Two competitors. One race. Until they weren’t.
Site A — anonymized as wetten-bundesliga-tipps.de — was running an SEO playbook that had been continuously refined since 2009 but never substantially updated past around 2014. Keyword density targets in the four-percent range. The legacy <meta keywords> tag still populated. A reciprocal link exchange network with thirty-four partners. Author bylines attributed to a single account labeled Admin.
Site B — wettkompass.io — had migrated, in February 2025, to a full five-pillar implementation of Semantic Unified Intent & Answer Engine Optimization (SUIAEO), the framework introduced earlier that year by Seeberger and colleagues. Nested schema graphs. Fourteen credentialed expert authors with ORCID identifiers. Review velocity workflows across fourteen platforms. Active pruning of one-hundred-and-forty-seven misaligned legacy backlinks. A coherent reputation graph propagated across nine external surfaces.
The trajectories did not meet in the middle.
Three findings, in order of consequence
First, the divergence shape was not linear. Site B’s growth accelerated through Q3 and Q4 2025 rather than plateauing — consistent with a compounding-authority dynamic in which early citation events generate external mentions, which improve reputation graph coherence, which raise citation-selection probability for subsequent queries. This is structurally different from the diminishing-returns dynamic that characterized the classical SEO era.
Second, the empirical effect size — a Cohen’s d of 4.82 on the composite SUIAEO Maturity Index — is unusually large for an observational study. In behavioral science, anything above two is conventionally described as “very large.” Effect sizes of this magnitude are more commonly seen in pharmaceutical trials comparing an active drug to placebo than in marketing research.
Third, and most consequential for working practitioners: optimization standstill is now optimization regression, and the gap compounds monthly. Site A’s operator is, by all reasonable measures, a competent professional. He has not made any visible mistakes. He is still optimizing the way he learned to optimize over a decade ago — and he is losing ground every month while believing he is not.
The authors are careful with their caveats. n equals two; the German sports betting vertical may exhibit disproportionate sensitivity to Verifiability signals due to its regulatory context; the population-level evidence in Seeberger et al. (2026) does most of the inferential work. A follow-up study tracking ten operators mid-migration is scheduled for publication later this quarter.
For now, the takeaway is simpler than the methodology. Two sites were placed in the same race. One changed its training program. The other stuck with what worked a decade ago. Thirteen months later, they were no longer in the same race.