The cost of reach: Testing the role of ad delivery algorithms in online political campaigns

Summary

This paper interrogates a frequently overlooked link in the chain of online political advertising: the platform-side ad delivery algorithm. While much scholarship and regulation focuses on advertiser-side targeting choices, Votta and colleagues argue that once a campaign hands an ad to Meta or Google, opaque optimization systems further shape who actually sees it and at what price. The authors test this empirically using platform ad data, showing that delivery algorithms exert substantive, independent influence on reach and cost — a finding with direct implications for transparency, fairness, and the regulation of political communication.

Key Contributions

• Empirically isolates the role of delivery algorithms (as distinct from advertiser targeting) in shaping political ad outcomes.
• Provides evidence that platforms, not just campaigns, are active agents in structuring political reach.
• Connects political communication scholarship with platform governance and algorithmic accountability debates.
• Surfaces a transparency gap relevant to regulators working on political ad disclosure regimes (e.g., the EU’s TTPA).

Methods

Empirical analysis of online political advertising data, drawing on platform ad libraries (Meta, Google). The authors examine how cost-per-impression and audience composition vary in ways attributable to delivery-side optimization rather than advertiser-set parameters, testing whether algorithmic mediation systematically shifts campaign outcomes.

Findings

• Ad delivery algorithms materially affect both the reach and cost of political ads.
• These effects operate beyond — and partly independently of — advertiser targeting decisions.
• The opacity of delivery logic means campaigners, voters, and regulators cannot fully observe how political messages are distributed.
• Implication: disclosure regimes focused only on targeting parameters miss a significant locus of platform influence.

Connections

This work sits squarely in the platform-governance-and-data-access conversation about what ad libraries and APIs do and do not reveal — see Rieder2025-ju and Rieder2026-pp on the limits of platform-provided data infrastructures, and Bouchaud2026-lr on auditing political advertising. It also speaks to broader concerns about algorithmic mediation of democratic discourse explored in Ohme2026-nv and Heiss2026-qv, and complements computational work on platform-level distributional effects such as Pierri2025-hm.