Answers are grounded in this paper and cite their sources.
This analysis reveals how AI dependence is built and priced among various global providers, indicating significant implications for consumer choice.
Consumer Reliance Architecture is the third leg of a reliance trilogy (institutional, generational, consumer). It profiles how individual AI dependence is built and priced across US, EU, Chinese and companion providers, measured through the commercial machinery that captures it: pricing tiers, free-tier design, usage caps, metering changes, and portability friction. The finding is reported as a per-provider profile, not a single composite score. Three results: 1. Three pricing archetypes. US closed labs (OpenAI, Anthropic, Google) monetise dependence through a steep tier-and-cap ladder; the EU hybrid (Mistral) keeps a ladder but open weights leave a self-host escape valve; the Chinese providers (DeepSeek, Qwen) run no consumer ladder at all, relocating reliance to data residency, geopolitics and an enterprise-cloud funnel. A cross-cutting bundling vector runs through the largest players (Gemini as the Android default, Copilot folded into Windows and Office). 2. The headline subscription price is the decoy. It has been flat or falling; the tightening shows up instead in free-tier ads, newly introduced caps, and added ladder rungs. 3. Revealed elasticity. Across dated squeeze events, a squeeze stuck only where users could not withhold money (free tiers), while paid and programmatic squeezes were reversed. Coercive efficiency is highest where the squeezed party has no cancellation lever, which is dependence rather than mere willingness to pay. Method: a dual-layer approach (a reproducible hard pricing layer plus a softer behavioural layer), with all input tables web-verified to current pricing pages and provider announcements, and a revealed-elasticity layer built from public Google Trends defection-search around dated squeeze events. Honest limits: pricing measures the machinery, not dependence itself; the elasticity signals are relative and confounded; this is a profile, not a ranking. Reproducibility: stdlib-only Python over hand-coded dated CSV tables. Run python3 consumer_reliance.py. Conflict of interest: written with AI assistance from a model made by Anthropic, itself one of the providers profiled here. Anthropic rows are marked in the data, load-bearing figures were verified against primary sources, and competitors are weighted equally. License: CC BY 4.0. ORCID 0009-0003-4213-7769.
No takes yet. Share an insight, caveat, or question.
N Milton (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: