Perplexity AI
Perplexity
Answer engine that routes queries across multiple frontier models with citations.
Impact Indicators
Job Displacement
No verified data attributes job losses to Perplexity. Labor research instead measures broader generative-AI occupational exposure and early employment shifts across AI-exposed occupations, not product-specific displacement. Exposure and relative employment shifts are not evidence that Perplexity displaced jobs.
Privacy & Surveillance
No regulator finding or enforcement action against Perplexity over personal-data handling has been verified. Independent security researchers demonstrated indirect prompt-injection weaknesses in the Comet browser that could cause its agent to disclose data it had access to; reporting indicates a patch followed disclosure. These are proof-of-concept demonstrations, not confirmed widespread exploitation. Perplexity's own privacy notice states it does not sell personal data or send conversation content to advertisers; company disclosure is not independent verification.
Water Usage
No measured Perplexity-specific water-use figure exists, and no facility identified as serving Perplexity has a verified water figure. Peer-reviewed and national-laboratory research establishes that U.S. data centers require substantial direct cooling water plus indirect water embodied in electricity generation, but those figures are modeled at industry level and cannot be attributed to Perplexity. Third-party provider water data remains provider-level.
Energy & Carbon
No independently verified Perplexity-specific energy or emissions figure exists, and Perplexity is absent from published inference-energy benchmarks. Company technical disclosures establish that its workloads run on third-party GPU and cloud infrastructure, while national-laboratory research documents U.S. data-center electricity demand at industry level. Cloud relationships, hardware counts and capacity are not measured electricity use or emissions.
U.S. reporting requirements do not yet provide standardized, model-level disclosure for many AI human and environmental impacts. When U.S. evidence is incomplete, DoctorKnow uses verifiable EU regulatory disclosures, public-sector data, independent research, and reputable journalism to provide the strongest evidence currently available.
Evidence scope is identified as model-specific, provider-level, facility-level, regional/grid, or industry-level.
Value alignment
52/100
Performance fit
82/100
Powered by: Claude (Anthropic), ChatGPT (OpenAI)
Visit site →Scores use a 0–100 scale. Higher scores indicate stronger performance or lower concern; 100 is the best possible rating.
Reasoning
82
Coding
70
Writing
80
Analysis
84
Speed
86
Scores use a 0–100 scale, where 100 represents the strongest outcome and 0 the weakest.
What Perplexity does
Perplexity is built by Perplexity AI and is positioned for research, chat and analysis. Our readings score it strongest on response speed and data analysis.
Its weakest measured axis is code generation (70/100), so plan around that when the work depends on it. Access is freemium, and the weights are closed, so you can only reach it through the vendor's own service.
Ethical stance
Transparency reads 52/100. Some model cards and evaluations are public, but training data and methods are only partly described.
Labor and job-displacement posture reads 48/100. There is partial disclosure on data work and workforce impact, with clear gaps.
Regulatory posture reads 52/100. The vendor cooperates with regulators selectively, and lobbies against parts of it.
Privacy and data handling
Privacy and data handling reads 60/100. Training opt-out and retention controls exist, but defaults or consumer tiers are less protective.
Because the weights are closed, every prompt is processed on vendor infrastructure — treat regulated or confidential data accordingly.
Energy, carbon and water
Carbon footprint reads 52/100. The energy mix is mixed, and disclosure covers some but not all serving regions.
Water and local impact reads 50/100. Some sites use evaporative cooling in regions with moderate water stress.
Speed and efficiency also matter here: a 86/100 speed reading means fewer compute-seconds per answer for routine work.
Diagnostic notes
Estimated scores pending confirmation. Multi-model aggregator, so impact varies with the routed provider.
Readings are DoctorKnow.ai's own 0–100 assessments, compiled from vendor documentation and public reporting, and revised as new disclosures land. They are a comparison aid, not a certification.