GEO ABCD Test — geo-lab
Epoch 2026-07-06 UTC · 14-day windows · currently window 0 (bots pinned to v1) · generated 2026-07-07 23:19:18 UTC
Human engagement by variant
Which version of the post: v1 control, v2 light, v3 balanced, v4 heavy. A badge with a gold ring is the variant bots see right now. |
Server-side sends of this variant — the blocker-proof denominator. A "low n" flag means the sample is still too small to trust. |
Beacon-reported pageviews. Blockers can suppress these. |
Views ÷ Serves. Below ~70% means blockers are eating the beacon, so engagement then describes only the non-blocking population. Delivery % colour ≥ 70% — healthy, compare freely 40–70% — watch, blockers biting < 40% — numbers describe only the non-blocking subset |
Distinct visitor IDs that produced a pageview. |
Mean seconds between the pv and end beacons. |
Mean of the deepest scroll depth reached per session. |
Pageviews with no end beacon of at least 10 seconds. |
|
|---|---|---|---|---|---|---|---|---|
| no data yet | ||||||||
Click any column heading (marked ) for its definition. A gold-ringed badge is the live variant; low n marks a sample too small to trust.
Beacon delivery by traffic source
How the visitor arrived: ai, search, other, or direct — classified server-side from Referer / utm_source. |
Server-side sends to this source class. Blocker-proof. |
Beacon pageviews from this source class. |
Views ÷ Serves for this class. A class far below the others is running blockers at a higher rate. Delivery % colour ≥ 70% — healthy, compare freely 40–70% — watch, blockers biting < 40% — numbers describe only the non-blocking subset |
|||||
|---|---|---|---|---|---|---|---|---|
| no data yet | ||||||||
A source class whose delivery sits far below the others (often 'ai' or 'search' audiences on blockers) is only partly measured — don't compare its engagement head-to-head with a high-delivery class.
Bot windows & AI referral attribution
Which variant did an engine ingest?
Referrer clicks can't tell you — pinning rotates variants and every index eventually holds all four. Instead, each variant carries one unique-but-true phrasing of the same fact. Grep an AI answer for these exact strings:
- v1
four rewrites of one article behind a single canonical URL, each pinned to a fourteen-day window - v2
a single canonical URL fronting four rewrites of the same article, one per fourteen-day window - v3
one article rewritten four ways, served from a single canonical URL on fourteen-day windows - v4
four versions of one article, one canonical URL, a fresh variant each fourteen-day window
14-day period index since epoch. The highlighted row is live now. |
Calendar span this window covers. |
The variant every bot saw during this window. |
Declared-crawler requests, classified by user-agent. |
Cloudflare-verified ÷ scored hits. n/a when the plan exposes no bot-management signal. |
Distinct crawlers seen in this window. |
Human pageviews referred by an AI engine, bucketed into this window. |
Which AI engines drove those clicks, with counts. |
|
|---|---|---|---|---|---|---|---|---|
| no data yet | ||||||||
AI clicks are bucketed into the window active when the click happened, so you can line up an engine's referral spike with the variant it was serving then. Confirm the ingested variant with the watermark strings under Attribution & watermarks above.