Can artificial intelligence shape the future of prediction markets as financial dynamics?

The rise of prediction markets has reopened the debate over the boundary between gambling and financial markets. While in the United States the regulatory discussion continues to oscillate between treating these products as bets or as financial instruments, this article examines the impact of artificial intelligence on that debate. Through the development of tools based on machine learning and advanced analytics, it is argued that AI may progressively reduce the element of chance by strengthening predictive capabilities and informational asymmetries between participants. From this perspective, the article argues that such evolution could shift prediction markets towards being understood as financial vehicles, unless the gambling industry develops regulatory alternatives capable of responding to these new dynamics.

Introduction

Over recent years, prediction markets have become one of the most sensitive debates within both the gambling industry and financial markets. In the United States, the regulatory discussion has largely focused on whether these products should be understood as traditional bets or as financial instruments based on information and economic projections. The debate remains open and is currently before the courts. As noted in the recent NEXT article “The Great Divide”[1], those defending these markets as part of the financial sector argue that they share important similarities with traditional financial mechanisms, including trading activities and the analysis of future events, while also attracting a different type of consumer from the traditional gambling user. However, much of this discussion continues to develop on the assumption that these markets remain fundamentally dependent on chance.

Within this debate, one element has still received limited attention, namely the impact of artificial intelligence, particularly machine learning. As these technologies evolve, predictive systems are becoming capable of processing vast amounts of information, identifying complex patterns and developing increasingly sophisticated probabilistic models. Several recent studies on machine learning applied to sports betting and prediction markets show that these systems not only improve the accuracy of certain predictions but also identify inefficiencies within markets. This raises an important question as to what happens when uncertainty no longer depends solely on human intuition or basic analytics but instead becomes increasingly influenced by advanced computational capabilities.

This article argues that artificial intelligence does not entirely eliminate chance. However, due to its rapid evolution, it may progressively transform the way uncertainty is distributed within certain gambling verticals. In environments driven by large volumes of data and real-time analytics, competitive advantage may begin shifting away from traditional chance and towards predictive dynamics, information processing and the analysis of algorithmic asymmetries. From this perspective, the evolution of AI could reopen, at least from a technical and regulatory standpoint, the discussion as to whether certain gambling verticals are gradually moving closer to structures commonly associated with prediction markets under a financial market logic.

Analysis

Those defending prediction markets from a financial perspective argue that these products share important similarities with certain dynamics commonly found in trading activities and derivatives markets. One of the main arguments supporting this position is that, unlike traditional gambling, participants may buy, sell or close positions before the final event occurs, much like certain financial instruments. This ability to manage positions in real time, while generating economic benefits, has become one of the main elements used to distinguish these models from traditional concepts of chance.

Within this logic, information plays a central role. In many trading environments, participants seek advantages through the analysis of data, trends, news, market behaviour, political developments and future projections, among other factors. A similar dynamic can be observed in certain prediction markets, where participants no longer act solely on intuition or entertainment, but rather on the interpretation of information and probabilities. As noted in NEXT’s article “The Great Divide”, one of the most common arguments advanced by supporters of these models is that their users behave more like traders or specialized analysts than traditional gambling consumers.

Closely linked to the role of data and the way it is processed within these markets, another element emerges which could substantially transform this discussion, namely the accelerated evolution of artificial intelligence, particularly through systems based on machine learning. According to Stanford University’s 2025 AI Index Report, the growth of these models over the last decade has been exponential. The report states that “parameter counts have risen sharply since the early 2010s, reflecting the growing complexity of their architecture, greater availability of data, improvements in hardware, and proven efficacy of larger models”[2].

The AI Index Report also highlights significant improvements in benchmarks designed to measure complex reasoning, programming and advanced problem-solving capabilities. Between 2023 and 2024, results achieved by AI models increased considerably across these types of tests. In certain scenarios, language model-based agents even outperformed humans in specific programming tasks under limited time constraints.

It is also important to highlight the increasingly sophisticated predictive capabilities being developed in real-world scenarios. The same report refers to models such as “Aurora” and “GenCast”, designed for advanced climate forecasting and capable of generating highly accurate predictions relating to cyclone trajectories, air quality, ocean waves and complex weather phenomena at significantly lower computational costs than traditional systems. In other words, modern machine learning systems are no longer limited to automating tasks or processing information, but are increasingly capable of optimizing complex predictive processes through the large-scale analysis of data, historical patterns and dynamic probabilities.

Nevertheless, the gambling industry has not remained isolated from the use of artificial intelligence systems based on machine learning. Initially, many of these tools were used in marketing processes, user segmentation, customer interaction and commercial optimisation within digital betting platforms. More recently, the evolution of these systems has also started to play an important role in responsible gambling initiatives and the early detection of at-risk players.

A study developed by researchers from the International Gaming Institute at the University of Nevada Las Vegas (UNLV), Harvard Medical School, the University of Calgary, the University of Sydney and Washington State University notes that artificial intelligence systems have become central tools in harm prevention efforts within the gambling industry. The study highlights that one of the main attributes of machine learning lies in “…its ability to learn complex and nonlinear relationships in the data”[3], allowing these models to “…target players according to markers defined by both their gambling patterns…”[4], thereby enabling the detection of potentially problematic gambling behaviours through the individualised analysis of user conduct.

Another example of these predictive capabilities can be observed in gambling verticals such as sports betting. A relevant example is the study developed by Luke Boll, a researcher at the University of Michigan, on neural networks applied to predicting outcomes in college football. In this paper, the author states that “In recent years, neural networks have emerged as a powerful tool for predicting sports outcomes due to their ability to learn complex relationships between statistical inputs and game outcomes. These challenges have led to the development of various prediction models that utilize machine learning and statistical techniques to extract insights from large volumes of data, with the goal of improving the accuracy of predictions in the ever-evolving landscape of college football.”[5]

The research compiled thousands of historical data points derived from sports statistics and Las Vegas betting lines, subsequently building a database containing more than 1,700 variables related to performance, historical results and betting behavior. Based on this information, the researchers trained different machine learning models, including neural networks, in order to compare their predictive capabilities against the betting lines traditionally used by the industry. The study concluded that tools such as lasso regression and feature selection[6], both commonly associated with machine learning techniques, significantly improved the accuracy and robustness of predictive models. However, the study also demonstrated that Las Vegas betting lines constituted one of the strongest predictors within the model, to the extent that removing this variable considerably reduced the performance of the analysed systems.

A study entitled “A Systematic Review of Machine Learning in Sports Betting: Techniques, Challenges, and Future Directions”[7]” states that “…machine learning has significantly impacted the sports betting landscape by improving both the accuracy of predictions and the efficiency of betting strategies”, further noting that these techniques “…have been employed to identify mispriced odds offered by bookmakers, presenting opportunities for savvy bettors to capitalize on these inefficiencies”.

This demonstrates how artificial intelligence is beginning to reshape the traditional understanding of chance within sports betting, opening the discussion as to whether the modern bettor increasingly resembles an analyst of information and variables, similar to certain profiles within the financial sector, where data processing capabilities and analytics may partially reduce uncertainty.

Nevertheless, it is essential to distinguish prediction markets from sports betting based on the nature of the underlying event. In prediction markets, the object of prediction is generally linked to structural, economic, political or social variables whose evolution may be modelled through aggregated information and rational analysis, allowing uncertainty to be progressively and cumulatively reduced.

By contrast, sporting events contain an irreducible element of randomness associated with inherently unpredictable factors, including real-time human decisions, errors, physical conditions, refereeing decisions and unforeseen events, all of which limit the scope of any predictive system. For this reason, although artificial intelligence may significantly improve accuracy within sports betting, we believe it is still incapable of eliminating chance in these events.

It may therefore become necessary to propose a regulatory test based on at least three variables for classification purposes, namely (i) the degree to which chance may be reduced, (ii) the role of information and predictive capability, and (iii) the possibility of actively managing positions.

Another regulatory alternative may lie not necessarily in the creation of entirely new legal categories, but rather in the reinterpretation or adaptation of existing industry models, such as certain exchange betting schemes whose operational logic increasingly resembles dynamics commonly associated with prediction markets.

This may represent only the beginning of AI’s impact on the gambling industry. As previously noted, the evolution of these technologies is occurring at remarkable speed, raising the possibility that gambling environments may soon involve artificial intelligence agents with a far more active role in digital interaction through the autonomous execution of tasks based on pre-established prompts.

In this context, artificial intelligence is already beginning to generate new asymmetries between users with access to advanced predictive processing tools and those who continue to participate solely through traditional human capabilities[8]. The discussion therefore moves beyond automation itself and towards the possibility that certain participants may develop structurally superior analytical advantages within prediction markets and gambling verticals, potentially disrupting the balance between users.

At the same time, advances in algorithmic personalisation models are transforming the way users interact with digital betting platforms. The article “AI Personalization and Its Influence on Online Gamblers’ Behavior”[9]  notes that modern machine learning systems are already capable of dynamically adapting content, recommendations and digital experiences based on each user’s individual behavior, thereby creating highly personalised environments in real time.

This raises the possibility of environments where bettors are no longer simple artificial intelligence agents, as is already the case today, but highly specialised and continuously trained systems capable of developing predictive and analytical advantages within increasingly personalised betting markets.

This demonstrates that the impact of machine learning within gambling is no longer limited to operational, commercial or responsible gambling processes. It is progressively transforming structural dynamics within the sector, as well as the traditional understanding of the gambling consumer itself. The growing ability of these systems to process vast amounts of information and generate predictive advantages is already having a direct impact on highly data-intensive verticals such as sports betting.

This phenomenon also appears to impacting the profile of users within certain gambling verticals. The traditional consumer, historically driven mainly by intuition, entertainment or chance, is beginning to coexist with participants using advanced analytical tools, data processing systems and predictive models to optimise decision-making. It may even mark the beginning of competitive dynamics between human users and AI agents capable of directly interacting with betting platforms.

The accelerated evolution of machine learning, together with the future development of increasingly specialised artificial intelligence agents and systems, will likely create new regulatory and conceptual challenges for the gambling industry. In particular, it may require a reconsideration of how uncertainty and chance are understood within certain digital markets. This also represents a challenge from a regulatory perspective, given the degree of agility and flexibility that future frameworks may require in response to the transformations expected within the industry.

These developments are already raising new technical and regulatory questions. If certain artificial intelligence tools become capable of generating increasingly sophisticated predictive advantages, it becomes necessary to examine whether gambling verticals such as sports betting or eSports, both highly data-intensive environments well suited to predictive AI systems, still operate under traditional concepts of chance or are moving progressively closer to information and prediction markets more commonly associated with financial market dynamics.

It may also become necessary to establish clearer rules governing the use of artificial intelligence within gambling platforms, both from the perspective of operators and users, including potential obligations relating to transparency, the identification of automated systems and the use of certain predictive models. In other words, the future regulatory challenge may not lie solely in controlling the use of artificial intelligence, but rather in determining the extent to which these technologies can alter the balance of uncertainty upon which the traditional concept of chance has historically been built.

Conclusion

The evolution of artificial intelligence is inevitably confronting the gambling industry with new challenges, particularly within verticals where prediction and variable analysis are becoming increasingly valuable. In this context, the discussion no longer appears limited to the existence of a random element, but rather to the impact that analytical and computational advantages may have on users interacting with betting operators, and whether both the industry and regulation are prepared for the emergence of a far more sophisticated user seeking to reduce chance and uncertainty.

This transformation may also strengthen some of the arguments advanced by those who consider prediction markets to be more closely aligned with financial dynamics than traditional gambling. If certain users increasingly rely on predictive tools, automation and sophisticated information analysis, it becomes necessary to question whether traditional gambling regulation remains suitable for these new realities or whether alternative regulatory approaches may eventually be required.

Everything suggests that, in the short and medium term, new regulatory demands will emerge concerning the use of artificial intelligence within the gambling industry. Issues such as transparency, the use of AI agents, advanced predictive tools, personalisation and equality of conditions between users are likely to generate discussions that will progressively reshape the way the sector is understood and regulated.

Given the speed of this transformation, both regulators and the industry itself may need to adopt a more proactive approach. Those capable of understanding the implications of these technologies at an early stage will likely shape how markets understand chance and prediction in the years ahead. Otherwise, there is a risk that certain economic dynamics may progressively migrate towards regulatory frameworks more closely associated with financial or technological sectors, moving part of this discussion outside the traditional boundaries of gambling regulation.

What is clear is that artificial intelligence is here to stay, as are prediction markets.

“…what you resist not only persists, but will grow in size.”

― Carl Jung 

[1] NEXT.io, “The Great Divide”, March 2026.

[2] Standford University, “Artificial Intelligence Index Report 2025”.

[3] Kasra Ghaharian et al, “The Need for Benchmarks to Advance AI-Enabled Player Risk Detection in Gambling”, International Gaming Institute (2025).

[4] Ibid.

[5] Boll Luke, “Gridiron Genius: Using Neural Networks to Predict College Football”, Michigan University,

[6] Lasso regression is a statistical technique used to identify relevant variables within large volumes of data, while feature selection refers to the process of identifying and retaining the most useful variables to improve the predictive performance of a machine learning model.

[7] Manassé Galekwa René et al. “A Systematic Review of Machine Learning in Sports Betting: Techniques,

Challenges, and Future Directions” (2024).

[8] Tshilidzi Marwala and Evan Hurwitz “Artificial Intelligence and Asymmetric Information Theory”, University of Johannesburg.

[9] Mihai Florin et al. “AI Personalization and Its Influence on Online Gamblers’ Behavior”. Review Behavioral Sciences”. (2025)

Juan Camilo Carrasco is Managing Partner at SORA Law in Colombia

Related articles on AI and prediction markets

Have a little faith in markets

Artificial Intelligence in Gaming