VPS Hosting
LIVE30 standardized discovery queries covering DevOps, budget, raw performance, managed onboarding and EU residency.
1DigitalOcean87.8
2Hetzner87.4
90 LIVE EVALUATIONSView Full Leaderboard
GLOBAL AI BENCHMARK • 100% DETERMINISTIC
See what AI
recommends.
We run standardized discovery queries through Claude, Gemini, and GPT every cycle, parse every named product out of the raw answers, and publish where the models actually agree. No affiliate links, no sponsored placements — just reproducible model consensus.
Zero Affiliate Bias
No commissions, no sponsored slots, no paid inclusion — ever.
Multi-Model Consensus
Three model families, weighted equally. One vendor never decides a rank.
Auditable Prompts
Every query is published, versioned and downloadable as JSON.
Continuous Evaluation
Re-run on a fixed cadence and on every model version bump.
TRACKED CATEGORIES
One leaderboard per decision
Each vertical gets its own frozen prompt set. A category only goes live once it clears 60 evaluations across active frontier model families.
Transactional Email
LIVE30 standardized discovery queries covering deliverability, React/Next.js SDKs, high-volume pricing and EU residency.
1Postmark90.9
2Resend80.6
90 LIVE EVALUATIONSView Full Leaderboard
VPN Services
LIVE30 standardized discovery queries covering privacy jurisdictions, verified no-logs audits, WireGuard and streaming.
1Proton VPN71.5
2Mullvad71.2
90 LIVE EVALUATIONSView Full Leaderboard
AI Code Assistants
LIVE30 standardized discovery queries covering IDE autocomplete, codebase indexing, terminal agents, local LLMs and refactoring.
1GitHub Copilot80.2
2Cursor79.1
60 LIVE EVALUATIONSView Full Leaderboard
LLM Gateways
PLANNEDRouters, proxies and observability layers — where the models are being asked to recommend their own infrastructure neighbours.
- prompt set
- scoping
- entity list
- scoping
- evaluations
- 0 / 60
Self-reference bias controls under design.
WHY THIS EXISTS
Traditional Google vs. AI recommendation
DIMENSION
GOOGLE SERP LISTICLE
AIPICKRANK CONSENSUS
Ranking driver
GOOGLE SERP LISTICLEBacklinks, domain authority and ad spend
AIPICKRANK CONSENSUSFrequency and declared position of a brand inside raw model answers
Incentive
GOOGLE SERP LISTICLECommission per signup, highest payout ranks first
AIPICKRANK CONSENSUSNone. No affiliate links, no sponsorships, no paid inclusion
Reproducibility
GOOGLE SERP LISTICLEPersonalised per user, location and session
AIPICKRANK CONSENSUSFrozen prompt set, temperature 0.0, run logs published per cycle
Coverage
GOOGLE SERP LISTICLEOne author's shortlist, usually the same ten brands
AIPICKRANK CONSENSUS2 model families × 20 providers × 60 evaluations per cycle
Update cadence
GOOGLE SERP LISTICLERewritten when organic traffic drops
AIPICKRANK CONSENSUSScheduled re-runs plus a forced re-run on every model version bump
Auditability
GOOGLE SERP LISTICLEMethod undisclosed, scores unexplained
AIPICKRANK CONSENSUSAll 30 prompts, parser rules and raw outputs downloadable as JSON
THE FRAMEWORK
Three steps, all of them auditable
Unbiased Prompts
Queries are written the way a buyer actually asks, never the way a brand would like to be asked. No brand names inside the prompt, no leading adjectives, no "best X for Y" template padding.
30 queries · frozen · versioned
Deterministic Parsing
Answers are parsed by rule, not by a second model: entity aliases are resolved against a fixed registry, and the declared order in the answer becomes the recorded position.
temp 0.0 · seed pinned · no judge model
Equal-Weight Consensus
Each model family contributes equally to the Pick Score. A provider that only one vendor loves cannot climb; agreement across families is the entire signal.
1/2 GPT · 1/2 Gemini