Global MDI / Moat Type Index
Compare fourteen types of competitive advantage
Compare fourteen moat types on research evidence, AI durability and momentum. This view names no companies.
Measured snapshot, not a live feed
Read the measures
Sort each measure separately; the columns are never blended into a hidden score.
- Evidence
- Distinct literature works in the graph associated with the type, from titles and available abstracts. Also shown: distinct works in the graph that cite them, and the publication-year span.
- AI durability
- The direction our standard expects for the type when agents can copy software fast: holds, depends, partly holds or erodes. A published expectation, not a measured outcome.
- Momentum
- Change in yearly research papers on a theme matched to the type, 2023–2025 average against 2018–2020 average. Literature activity, not market performance.
Evidence × AI durability
Evidence uses a logarithmic scale. Durability is an ordered expectation, not a numerical outcome. The table below remains the source of truth.
- Physical & tacit process37 works · depends
- Identity-bound network55 works · depends
- System-of-action embedding32 works · depends
- Rights-cleared data3 works · holds
- Licences & approvals123 works · holds
- Scarce physical, energy & supply positions12 works · holds
- Human-chosen brand393 works · erodes
| Rank | Moat type | |||
|---|---|---|---|---|
| 1 | Human-chosen brandTrust & accountability | 393works · cited by 792 · 1976–2025 | erodesAgents and AI answers cut how often a human picks by brand. | No matched theme yet |
| 2 | Licences & approvalsRights, positions & complements | 123works · cited by 507 · 1934–2026 | holdsAgents speed up paperwork, not a regulator's calendar; approvals can still become routine or be revoked. | +85%yearly papers |
| 3 | Identity-bound networkCustomer lock & network | 55works · cited by 133 · 1985–2026 | dependsHolds for networks of identified, transacting parties; erodes for networks built on content alone. | +156%yearly papers |
| 4 | Physical & tacit processCost & learning | 37works · cited by 701 · 1981–2018 | dependsHolds when the know-how lives in physical operations; erodes once it is written down where agents can read it. | No matched theme yet |
| 5 | System-of-action embeddingCustomer lock & network | 32works · cited by 604 · 1982–2017 | dependsHolds where replacing it means re-validating real work; erodes where lock-in rests on fees or data export. | +58%yearly papers |
| 6 | Scarce physical, energy & supply positionsRights, positions & complements | 12works · cited by 15 · 1999–2021 | holdsAgents cannot create sites, grid capacity or supply; positions can still be cyclical or revoked. | No matched theme yet |
| 7 | Rights-cleared dataRights, positions & complements | 3works · cited by 1 · 2004–2016 | holdsExclusive, documented rights do not get easier to copy as agents improve. | No matched theme yet |
| 8 | Capex-bound scaleCost & learning | 0Not yet covered | dependsHolds when tied to plant, energy or regulated capacity; fades when the scale is only compute spend. | +120%yearly papers |
| 8 | Proprietary learning loopCost & learning | 0Not yet covered | partly holdsHolds only while exclusive outcomes keep it ahead of each new model generation. | No matched theme yet |
| 8 | Co-specialised complementary assetsRights, positions & complements | 0Not yet covered | holdsWhen software is easy to copy, value moves to the assets a rival still needs to reach the customer. | No matched theme yet |
| 8 | Audited track record & verified reputationTrust & accountability | 0Not yet covered | holdsDated, third-party records take calendar time that a clone cannot buy. | No matched theme yet |
| 8 | Liability absorptionTrust & accountabilityHypothesis | 0Not yet covered | holdsStanding behind outcomes gains value as liability lands on whoever ships AI; an early, untested market. | No matched theme yet |
| 8 | Counter-positioningPosition in the agent economy | 0Not yet covered | partly holdsA strong entry move with a shorter life: it fades as incumbents adapt. | No matched theme yet |
| 8 | Trust-anchor positionPosition in the agent economyHypothesis | 0Not yet covered | holdsAgents rely on issuers and certifiers they can check; untested, and open protocols may add rivals. | No matched theme yet |
Sorted by evidence, most associated works first.
Momentum reference: the median research theme in the graph changed +120% over the same windows (11 themes). Evidence covers 7 of 14 types and momentum 4 of 14 so far; the graph keeps growing.
Limits
Read them first
- Associations are not causal evidence: a matched work discusses the capability, it does not prove the moat works.
- Zero matched works is a coverage gap in this snapshot, not evidence against a moat type.
- Momentum counts come from broad literature searches, which grow every year; compare types with each other and with the reference, not with zero.
- AI durability is the standard's expectation, worded at capability level; no moat type has yet been shown to predict profit persistence in the AI era.
- No company is named or ranked. Not investment advice; no prediction of funding, revenue, valuation or unicorn outcomes.
The method behind the Moat Index and the data behind the graph stay closed; this page publishes measured counts only.