InfoFi in-depth research report: Attention finance experiment in the AI ​​era

I. Introduction: From information scarcity to attention scarcity, InfoFi was born

The information revolution in the 20th century brought explosive knowledge growth to human society, but it also triggered a paradox: when information acquisition is almost free, what is truly scarce is no longer the information itself, but the cognitive resource we use to process information - attention. As Nobel Prize winner Herbert Simon first proposed the concept of "attention economy" in 1971, "information overload leads to attention poverty", and modern society is deeply trapped in it. Faced with the overwhelming content instilled by Weibo, X, YouTube, short videos, and news push, the cognitive boundaries of human beings are being continuously squeezed, and screening, judgment, and assignment are becoming increasingly difficult.

And this scarcity of attention has evolved into a battle for resources in the digital age. In the traditional Web2 model, the platform firmly controls the traffic entrance through algorithmic distribution, and the real creators of attention resources - whether users, content creators or community evangelists - are often just "free fuel" in the platform's profit logic. The head platforms and capital parties harvest layer by layer in the chain of attention realization, while ordinary individuals who truly promote information production and diffusion find it difficult to participate in value sharing. This structural split is becoming the core contradiction in the evolution of digital civilization. The rise of information financialization (InfoFi) is happening in this context. It is not an occasional new concept, but a bottom-level paradigm shift with blockchain, token incentives and AI empowerment as the technical foundation and the goal of "reshaping the value of attention". InfoFi attempts to transform users' unstructured cognitive behaviors such as opinions, information, reputation, social interaction, trend discovery, etc. into quantifiable and tradable asset forms, and through a distributed incentive mechanism, every user who participates in creation, dissemination, and judgment in the information ecosystem can share the resulting value. This is not just a technological innovation, but also an attempt to redistribute power about "who has attention and who dominates information."

In the narrative genealogy of Web3, InfoFi is an important bridge connecting social networks, content creation, market games and AI intelligence. It inherits the financial mechanism design of DeFi, the social drive of SocialFi and the incentive structure of GameFi, while introducing AI's capabilities in semantic analysis, signal recognition and trend prediction to build a new market structure centered on "financialization of cognitive resources". Its core is not simple content distribution or likes and rewards, but a complete set of value discovery and redistribution logic centered on "information → trust → investment → return".

From the agricultural society with "land" as the scarce element, to the industrial era with "capital" as the growth engine, to today's digital civilization with "attention" as the core means of production, the resource focus of human society is undergoing a profound shift. And InfoFi is the concrete expression of this macro paradigm shift in the on-chain world. It is not only a new outlet for the crypto market, but also the starting point for the deep reconstruction of the digital world's governance structure, intellectual property logic and financial pricing mechanism.

But any paradigm shift is not linear, it is inevitably accompanied by bubbles, hype, misunderstandings and vacillations. Whether InfoFi can become a truly user-centered attention revolution depends on whether it can find a dynamic balance between incentive mechanism design, value capture logic and real needs. Otherwise, it will just be another fantasy of sliding from "inclusive narrative" to "centralized harvesting".

2. InfoFi's ecological composition: a ternary cross market of "information × finance × AI"

The essence of InfoFi is to build a composite market system that simultaneously embeds financial logic, semantic computing and game mechanism in the contemporary network context where information is highly flooded and value is difficult to capture. Its ecological architecture is not a single-dimensional "content platform" or "financial protocol", but the intersection of information value discovery mechanism, behavior incentive system and intelligent distribution engine - forming a full-stack ecosystem integrating information trading, attention incentives, reputation rating and intelligent prediction.

From the underlying logic, InfoFi is an attempt to "financialize" information, that is, to transform cognitive activities such as content, opinions, trend judgments, social interactions, etc. that were originally unpriced into measurable and tradable "quasi-assets" and give them market prices. The intervention of finance makes information no longer scattered and isolated "content fragments" in the process of production, circulation and consumption, but a "cognitive product" with gaming attributes and value accumulation capabilities. This means that a comment, a prediction, and a trend analysis can be an expression of individual cognition, or a speculative asset with risk exposure and future income rights. The popularity of prediction markets such as Polymarket and Kalshi is an example of the implementation of this logic at the level of public opinion and market expectations.

However, financial mechanisms alone are far from enough to solve the problem of noise flooding and bad money driving out good money caused by the explosion of information. Therefore, AI has become the second pillar of InfoFi. AI mainly plays two roles: one is semantic screening, which serves as the "first line of defense" for information signals and noise; the other is behavior recognition, which achieves accurate evaluation of information sources by modeling multi-dimensional data such as user social network behavior, content interaction trajectory, and originality of opinions. Platforms such as Kaito AI, Mirra, and Wallchain are typical representatives of introducing AI technology into content evaluation and user portraits. In the Yap-to-Earn model, they play the role of "algorithmic referee" for incentive distribution, deciding who should receive token rewards and who should be blocked or demoted. In a sense, the function of AI in InfoFi is equivalent to the market maker and clearing mechanism in the exchange, which is the core of maintaining ecological stability and credibility.

And information is the foundation of all this. It is not only the subject of the transaction, but also the source of market sentiment, social connection, and consensus shaping. Unlike DeFi, InfoFi's asset anchors are no longer on-chain hard assets such as USDC and BTC, but "cognitive assets" such as opinions, trust, topics, trends, and insights, which are more liquid, looser in structure, but more timely. This also determines that the operating mechanism of the InfoFi market is not a linear stack, but a dynamic ecology that is highly dependent on social graphs, semantic networks, and psychological expectations. In this framework, content creators are equivalent to the "market makers" of the market. They provide opinions and insights for the market to judge their "prices"; users are "investors" who express their value judgments on certain information through behaviors such as likes, reposts, bets, and comments, driving it to rise or sink in the entire network; and platforms and AI are "referees + exchanges" responsible for ensuring the fairness and efficiency of the entire market.

The coordinated operation of this ternary structure has spawned a series of new species and new mechanisms: prediction markets provide clear targets for gaming; Yap-to-Earn encourages knowledge mining and interaction output; reputation protocols such as Ethos transform personal on-chain history and social behavior into credit assets; attention markets such as Noise and Trends attempt to capture the "emotional fluctuations" of on-chain communication; and token-gated content platforms such as Backroom rebuild the information payment logic through the permission economy. Together, they constitute the multi-layered ecology of InfoFi: it contains both value discovery tools and value distribution mechanisms, and also embeds a multi-dimensional identity system, participation threshold design, and anti-witch mechanisms.

It is in this cross-structure that InfoFi is no longer just a market, but a complex information game system: it uses information as a transaction medium, finance as an incentive engine, and AI as a governance center, and ultimately intends to build a self-organizing, distributable, and adjustable cognitive collaboration platform. In a sense, it attempts to become a "cognitive financial infrastructure", not only for content distribution, but also to provide a more efficient information discovery and collective decision-making mechanism for the entire crypto society.

However, such a system is also destined to be complex, diverse and fragile. The subjectivity of information determines the non-uniformity of value assessment, the game nature of finance increases the risk of manipulation and herd effect, and the black box nature of AI also challenges transparency. The InfoFi ecosystem must constantly balance and self-repair between the three tensions, otherwise it is very easy to slide into the opposite side of "disguised gambling" or "attention harvesting field" driven by capital.

The ecological construction of InfoFi is not an isolated project of a certain protocol or platform, but a co-performance of a whole set of social-technical systems. It is a deep attempt of Web3 in the direction of "governing information" rather than "governing assets". It will define the information pricing method of the next era, and even build a more open and autonomous cognitive market.

III. Core Game Mechanism: Incentive Innovation vs. Harvesting Trap

In the InfoFi ecosystem, behind all the prosperous appearances, it is ultimately the design game of incentive mechanisms. Whether it is the participation in the prediction market, the output of verbal behavior, the construction of reputation assets, the transaction of attention, or the mining of on-chain data, it is essentially inseparable from a core question: Who works? Who gets dividends? Who bears the risk?

From an external perspective, InfoFi seems to be a "production relationship innovation" for the migration from Web2 to Web3: it attempts to break the exploitation chain between "platform-creator-user" in traditional content platforms and return value to the original contributors of information. However, from the internal structure, this value return is not inherently fair, but a delicate balance based on a series of incentives, verification and game mechanisms. If designed properly, InfoFi is expected to become an innovative experimental field for win-win users; if the mechanism is unbalanced, it will easily become a "retail investor harvesting field" dominated by capital + algorithms.

The first thing to examine is the positive potential of "incentive innovation". The essential innovation of all sub-tracks of InfoFi is to give "information", an intangible asset that was difficult to measure and financialized in the past, clear tradability, competitiveness and settlement. This transformation relies on two key engines: the traceability of blockchain and the evaluability of AI.

The prediction market monetizes cognitive consensus through the market pricing mechanism; the mouth-pushing ecosystem turns speech into economic behavior; the reputation system is building a kind of inheritable and mortgageable social capital; the attention market takes hot trends as the transaction target, and redefines the content value through the logic of "information discovery -> betting on signals -> obtaining price differences"; and the AI-driven InfoFi application attempts to build an information financial network driven by data and algorithms through large-scale semantic modeling, signal recognition, and on-chain interaction analysis. These mechanisms make information have the attribute of "cash flow" for the first time, and also make "saying a word, forwarding a tweet, endorsing someone" become real production activities.

However, the stronger the incentive system, the more likely it is to give rise to "game abuse". The biggest systemic risk faced by InfoFi is the alienation of incentive mechanisms and the reproduction of arbitrage chains.

Take Yap-to-Earn as an example. On the surface, it rewards the value of user content creation through AI algorithms, but in actual implementation, many projects quickly fall into "information haze" after briefly attracting a large number of content creators in the early stage of incentives - robot matrix account flooding, big V internal testing early participation, project parties directional manipulation of interaction weights and other chaos frequently occur. A top KOL said bluntly: "Now you can't get on the list without brushing the volume. AI has been trained to identify keywords and take advantage of the popularity." A project owner also revealed: "I invested 150,000 US dollars in a round of Kaito's mouth-pushing, but 70% of the traffic was AI and water army. The real KOLs didn't participate. It's impossible for me to invest again." Under the opaque mechanism of the points system and token expectations, many users have become "free workers": posting tweets, interacting, going online, and building groups, but in the end they are not eligible to participate in airdrops. This kind of "backstab" incentive design not only damages the reputation of the platform, but also easily leads to the collapse of the long-term content ecology. The comparison case of Magic Newton and Humanity is particularly typical: the former has a clear distribution mechanism in the Kaito mouth-pushing stage, and the token value return is rich; the latter has an unbalanced distribution mechanism and lack of transparency, which has caused a crisis of trust in the community and doubts about "anti-pushing". This structural injustice under the Matthew effect has greatly reduced the enthusiasm of the tail creators and ordinary users to participate, and even spawned the ironic identity of "algorithm-sacrificing mouth-pushing players".

What is more noteworthy is that the financialization of information does not mean the consensus of value. In the attention market or reputation market, those contents, characters or trends that are "long" may not be signals of real long-term value. In the absence of real demand and scenario support, once the incentives ebb and subsidies stop, these financialized "information assets" often quickly return to zero, and even form a Ponzi dynamic of "short-term speculation narrative, long-term return to zero". The short life of the LOUD project is a microcosm of this logic: the market value exceeded 30 million US dollars on the day of its launch, and fell to less than 600,000 just two weeks later, which can be called the InfoFi version of "pass the parcel".

In addition, in the prediction market, if the oracle mechanism is not transparent enough or is manipulated by large capital holders, it is very easy to form information pricing deviations. Polymarket has repeatedly caused user disputes due to "unclear explanation of event settlement", and in 2025, there was even a large-scale compensation storm caused by a voting vulnerability in the oracle. This reminds us that even if it is a prediction mechanism based on "real-world information", it must find a better balance between technology and game.

Finally, whether InfoFi's incentive mechanism can break out of the confrontational narrative of "financial capital vs. retail attention" depends on whether it can build a triple positive feedback system: information production behavior can be accurately identified -> value distribution mechanism can be transparently executed -> long-tail participants can be truly motivated. This is not only a technical issue, but also a test of institutional engineering and product philosophy.

In summary, InfoFi's incentive mechanism is both its greatest advantage and its greatest source of risk. In this market, every design of incentives may create an information revolution or trigger a collapse of trust. Only when the incentive system is no longer just a game of traffic and airdrops, but becomes an infrastructure that can identify real signals, incentivize high-quality contributions, and form a self-consistent ecosystem, can InfoFi truly achieve the transition from "gimmick economy" to "cognitive finance".

Fourth, typical project analysis and recommended focus

InfoFi's ecosystem currently presents a pattern of flourishing and hot spots rotating. Different projects have evolved differentiated product paradigms and user growth strategies around the core path of "information → incentives → market". Some projects have initially verified the business model and become the key anchor of InfoFi's narrative; while others are in the proof-of-concept stage and are still looking for breakthroughs in the process of user education and mechanism optimization. In the complex track, we try to select projects from five representative directions for analysis and propose potential camps that are worth continuing to track.

1. Predicting market direction: Polymarket + Upside

Polymarket is one of the most mature and iconic projects in the InfoFi ecosystem. Its core model is to achieve collective expected pricing of real events by buying and selling contract shares with different outcomes through USDC. The reason why it is called "the prototype of information finance" by Vitalik is not only because its trading logic is clear enough and its financial design is robust enough, but also because it has begun to have "media functions" in the real world - for example, during the 2024 US election, the probability of winning or losing reflected by Polymarket was better than traditional polls many times, which triggered heated discussions and reposts including Musk.

With the official cooperation between Polymarket and X, its user growth and data visibility have been further enhanced, and it is expected to become a "super hub platform" for the integration of social public opinion and information pricing. However, the challenges currently faced by Polymarket still include compliance risks (CFTC has repeatedly launched attacks), oracle disputes, and insufficient participation in niche topics.

In contrast, Upside focuses on social prediction and is an emerging project invested by well-known capital such as Arthur Hayes. It attempts to marketize content prediction through the mechanism of likes and voting, so that creators, readers, and voters can share the benefits. Upside emphasizes light interaction, low threshold, and de-financialized user experience, and explores the integration model between InfoFi and content platforms. It is worth paying attention to its subsequent performance in user retention and content quality maintenance.

2. Yap-to-Earn direction: Kaito AI + LOUD

Kaito AI is one of the most representative platforms in the Yap-to-Earn model, and is also the project with the largest number of InfoFi users. It has attracted more than 1 million registered users and more than 200,000 active Yappers. Its innovation lies in the use of AI algorithms to evaluate the quality, interactivity, and project relevance of content posted by users on X (formerly Twitter), thereby distributing Yaps (points), and cooperating with projects based on the rankings to airdrop or reward tokens.

The Kaito model forms a closed loop: projects use tokens to incentivize community dissemination, creators use content to compete for attention, and the platform uses data and AI models to control distribution and order. However, with the surge in users, it has also encountered structural problems such as content signal pollution, robot proliferation, and point distribution disputes. The founder of Kaito has recently begun to iterate algorithms and optimize community mechanisms for these problems.

And LOUD is the first project to use the Yap-to-Earn points list for IAO (initial attention issuance). Before going online, it monopolized 70% of the attention on the Kaito list through word-of-mouth activities. Although its airdrop strategy created a lot of social voice in the short term, it was criticized by the community as "harvesting by passing the flower" due to the rapid plunge in the subsequent token price. The ups and downs of LOUD show that the Yap-to-Earn track is still in the trial and error stage, and the maturity of the mechanism and the fairness of incentives still need to be polished.

3. Reputation Finance: Ethos + GiveRep

Ethos is the most systematic and decentralized attempt in the current reputation finance track. Its core logic is to build a verifiable "credit score" on the chain. It not only generates scores through interactive records and comment mechanisms, but also introduces a "guarantee mechanism": users can pledge ETH to endorse others and take certain risks, thus forming a Web3-like trust network.

Another major innovation of Ethos is the launch of a reputation speculation market, allowing users to "go long or short" the reputation of others, forming a new dimension of financial instruments-trust monetization. This mechanism opens up imagination space for the integration of reputation scoring with the lending market, DAO governance, and social identity recognition in the future. However, its invitation-based mechanism also slows down the expansion of users. How to lower the threshold and improve anti-witch capabilities in the future is the key to the development of the platform.

Compared with Ethos, GiveRep is more lightweight and community-oriented. Its mechanism is to score content creators and commenters by commenting @ official accounts, with a limited number of comments per day. In conjunction with the active ecology of the X community, it has achieved a certain scale of dissemination on Sui. This model is more suitable for projects to do lightweight testing of social fission and reputation scoring, and can also serve as a trust basis for future integration of governance weights, project airdrops and other mechanisms.

4. Attention market direction: Trends + Noise + Backroom

Trends is a platform that explores "content assetization", allowing creators to cast their X posts into tradable "Trends", set up trading curves, and community members can buy and go long on the popularity of the post, while creators receive commissions from the transaction. It creatively transforms "explosive posts" into liquid assets, which is a typical attempt at "social financialization".

Noise is an attention futures platform based on MegaETH. Users can bet on the popularity changes of a topic or project. It is a direct investment platform for attention finance. In the closed beta test that requires an invitation code, some of its prediction models have demonstrated early market discovery capabilities. If AI models are introduced to predict popularity trends in the future, it may become a "weather vane" tool in the InfoFi ecosystem.

Backroom represents an InfoFi product of "paid unlocking + screening of high-value content". Creators can publish high-quality content based on token thresholds, and users can unlock access by purchasing Keys. At the same time, Keys themselves are tradable and have value volatility, forming a closed loop of content finance. Against the backdrop of the popularity of NoiseFi, this model focuses on "reducing noise and screening signals" and is becoming a new tool for knowledge creators.

5. Data Insight and AI Agent Platform: Arkham + Xeet + Virtuals

Arkham Intel Exchange has become synonymous with the financialization of on-chain intelligence, allowing users to issue bounties and incentivize "on-chain detectives" to disclose address ownership information. Its logic is similar to that of the traditional intelligence market, but it is decentralized and tradable for the first time. Although there are constant controversies (such as privacy violations and witch hunts), it has established the basic paradigm of data insight InfoFi.

Although Xeet has not yet been fully released, its founder Pons has publicly stated that he wants to be the "noise reducer" of InfoFi. By introducing mechanisms such as the Ethos reputation system, KOL recommendations, and private content recommendations, he has created a more authentic and spam-free signal market, which is a direct counterattack against the noise problem of Yap-to-Earn.

Virtuals' innovation lies in using AI agents as new InfoFi participants, injecting "non-human productivity" into the InfoFi ecosystem by launching tasks, completing evaluations, and generating interactive data. The Yap-to-Earn stage in its Genesis Launch mode is linked with Kaito, which also shows the trend of ecological linkage between InfoFi projects.

V. Future Trends and Risk Outlook: Can Attention Become the "New Gold"?

In the deep waters of the digital economy, information is no longer scarce, but effective information and credible attention are becoming more and more precious. In this context, InfoFi is called the "next narrative engine" and even a potential asset of "new gold" by many industry insiders. The logic behind it is: in today's world where AI computing power is becoming increasingly rampant and content costs are approaching zero, what is scarce is not content, but the "signal" that can accurately guide actions, and the real attention itself that focuses on this signal. Whether the future InfoFi can move from concept to assetization, and from short-term "mouth incentive" to long-term "on-chain influence standards", the key lies in the struggle and evolution of three major trends and three major risks.

First, the deep integration of AI and prediction market will open a new era of "reasoning capital". At present, the combination of Polymarket, X and Grok has taken the lead in implementing this model: real-time public opinion + AI analysis + real money game results, building a flywheel between effectiveness, authenticity and market feedback. If the future InfoFi project can use AI to provide event modeling, signal extraction and dynamic pricing, it will greatly enhance the credibility of the prediction market in governance, news verification, trading strategies and other aspects. For example, the governance DAO under the Futarchy model may use a combination of AI + prediction market to formulate policies in the future.

Second, the intersection of reputation, attention and financial attributes will trigger a major outbreak of decentralized credit system. The current exploration of reputation InfoFi projects (such as Ethos and GiveRep) is building a set of on-chain "reputation points" that do not require third-party credit intermediaries. In the future, reputation points are expected to become the basis for DAO voting rights, DeFi collateral, content distribution priorities, etc., and become a real on-chain "social capital". If cross-platform mutual recognition, anti-sybil attack and traceable credit trajectory can be achieved, the attention reputation system will rise from an auxiliary indicator to a core asset.

Third, the tokenization and derivativeization of attention assets are the ultimate form of InfoFi. The current Yap-to-Earn model is still at the stage of exchanging content and influence for points, while a truly mature InfoFi should be able to convert every valuable content, a KOL's "attention bond", and a set of on-chain signals into tradable assets, and allow users to "go long", "go short" and even "construct ETFs". This will also form a brand new financial market: from narrative-based Meme Tokens to derivative assets based on attention dynamics.

But at the same time, InfoFi still faces three major structural risks if it wants to be truly sustainable.

The first is that the imperfect mechanism design leads to the proliferation of "mouth-pushing traps". If the incentive is too biased towards "quantity rather than quality", the platform algorithm is not transparent, and the airdrop expectations are unreasonable, it will lead to extremely high enthusiasm in the early stage of the project, but the attention will drop sharply in the later stage, forming the SocialFi-style fate of "airdrop is the peak". For example, LOUD initially attracted users with Yap ranking incentives, but after the token was launched, the market value plunged and the participation rate dropped sharply, reflecting the lack of long-term mechanisms in the ecosystem.

The second is that the "Matthew effect" intensifies and causes ecological fragmentation. The data of most current mouth-pushing platforms have revealed that more than 90% of the rewards are concentrated in the hands of the top 1% of users, and long-tail users can neither benefit from the interaction nor break through the KOL class, and eventually choose to exit. If this structure cannot be broken through reputation weighting, credit flow and other mechanisms, it will weaken the willingness of users to participate and turn InfoFi into another "platform oligopoly" system.

The third is the dual dilemma of regulatory risk and information manipulation. For emerging products such as prediction markets, reputation trading, and attention speculation, major jurisdictions around the world have not yet formed a unified regulatory framework. Once a platform involves gambling, insider trading, false propaganda or market manipulation, it is very easy to trigger high regulatory pressure. For example, Polymarket has encountered dual scrutiny from the CFTC and FBI in the United States, and Kalshi has also taken a differentiated path due to its compliance advantages. All this means that the InfoFi project must consider the "regulatory-friendly" path from Day One design to avoid going to the edge of illegality.

In summary, InfoFi is not just a next-generation content distribution protocol, but also a new attempt to financialize attention, information, and influence. It is a challenge to the traditional platform value possession model and a collective experiment of "everyone is the discoverer of Alpha." Whether InfoFi can become the "new gold" in the Web3 world in the future depends on whether it can find the optimal solution among the fair mechanism, incentive design and regulatory framework, and truly turn the "attention dividend" from the prey of a few people into the assets of the majority.

Sixth, Conclusion: The revolution is not yet complete, InfoFi still needs to be cautiously optimistic

The emergence of InfoFi is another cognitive evolution of the Web3 world after going through multiple cycles such as DeFi, NFT, and GameFi. It attempts to answer a core question that has been ignored for a long time-in an era of information overload, free content, and algorithm proliferation, what is really scarce? The answer is: human attention, real signals, and credible subjective judgment. This is exactly what InfoFi tries to give value, incentive mechanisms, and market structures.

In a sense, InfoFi is a "reverse power revolution" against the traditional attention economy system - instead of allowing platforms, giants, and advertisers to monopolize data and traffic dividends, it attempts to redistribute the value of attention to real creators, communicators, and identifiers through blockchain, tokenization, and AI protocols. This structural redistribution of value gives InfoFi the potential to change the content industry, platform governance, knowledge collaboration, and even social public opinion mechanisms.

However, potential does not mean reality. We still need to be cautiously optimistic.

The revolution is not yet complete, but it has begun. The future of InfoFi is not defined by a certain platform or track, but is jointly shaped by all creators, observers, and identifiers of attention. If DeFi is a revolution about the flow of value, then InfoFi is a revolution about the way value is perceived and distributed. On the long-term path of de-platforming and de-intermediation, we should maintain calm judgment and prudent participation, but we should not ignore its potential as the soil for the next generation of Web3 to grow a new narrative forest.