Executive Summary

• AI prediction markets are becoming an information layer for the AI economy. Combined notional volume across major platforms reached US$162.84 million, while accompanying market volume grew 279% between August and January 2026.
• The market prices technical milestones over AI diffusion. Capability, rankings, and frontier-intelligence contracts account for 56.6% of classified volume, followed by product releases and operating performance at 20.7%, corporate finance at 11.1%, and adoption, usage, and competitive positioning at 5.9%.
• This concentration creates an adoption gap: a model can top leaderboards, launch products, or achieve benchmarks without demonstrating broad usage, enterprise deployment, revenue, productivity gains, or durable economic value.
• Existing adoption-related contracts provide useful signals, such as usage, spending, requests, sign-ups, market share, and sector adoption, but remain fragmented and rely on different definitions, samples, data providers, and settlement rules.
• The AI Prediction Market proposed by OG.com and PYMNTS is designed as a demand-side complement. 
• • It converts recurring data from approximately 4,000 U.S. adults, 500 U.S. enterprise decision-makers, and public company disclosures into clearly defined, tradable contracts.
  • Its value lies not in replacing conventional AI benchmarks or prediction markets, but in connecting what AI systems can do, what organizations deploy, and what economic outcomes companies disclose. Pre-registered definitions, dual reviews, and documented settlement procedures can improve comparability, though self-reporting bias, panel selection, denominator changes, threshold cliffs, liquidity, and settlement concentration remain material risks.
• OG.com’s Alpha AI co-pilot adds a separate user-facing research layer, enabling traders to analyze markets, and build parlays within a single conversation. While Alpha aids analysis, it does not replace contract rules, settlement sources, risk disclosures, or the exchange’s final determinations.

1. The AI Market Became a Market About AI

The existing artificial intelligence (AI) category in prediction markets encompasses a wide range of questions: AI-company announcements, model rankings, product releases, AI-bubble scenarios, and capability milestones. The growth of these markets reflects a broader change in how AI progress is understood. Rather than treating technological development as a sequence of surprises, prediction markets convert it into a series of contracts with explicit deadlines, measurable outcomes, and continuously updated probabilities. A trader who buys a “Yes” contract at 30 cents effectively takes a position that the event has roughly a 30% chance of occurring, subject to market structure, liquidity, fees, and the trader’s ability to exit before settlement.

While these markets remain small compared to sports or cryptocurrency markets, they are large enough to attract sustained attention. Combined notional trading volume exceeded $162 million across major prediction market platforms from January to September 20, with monthly volume surging 279% from January to August 2026.

These contracts are not simply technology polls. They create financial incentives for participants to investigate company statements, product roadmaps, benchmark results, release patterns, regulatory developments, and competitive dynamics. 

2. Current AI Markets Traded in Prediction Markets

Traders buy a mix of contracts tied to observable AI outcomes. Most focus on capability milestones, model rankings, benchmark scores, product releases, corporate announcements, valuations, or policy events. A smaller but meaningful segment measures adoption and usage directly through indicators such as AI-adoption rates, model provider market share, AI spending per employee, requests, and sign-ups. While these contracts can provide evidence regarding specific aspects of adoption, they typically serve as narrow proxies rather than comprehensive measures of enterprise deployment, revenue, productivity, or automated task volume.

Our data indicates that AI capability-related contracts dominated the notional volume with an over 56% share, followed by product releases (20.7%) and corporate finance (11.1%). Adoption-related contracts accounted for approximately 5.9% of total volume.

Primary-Category Allocation of Current AI-Related Prediction Markets (As of September 20, 2026)

Risks of Interpreting These Markets

• Definition Risk: “Best AI,” “AGI,” “adoption,” and “market share” can carry inconsistent definitions across contracts.
• Benchmark Risk: Scores depend on versioning, benchmark contamination, hidden prompts, tool access, inference budgets, and leaderboard methodologies.
• Announcement Risk: Companies can strategically choose when to use or avoid terms like “AGI,” “release,” “launch,” or “open source.”
• Source Risk: Leaderboards, indices, valuation pages, or data providers can change, go offline, revise historical data, or employ opaque methodologies.
• Timing Risk: Release or scores recorded immediately before or after a cutoff can dictate a binary payout.
• Overlap Risk: Related monthly, threshold, and nested markets can inflate platform-level volume totals relative to the underlying independent information.
• Adoption Gaps: Correctly forecasting a top-ranked model or benchmark milestone reveals little on its own about real-world usage, revenue, workflow integration, inference volume, or productivity.

3. AI Adoption Markets

Existing market inventory provides several forms of AI adoption exposure, primarily tracking AI company revenues, usage of third-party API and AI model routers, model market-share contracts, compute demand, and related economic activity. However, these adoption indicators remain fragmented, inconsistently defined, and comparatively narrow.

3.1 Contract Design

The AI Prediction Market proposed by PYMNTS and Crypto.com turns recurring measurements of AI adoption into tradable contracts on OG.com Prediction Markets. Its differentiation is a unified measurement-and-settlement framework built around three core evidence streams:

• Consumer Behavior Monitor: Roughly 4,000 U.S. adults are surveyed monthly regarding reported AI use across work, healthcare, education, commerce, finance, and daily life. Behavioral questions generally employ a previous-day recall period, with results weighted to specified U.S. adult-population benchmarks.
• Enterprise Deployment Monitor: Approximately 500 U.S. enterprise decision-makers are surveyed quarterly on production agents, workflow automation, capital investment, software replacement, productivity, ROI, business outcomes, workforce expectations, and agentic permissions.
• Corporate Disclosure Monitor: SEC filings, earnings materials, and official corporate communications are structured and validated to measure disclosed AI investment, revenue, productivity, workforce effects, and capital allocation.

The design tracks AI through three stages often disjointed in existing markets: consumer usage, enterprise deployment, and public company reporting. Consequently, these contracts measure economic diffusion rather than just model rankings, benchmark scores, product releases, or corporate announcements.

3.2 Settlement

Each contract publishes its definition, measurement source, denominator, threshold, close date, settlement date, rounding rule, and calculation method before trading begins. Fresh monthly or quarterly measurements determine settlement, supported by nearly 30 months of historical data establishing the benchmark and methodology.

The settlement process follows a six-stage workflow: data acquisition, independent determination by two researchers, verification, sign-off by the Head of Research, embargoed delivery to the exchange, and public release. The methodology also enforces respondent-quality checks, source provenance, dual extraction for corporate figures, heightened review when disclosures near a threshold, and a finality rule for delivered values.

This structure transforms the product into a recurring measurement index wrapped in event contracts rather than a standard headline-driven prediction market. The exchange supplies the trading and clearing infrastructure, while PYMNTS Intelligence provides the settlement evidence.

3.3 Alpha: An AI Co-Pilot

Alongside its prediction market products, OG.com offers Alpha, an AI co-pilot purpose-built for prediction market research and trading. Alpha enables users to research markets and build parlays within a single conversation. In the context of AI adoption markets, it represents a user-facing intelligence layer that helps traders organize information, evaluate market-implied probabilities, and develop trading theses. However, Alpha must be distinguished from the underlying market and its settlement process: it does not replace contract rules, designated data sources, or the exchange’s final settlement, and its analysis does not guarantee trading outcomes. 

3.4 Implications for Traders

For users, the primary feature is interpretability. A contract should specify the exact population, behavior, threshold, and date matter. That precision contrasts with trading a vague thesis such as “AI will transform healthcare.” The framework also offers three practical features:

• Demand-Side Complement to Capability Markets: An AI model can rank first without achieving broad usage, revenue, workflow integration, or productivity gains. PYMNTS contracts are designed to target these demand-side outcomes directly.
• Recurring Signal Rather than a One-Off Headline: Monthly consumer data alongside quarterly enterprise and disclosure readings show whether adoption is accelerating, slowing, or diverging between consumers, enterprises, and public companies.
• Live Forecast of a Measurable Number: Research provides historical readings, while contract prices express where traders expect the next reading to land. Consequently, price changes can reflect shifts in market conviction before underlying surveys or disclosures are released.

The framework can also capture “missing receipts,” instances where AI replaces searches, support interactions, or human tasks without generating a clear transaction record. Survey-based behavioral measurement can reveal these shifts earlier than financial statements, though it should be read as reported behavior rather than verified usage telemetry.

3.5 Limitations and Risks

The PYMNTS design differs from  typical headline-driven prediction markets by predefining measurement sources, contract terms, settlement processes, and conflict controls. However, methodological transparency does not eliminate measurement error. The main risks stem from what is measured, who is sampled, how results are interpreted, and how continuous economic phenomena are converted into binary payouts.

4. Conclusion

AI prediction markets are becoming a meaningful information layer for the AI economy, but they are not yet a clean proxy for AI adoption. The current market is dominated by capability milestones and headline outcomes. Although platforms offer adoption and usage contracts, these remain comparatively narrow, fragmented, and dependent on inconsistent data sources and settlement rules.

OG.com’s broader product ecosystem pairs recurring prediction markets with Alpha, an AI co-pilot designed to help users research markets and build parlays around those opportunities.

OG.com and PYMNTS are developing an integrated market for AI adoption indicators, with PYMNTS providing the measurement framework and OG.com Prediction Markets delivering the trading and clearing infrastructure. OG.com’s Alpha adds a distinct, user-facing research and analysis layer. The overarching opportunity is to convert these measurements into a recurring market for AI diffusion.

The investment case rests on PYMNTS’ methodological integration. Pre-registered definitions, explicit denominators, recurring observations, source documentation, independent reviews, and conflict controls make AI adoption contracts more comparable and actionable than isolated event markets. If sufficient liquidity develops, the contracts can provide early signals of where adoption is accelerating, where enterprise deployment is stalling, and whether AI investment is translating into measurable economic outcomes.

However, the opportunity remains unproven with notable risks. Survey-based data remains vulnerable to sampling and self-reporting bias. Corporate disclosures reflect management’s communication choices rather than full operational realities. Conditional enterprise samples can be small, binary thresholds can create artificial payout cliffs, and PYMNTS remains a concentrated source of measurement and settlement judgment. While transparent rules reduce settlement uncertainty, they do not eliminate measurement risk.

Ultimately, PYMNTS’ AI adoption contracts offer a compelling demand-side complement to capability-focused AI markets. While the underlying methodology establishes procedural safeguards, the product’s practical credibility will depend on its observed settlement record, liquidity, and ability to produce information beyond existing data sources. If PYMNTS achieves these measures, it could establish a valuable new market category: tradable intelligence on AI’s economic diffusion.

Read the full report: Pricing AI’s Progress: From Frontier Milestones to Economic Adoption

Interested to know more? Access exclusive reports by signing up as a Private member, joining our Crypto.com Exchange VIP Programme, or collecting a Loaded Lions NFT.

Authors

Crypto.com Research and Insights team

Get the latest market, DeFi & NFT updates delivered to your inbox:

Be the first to hear about new insights: