The Rise of AI Search: Implications for Information Markets and Human Judgement at Scale

We executed 24,000 search queries in 243 countries, generating 2.8 million AI and traditional search results in 2024 and 2025. We found a rapid global expansion of AI search and key trends that reflect important, previously hidden, policy decisions by AI companies that impact human exposure to AI search worldwide. From 2024 to 2025, overall exposure to Google AI Overviews (AIO) expanded from 7 to 229 countries, with surprising exclusions like France, Turkey, China and Cuba, which do not receive AI search results, even today. While only 1% of Covid search queries were answered by AI in 2024, over 66% of Covid queries were answered by AI in 2025 – a 5600% increase signaling a clear policy shift on this critical health topic. Our results also show AI search surfaces significantly fewer long tail information sources, lower response variety, and significantly more low credibility and right- and center-leaning information sources, compared to traditional search, impacting the economic incentives to produce new information, market concentration in information production, and human judgment and decision-making at scale. The social and economic implications of these rapid changes in our information ecosystem necessitate a global debate about corporate and governmental policy related to AI search.

💡 Research Summary

The paper presents a large‑scale empirical investigation of the rapid diffusion of AI‑driven search and its consequences for information markets and human judgment. The authors executed 24,000 queries across 243 countries in 2024 and 2025, generating 2.8 million search results from both AI‑augmented (Google AI Overviews, AIO) and traditional keyword‑based engines. By running the exact same 12,000 queries in both years, they isolate platform policy effects from changes in user behavior.

Key findings:

1. Geographic expansion – AI Overviews were available in only seven countries in 2024 but reached 229 countries by 2025. Notable exclusions (France, Turkey, China, Cuba) persisted, while countries such as Ukraine, Russia, Israel, and Venezuela moved from zero exposure to over 55 % AI answer rates. The variation is driven largely by corporate policy rather than internet penetration.
2. Topic and query‑type effects – Question‑style queries returned AI answers 60 % of the time, statements 37 %, and navigational searches only 12 %. Health and general‑knowledge queries saw AI exposure rise from roughly half to two‑thirds of all results worldwide. The most striking shift concerns COVID‑19 queries: AI answered just 1 % of such queries in 2024 but 66 % in 2025—a 5,600 % increase that aligns with a U.S. executive order reversing prior misinformation‑suppression policies.
3. Content diversity and source quality – AI results dramatically reduce “long‑tail” exposure: they link far more often to the top 1 000 sites and far less to the 1 000–1 000 000 range compared with traditional SERPs. Consequently, traffic to smaller publishers collapses, raising concerns about market concentration. Moreover, AI‑cited sources are statistically more likely to be low‑credibility, right‑leaning or centrist outlets, and less likely to be left‑leaning, indicating a systematic bias in the synthesis layer.
4. User interaction – Presence of an AI summary cuts outbound click‑through roughly in half (8 % vs. 15 % in a July 2025 Pew study) and raises the zero‑click rate to 80 % versus 60 % without AI. This “one‑voice” presentation reduces exposure to diverse viewpoints and threatens the ad‑revenue model that sustains original journalism and expert curation.
5. Trust paradox – Adding references and citations to AI answers boosts user trust even when those references are inaccurate or hallucinated. The trust effect is strongest among users with lower education or non‑technical occupations, making vulnerable populations more prone to accept misinformation.

Methodologically, the study combines large‑scale data collection via the serpAPI, logistic regression to assess the predictive power of country, topic, and query style, and content analysis using the Media Bias/Fact‑Check database to quantify credibility and political leaning. The authors also reference complementary experiments (e.g., reduced time‑on‑task, over‑reliance trade‑offs) to contextualize cognitive impacts.

Policy implications are foregrounded: the authors call for mandatory transparency of AI‑search rollout decisions, enforced disclosure of source URLs, and public‑sector auditing mechanisms. They warn that unchecked corporate discretion can silently reshape the global information ecosystem, concentrating attention on a handful of dominant publishers while marginalizing niche or dissenting voices.

In conclusion, the paper documents a swift, policy‑driven expansion of AI search that reshapes how billions access information, alters economic incentives for content creation, and introduces new biases into human decision‑making. It urges a coordinated global debate involving regulators, AI firms, and civil society to balance AI’s convenience with the preservation of a pluralistic, credible information marketplace.