AI-search citation bias measurable through $14 GEO purchase
Researchers have demonstrated that ordinary publication on low-barrier platforms can steer AI-search outputs, and that commercial services already sell this influence for pocket change. A paper submitted to arXiv on October 8, 2026, presents a measurement framework that quantifies how public posts become AI-search citations across ten platforms.
How AI search differs from traditional search
Traditional search maps keywords to ranked pages. AI search retrieves pages, filters sources, selects citations, and generates answers before users see any source list. This selection layer amplifies source bias and turns source choice into a security question. Platforms reveal little about citation selection logic. Citations change over time. The web contains so much background content that attributing later answer changes to specific posts is difficult without controlled experiments.
Measurement framework combines three techniques
The framework uses cross-platform citation mapping, publication-barrier testing, and marker-controlled publication experiments. Across ten AI-search platforms, the team analyzed 17,211 citation instances spanning 6,356 unique source domains. They tested 22 publication platforms tied to cited domains for account-setup and posting barriers.
Citations concentrate in a few platform-specific domains
The top-20 domains captured between 20.5 percent and 70.8 percent of per-platform citations. Fifteen of the 22 tested publication platforms had low or medium barriers for both account creation and posting. This means anyone can publish on the very sources that AI-search platforms cite most heavily.
Fabricated concept cited by eight platforms within a week
In controlled experiments, researchers published a fabricated concept on preferred platforms. Eight of ten AI-search platforms cited that concept within seven days. One article on a high-preference platform generated more citation impact than over twenty matched posts on low-preference platforms. Platform preference outweighs content volume for citation influence.
Commercial GEO service injects marked content in under an hour
A fourteen-dollar generative-engine-optimization purchase produced thirteen public posts. One AI-search platform cited the GEO-posted content with the researchers' designed markers within one hour. The path from payment to citation is short, cheap, and operational today.
What this means for developers building on AI search
Teams that rely on AI-search citations for grounding, fact-checking, or retrieval-augmented generation should treat citations as noisy signals. A citation does not imply editorial vetting. It may reflect only that the source domain has low publication barriers and high platform preference. Developers should add provenance tracking, cross-reference citations against authoritative sources, and consider human-in-the-loop review for high-stakes applications.
Framework enables ongoing audit of citation integrity
The measurement approach gives platform operators and third parties a repeatable method to audit citation pathways. Regular marker-controlled publication experiments can detect shifts in platform preference, barrier changes, or new commercial injection services. Without such auditing, the gap between what appears authoritative and what is merely accessible will continue to widen.