A study conducted over 14 weeks across 467 locations in the United States found that Google Search results for abortion services were less likely to return abortion clinics in poorer and rural areas. The research highlights how search engine algorithms can create disparities in health access by failing to consistently direct users to legitimate abortion providers, instead sometimes surfacing crisis pregnancy centers. The study noted that performance improved following a major algorithm update, suggesting the issue was tied to the search system's ranking logic.
Google Search reportedly returned fewer abortion clinics for searches from poorer and rural areas, particularly ones with Targeted Regulation of Abortion Providers (TRAP) laws.
Risk classification
- Primary risk domain: 1 Discrimination & Toxicity
- Primary risk subdomain: 1.3 Unequal performance across groups
Google Search performed worse (returned fewer legitimate abortion clinics) for users in poorer and rural areas, representing unequal performance across demographic and geographic groups.
Causal factors
- Entity: AI
- Intent: Unintentional
- Timing: Post-deployment
The search ranking algorithm (AI) unintentionally returned fewer legitimate abortion clinic results for poorer and rural areas post-deployment.
EU AI Act risk tier
- Risk tier: 4 Minimal or No Risk
Minimal or No Risk: Google Search is a general-purpose search engine and does not fall under the high-risk categories like medical diagnostics or critical infrastructure, thus posing minimal direct regulatory risk under the EU AI Act.
AI system and alleged parties
- AI system: Google Search (Google)
- AI purpose: Content Search; Content Curation
- Behaviour type: Assistant
- Alleged developer: Google
- Alleged deployer: Google
- Alleged harmed parties: women in need of abortion services, women having unexpected or crisis pregnancies
Harm severity
Severity ratings are based on this scale.
Highest direct severity in any category: Minor. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.
- Physical: direct Negligible, indirect Negligible
- Infrastructure: direct Negligible, indirect Negligible
- Property: direct Negligible, indirect Negligible
- Financial: direct Negligible, indirect Negligible
- Environmental: direct Negligible, indirect Negligible
- Malicious content: direct Negligible, indirect Negligible
- Differential treatment: direct Minor, indirect Minor
- Civil rights: direct Negligible, indirect Minor
- Democracy: direct Negligible, indirect Negligible
- Privacy: direct Negligible, indirect Negligible
- Psychological: direct Negligible, indirect Negligible
- Epistemic: direct Negligible, indirect Negligible
- Child sexual exploitation and abuse: direct Negligible, indirect Negligible
Differential treatment
Reported: Yes, the report explicitly describes differential treatment where fewer abortion clinic results were returned for searches from poorer and rural areas.
Directly caused: Google Search returned fewer legitimate abortion clinic results for queries originating from poorer and rural areas, as well as areas with TRAP laws.
Indirectly caused: This algorithmic disparity resulted in unequal access to healthcare information based on geographic and socioeconomic status.
Inferred additional harm: We infer that thousands of low-income and rural search engine users experienced systematically worse search utility and reduced access to reproductive healthcare information compared to wealthier, urban users.
Civil rights
Reported: Yes, the report notes important implications concerning health access quality and equity, which relates to reproductive rights and equal access to healthcare.
Indirectly caused: The algorithmic bias indirectly restricted access to reproductive healthcare services for women in poorer and rural areas, impacting their reproductive autonomy.
Inferred additional harm: We infer that many women's reproductive rights and health equity were compromised due to being systematically directed away from legitimate abortion providers.
People affected
Potential causes
Technology
- Algorithmic Bias: Search algorithm favored closer CPCs over actual abortion clinics.
- Legacy Search Algorithm: Performance was poorer before a major search algorithm update.
Data Inputs
- Geographic Disparities: Fewer AC results in data for poorer and rural search areas.
Process and Methods
- Inadequate Result Filtering: System failed to filter out misleading CPCs for abortion service queries.
Regulatory Environment
- TRAP Laws Influence: Targeted regulation of abortion providers limited available AC data.
Information quality
- Classification confidence: High
- Reason for confidence: The report is based on a structured academic study with clear empirical data, methodology, and statistical findings regarding Google Search results across multiple locations and weeks. There is little ambiguity about the findings, though the real-world human impact must be inferred.
- Ambiguities identified: The exact internal mechanics of Google's ranking algorithm and the specific changes in the algorithm update are not detailed.
- Alternative interpretations: The disparities could be interpreted as reflecting underlying real-world disparities in clinic availability rather than purely algorithmic bias, though the study controls for location and notes improvement after an algorithm update.
An academic study revealed algorithmic bias in Google Search, where queries in poorer and rural US areas returned fewer legitimate abortion clinics compared to wealthier areas. While highlighting significant issues of healthcare access equity and algorithmic fairness, the incident does not pose a direct threat to US national security.
- Overall national security impact: Minor
- Scope: Single nation
- Primary target: United States
- Alleged perpetrator: Unknown
Threat characteristics
- Imminence: Long-term. Represents an ongoing algorithmic performance issue that was subsequently improved via updates, rather than an active, imminent threat.
- Autonomy: Human-supervised. The search engine ranking algorithm operates autonomously to generate results, but is subject to human oversight and algorithmic updates.
- Novelty: Established threat. Algorithmic bias and demographic disparities in search engine results are well-documented, established issues in AI systems.
Impact by dimension
- Physical security: Negligible. No physical systems, critical infrastructure, or human safety systems were compromised or targeted in this incident.
- Information security: Negligible. No evidence of state-sponsored disinformation, intelligence compromise, or adversarial information warfare operations.
- Sovereignty: Negligible. Does not impact state authority, territorial control, electoral systems, or core government decision-making processes.
- Economic security: Negligible. No strategic technology theft, financial system attacks, or critical supply chain disruptions occurred.
- Societal stability: Minor. Algorithmic bias restricted access to reproductive healthcare information for vulnerable rural and low-income populations, representing a localized civil rights and public health equity issue rather than a national security crisis.