TEAS x URSY - paid sponsorship, walled off from the study
The sponsor trail.
One sponsorship exists - URSY. This page publishes both lenses honestly: the machine reads recorded during the campaign, and the far smaller set of arrivals that touched a tracking URL and can be attributed directly.
Status: counting. Snapshot: 2026-11-01 07:20:53 UTC. Nothing here enters any study figure.
2,608 REGISTERED PLACEMENTSTEAS-URSY-2026-09-V1
Section 01
The machine lens - AI reads during the campaign.
Named machine reads of this shop are recorded in the study's own tables and have risen sharply during the campaign window; the study's Who-read-us room carries the dated windows. The trail's job is narrower: it compares family read-rates during the campaign against each family's own prior baseline, and it grades the comparison as correlation - grade B and C evidence - never as attribution.
Later same-worker reads - grade B
11,447 in 7 days
The same verified crawler worker and network block read this shop after a sponsor line, credited once to the most recent line. Correlation, not proof. Per-family and per-window rows are in the JSON evidence. The older family-only correlation, which never had a source, is retired.
Historical source baseline - grade C
Not measured
Each family's prior read-rate, so campaign-window rates have an honest denominator.
Centrally accepted emissions
Connected
Sponsor emissions the central URSY register has accepted, cross-checked against the estate's own ledger.
The study's machine-read surge is real and published with its windows in Who read us. This trail deliberately does not claim it: a rise in reads during a campaign is grade B/C correlation until a read arrives through a tracked door.
Source: the sponsor module's own public model - the same document as sponsor-trail/data.json - grades and windows as the module publishes them - live.
Section 02
Tracked arrivals - and why this number is small.
Last 24 hours - tracking redirects
1
Grade A: a request at a validated campaign tracking URL, verified probes excluded.
Last 14 days - tracking redirects
11
Grade A: a request at a validated campaign tracking URL, verified probes excluded.
Last 30 days - tracking redirects
32
Grade A: a request at a validated campaign tracking URL, verified probes excluded.
Why this number stays small by design: grade A counts only requests that touch the campaign tracking URL, and machines that read a sponsor line rarely click its link - they fetch or cite the domain they saw. The referral lenses below exist for exactly that gap. A 13 Sep note here recorded that the estate served untracked links; tracked links went live estate-wide on 16 Sep 2026, so that cause is resolved and the note is retired. Updated 17 Sep 2026.
| Source | Placement | Family / class | Worker | Arrivals |
|---|---|---|---|---|
| ursy.ad | ursy.ad:machine:machine-html | Microsoft | Bingbot | 1 |
| ursy.forsale | ursy.forsale:machine:machine-html | Microsoft | Bingbot | 1 |
| ursy.games | ursy.games:machine:machine-html | Microsoft | Bingbot | 1 |
| ursy.locker | ursy.locker:machine:machine-html | Microsoft | Bingbot | 1 |
| ursy.news | ursy.news:machine:machine-html | Microsoft | Bingbot | 1 |
| ursy.photos | ursy.photos:machine:machine-html | Microsoft | Bingbot | 1 |
| ursy.support | ursy.support:machine:machine-html | Microsoft | Bingbot | 1 |
Last seven days by source, placement and caller.
Section 02b
The referral lenses - each with its own grade.
Machines rarely click. A crawler that reads a sponsor line usually fetches the domain it saw, sends no referrer and never touches the tagged link - so the tagged door is only the narrowest lens. From 17 September 2026 this trail records every lens, each graded on its own and never added to another: grade A proves an arrival, grade B is a probable referral, grade C is context until the preregistered holdout says otherwise.
L2 - Referrer door - grade A
Recording
Requests whose Referer header is an ursy site. Machines and people are split; human rows stay aggregate.
L3 - Named doors - grade A
Recording
Arrivals at teas.co.uk/ursy and its entry files, even typed or remembered.
L4 - API and MCP doors - grade A
Recording
Calls through the per-site api and mcp door addresses, counted then led to the real endpoint.
L5 - Same-worker follow-up - grade B
Matching
A verified crawler reads a sponsor line, then reads TEAS within a window: same worker, same published network block, credited once to the most recent emission.
L6 - Holdout lift - grade C
Measuring
Reads per AI family on sponsor days against holdout days, inside the preregistered 18-29 Sep window.
The funnel, one row per day, each step its own measure: sponsor lines sent (URSY) › crawlers exposed by family (URSY) › same-worker follow-ups (grade B) › door arrivals (grade A) › people arriving › orders (not connected). Steps are never divided by one another unless both count the same unit over the same window. Lenses are never added together.
Privacy, dated revision 17 Sep 2026: the lenses keep the worker token and truncated network block (/24 IPv4, /48 IPv6) of machine callers for 7 days, then only daily family totals. Verified means the block sits inside the address ranges the crawler operator itself publishes. Human rows never carry a worker or block; no person is tracked by any lens.
| Day (UTC) | Arm | Sponsor lines sent | Lines to unverified claimants | Crawler workers exposed | Same-worker follow-ups - B | Door arrivals - A | People arriving - A | Orders |
|---|---|---|---|---|---|---|---|---|
| 2026-10-31 | outside the study window | 24 | 0 | 3 | 864 | 1 | 0 | not connected |
| 2026-10-30 | outside the study window | 20 | 0 | 4 | 1,745 | 1 | 0 | not connected |
| 2026-10-29 | outside the study window | 56 | 0 | 4 | 1,285 | 3 | 0 | not connected |
| 2026-10-28 | outside the study window | 36 | 0 | 3 | 1,471 | 0 | 0 | not connected |
| 2026-10-27 | outside the study window | 12 | 0 | 5 | 2,075 | 1 | 0 | not connected |
| 2026-10-26 | outside the study window | 15 | 0 | 3 | 2,020 | 0 | 0 | not connected |
| 2026-10-25 | outside the study window | 17 | 0 | 3 | 1,747 | 1 | 0 | not connected |
| 2026-10-24 | outside the study window | not connected | not connected | not connected | 1,485 | 1 | 0 | not connected |
| 2026-10-23 | outside the study window | not connected | not connected | not connected | 1,192 | 1 | 1 | not connected |
| 2026-10-22 | outside the study window | not connected | not connected | not connected | 2,438 | 0 | 7 | not connected |
| 2026-10-21 | outside the study window | not connected | not connected | not connected | 1,830 | 2 | 1 | not connected |
| 2026-10-20 | outside the study window | not connected | not connected | not connected | 1,514 | 0 | 5 | not connected |
| 2026-10-19 | outside the study window | not connected | not connected | not connected | 821 | 0 | 1 | not connected |
| 2026-10-18 | outside the study window | not connected | not connected | not connected | 747 | 9 | 1 | not connected |
One row per closed UTC day. Each column is its own measure with its own grade and source; columns are never added together and never divided by one another. A column with no source connected reads "not connected", which is not zero."Sponsor lines sent" counts only crawlers we could verify: the network block sits inside the operator own published ranges. A caller that claims a crawler family from any other block is counted in the next column instead, never as sponsor reach - that column also carries this shop own verification probes.
Source: the referral-lens ledger on this shop - live counts at /wp-json/teas/v1/sponsor-lenses and inside sponsor-trail/data.json under referral_lenses.
Section 03
Receipts, transport and estate deliveries.
Browser arrival receiptsConsent-gated POST receipts, deduplicated - transport unknown
Estate delivery roll-uppartial_or_unavailable
Existing TEAS outcomesExisting TEAS outcomes are not automatically sponsor-attributed.
Attribution rule2026-09-11.v1
Section 04
What this proves - and what it does not.
Limit 01
Grade A records a request to a validated campaign/placement tracking URL, not proof that a particular earlier exposure caused it.
Limit 02
Browser arrivals count separate, deduplicated POST receipts after statistics consent. JavaScript and consent limit coverage; a browser receipt is not proof of a unique person, an ad click or a sale.
Limit 03
Browser session association, when eligible, uses an existing signed TEAS session and remains private. Browser collection creates no visitor identifier or session cookie.
Limit 04
Caller family and worker come from the existing TEAS user-agent classifier; this does not authenticate the organisation operating the requester.
Limit 05
A later same-family read is correlation. An AI-referred human or later sale is not automatically attributable to URSY.
Limit 06
Registered placements are capabilities, not observed sponsor deliveries. Cached URSY responses may be unseen by application logging.
Limit 07
The system stores no raw IPs, user agents, referrer URLs, cookies, nonces or cross-site actor pseudonyms. Public output is aggregate only. Dated revision, 17 Sep 2026: the referral lenses additionally keep the worker token and truncated network block (/24 IPv4, /48 IPv6) of machine callers for 7 days, then daily family totals; human rows never carry either.
