Three Verticals, One Curve: The Ingestion Sequence Replicates
The strongest objection to any single case study is “that’s just your niche.” This note retires the objection with replication: three deployments in three unrelated verticals, launched in February, April, and May–July 2026, produced the same ordered sequence of ingestion milestones. Where the same event could be timed on multiple deployments, the timings agree. The ingestion curve belongs to the protocol, not the vertical.
Key Findings
- Full AI-crawler ecosystem assembly (major search + AI agents present in logs) occurred within the first days of each deployment becoming ingestible — all three verticals.
- On the two deployments where a large corpus release could be precisely timed, crawler swarms arrived the same day — in the most recent case, 60% of the month’s crawled pages occurred within 24 hours of release.
- Every deployment then entered the staircase: impressions before positions, positions before clicks (FN-002).
- Entry-page decentralization followed on the deployment old enough to measure it: 62% → 17.6% homepage entry share (FN-003).
- Crawler-class behavior (FN-001) reproduced identically on all three — same binge ingesters, same compounding re-crawler.
The replication table
| Milestone | Local service (Feb 2026) | B2B reference (Apr 2026) | Consumer app (May–Jul 2026) |
|---|---|---|---|
| Crawler ecosystem assembly | Days | Days | Days |
| Same-day swarm on corpus release | Documented (month 2) | — (staged build) | Documented (Jul 5: 60% of month’s pages in 24h) |
| First search-classification impressions | Week 1 | Week 2 | Week 1 |
| Staircase step 1 (deep impressions at volume) | Month 2 | Month 2 | Days 1–5 post-corpus (early) |
| Interior pages become leading entries | Month 4–5 | Month 3 (case study #2 entry) | Days — guides entering top entries immediately |
| Binge/compounding crawler split (FN-001) | Reproduced | Reproduced | Reproduced |
| Platform citation activity (live per-user fetches) | Yes — multi-month | Yes — incl. July fan-out (FN-006) | Early signals |
Why replication is the finding
Any single deployment’s results admit a niche explanation: an uncompetitive market, a lucky topic, a seasonal tailwind. Three verticals with different competition profiles, different query economics, and different audiences — producing the same ordered sequence on comparable clocks — do not. The sequence (ecosystem assembly → ingestion events → classification → decentralization → citation) is the observable outer surface of one evaluation cycle after another, and it recurs because the corpora share the same extraction readiness — the property the protocol engineers for.
What the protocol is, this series does not say. That the curve replicates, the logs of three unrelated businesses now do. The two public write-ups — the local-service 60-day deployment and the travel-content velocity deployment — carry the per-deployment detail.
Terms Demonstrated in This Note
- Evaluation cycle
- A search or AI system’s recurring pass over a corpus in which classification, position, and citation decisions are revisited.
- Extraction readiness
- The degree to which a corpus’s content can be ingested, chunked, and reused by machine systems without friction — the engineered property that ingestion behavior responds to.
Related Field Notes
FN-001: Two Crawler Classes: Binge Ingesters and Compounding Re-Crawlers · FN-002: The Staircase Effect, Confirmed in Google’s Own Data · FN-003: Entry-Page Decentralization: A Server-Side Gauge of Topical Authority · FN-006: The Citation Fan-Out: What It Looks Like When a Platform Distributes Your Page