Outrigger AI Visibility Index
Study 1 was the April 2026 baseline on a Phase 1 sub-sample of 1,004 businesses. The v3 paper supersedes it, and the Phase 1 dataset record was withdrawn on 7 May 2026. The Phase 1 figures previously published on this page are no longer shown; what remains below is what the v3 paper itself prints about Phase 1.
Observational findings show association, not causation.
What the v3 paper retains from Phase 1
Three passages in the v3 paper refer to Phase 1. They are reproduced here exactly, with their section numbers.
“This paper supersedes earlier reports of the same dataset. Version 1 (April 2026) reported headline statistics on a Phase 1 sub-sample of 1,004 businesses.”
“Sampling proceeded in two waves. Phase 1 (n = 1,004) sampled businesses URL-first, beginning with an industry-city directory query and verifying first-party domain ownership. Phase 2 (n = 1,725) sampled name-first, beginning with a business name and resolving the URL post-hoc via Apify Google SERP.”
“Domain Authority has a raw Pearson correlation of 0.337 with the composite visibility score on the Phase 1 sub-sample (n = 1,004) and 0.217 on the full sample (n = 2,200 businesses with non-null DA). After BH-FDR correction it remains significant as a raw correlation. In the multivariate L2-regularised logistic regression on n = 2,729 — which simultaneously estimates 22 features — its standardised coefficient is +0.139 with a bootstrap 95% confidence interval of [−0.031, +0.316], which crosses zero.”
Why the other Phase 1 figures were removed
The Phase 1 dataset record (DOI 10.5281/zenodo.20048614) and the v2 paper (DOI 10.5281/zenodo.20048424) were both withdrawn on 7 May 2026. The v3 paper describes why the earlier analysis was replaced:
“Version 2 (May 2026, DOI 10.5281/zenodo.20048424) extended the analysis to a Phase 2 expansion but contained methodological issues including unreported skew in count features, multicollinearity in the logistic-regression design matrix, no multiple-comparison correction, and an aggregation of outcomes across models that masked the model-specific structure described in §4. The v2 record has been withdrawn.”
Figures from the Phase 1 report that the v3 paper does not reprint are therefore not published on this site. The full re-analysis on all 2,729 businesses is summarised on the Study 2 page, and in full in the paper.
Where does your business stand?
Run a free AI visibility audit. Timing depends on provider availability.
Analyze your brand