Citation monoculture

Alemohammad et al.'s own coined term (arXiv 2608.19230, 2026) for the cross-vendor agreement they measure at round 0 of their recursive citation-selection benchmark — different LLMs from different vendors converging on the same citation preferences for real papers. The paper's own finding complicates the term's implication of a uniform, self-shared bias: under recursion the agreement (0.68 cross-model correlation) holds for real papers but collapses (0.20) for AI-generated ones, so the "monoculture" tracks something like a shared external anchor in real scholarship rather than a general convergent preference. Distinct from, but placed by the paper under, the broader pre-existing term "algorithmic monoculture" (not coined here, not yet its own vault entity). Watching for recurrence before promotion to a full hub — this is the term's first appearance in the vault.

References

• claim-alemohammad-2026-recursive-citation-benchmark-dilution-concentrates-attention
• claim-alemohammad-2026-cross-vendor-citation-monoculture-collapses-under-recursion
• Capture: 10-inbox/raw/2026-09-14-does-llm-citation-popularity-bias-measurably-compound-across.md