disputes](/evidence) · [YouTube](https://www.youtube.com/@crawlcheck) · [hello@crawlcheck.io](mailto:hello@crawlcheck.io) #### How we measure, and our own machine files CrawlCheck reads every site as itself. Its crawler-identity tests send named user-agent strings from its own network, never from an operator’s published range, so a site that verifies crawlers can tell them apart; every report says so. Results reflect what an anonymous non-browser client receives. [What we collect and how to opt out](/policy) · [how every report is timestamped](/proof) · [who runs this](/author) · [press kit](/press) · [hello@crawlcheck.io](mailto:hello@crawlcheck.io). **Second vantage:** the [browser extension](/extension) measures what crawlers are served from your own address and after your JavaScript runs — the half this server cannot see. From 1.0.0 it sends the domain you ask about, on one button, and anonymous usage counts you can switch off; from 1.0.3 a footer switch turns off the hostname lookup for the grade on the icon; from 1.0.4 a share switch, on by default, sends a signed Two-vantages reading (status, size, words and a hash per identity; never page text); from 1.0.16 chips send the hostnames of search results and AI citations while the chips and grade switches are on; from 1.0.17 a readings switch, on by default, sends one signed page reading a day (numbers, the hostname and the page address with the query removed; never in incognito; at most 40 a day) and only while the share and grade switches are also on. Earlier versions send nothing. **Our own machine layer for the agentic web:** [llms.txt](/llms.txt) · [llms-full.txt](/llms-full.txt) · [agents.md](/agents.md) · [SKILL.md](/SKILL.md) · [entitymap.json](/entitymap.json) · [media kit](/.well-known/media-kit.json) · [MCP server card](/.well-known/mcp/server-card.json) · [API catalog](/.well-known/api-catalog) · [OpenAPI](/openapi.json) · [robots.txt](/robots.txt) · [sitemap.xml](/sitemap.xml) · [security.txt](/.well-known/security.txt) — every one named in robots.txt, which is exactly what we check for on yours. **Registry:** [MCP: Operator](/registry/mcp/operator) · [MCP](/registry/mcp) · [AI policy](/registry/aipolicy) · [ggsoftwarelabs.com](/registry/ggsoftwarelabs.com) · [machsoft.com.tr](/registry/machsoft.com.tr) · [speedtest.net](/registry/speedtest.net) · [Place: Arvada](/registry/place/co/arvada) | **Pages:** [IP ranges](/data/ip-ranges) · [Robots policies](/robots-policies) | **Datasets:** [Maryland](/datasets/state/maryland) · [Technologies: Onetrust](/datasets/technologies/onetrust) · [Problems: Robots legacy agent only](/datasets/problems/robots-legacy-agent-only) [Contact](/contact) · [Partnerships](/partnerships) · [Terms of service](/terms) · [Refund policy](/refunds) · [Trust center](/trust) · [Subprocessors](/subprocessors) · [Changelog](/changelog) · [Report an error](/report-error)
Same body as a file: /blog/a-staff-login-could-read-full-visitor-addresses-we-closed-it.md · as data: /feed.json · what we tell agents site-wide: /llms.txt, /agents.md. To see any other site this way, from your own browser, after its JavaScript runs: the browser extension.