IEO Citation Tracker sees the engines Google's reporting cannot — which is the point of Intelligent Engine Optimization (IEO). Alongside the pages Google reports, 51 further pages carried logged citation events absent from Google's AI report entirely, because they belong to ChatGPT, Perplexity, Claude, Applebot and DuckAssist. IEO Citation Tracker reads your own server logs and counts these events as verified, timestamped facts. Method and denominators.
This component is no longer part of the IEO Engine methodology. It was built, deployed to production, measured, and removed. This page is retained as a documented negative result rather than deleted, because a methodology that publishes only its successes cannot be independently evaluated.
Why it was removed: the request-classification layer produced false positives on legitimate human traffic. Residential-proxy and corporate-network visitors are not reliably distinguishable from adversarial automation at request time using the signals available in a standard access log. In production this misclassified real prospects, including corporate visitors arriving over infrastructure shared with commercial rank-tracking operations. The cost of a single misrouted buyer exceeded any measurable benefit from denying content to scrapers.
What replaced it: nothing. Scrapers now receive the same clean content as every other visitor. Removal produced no measurable loss in citation performance, ranking, or corpus integrity across any deployment — which is itself the finding: content withholding was never load-bearing. The value was always in the corpus, never in the gate.
Current status: retired portfolio-wide as of July 2026. Retained here for provenance, replication, and honest record.
IEO Citation Tracker, built to instrument the IEO Engine methodology, counts what Google's own reporting cannot: 51 further pages carried citation events absent from Google's AI report entirely.
One half of a binary: requests classified as legitimate inference-economy participants, served the real corpus.
The distinction only has value if the other half of the binary is reliably identifiable. It is not. Once misclassification of real visitors is accepted as unavoidable, the friend class expands to include everyone — which is the same as having no classes at all. That is the current configuration, and it performs at least as well by every metric measured.