The Coasean Singularity? Demand, Supply, and Market Design with AI Agents

May combine two claims.

An agent can solicit and compare many quotes in parallel, then book and monitor follow-through at far lower marginal cost than a human.

“An agent can solicit and compare many quotes in parallel, then book and monitor follow-through at far lower marginal cost than a human.”

Probably not a substantial claim: may not pass S4 (based).

By collapsing the costs of eliciting preferences, enforcing commitments, and verifying identity, agents make mechanisms that were once only theoretically attractive now feasible at scale.

“By collapsing the costs of eliciting preferences, enforcing commitments, and verifying identity, agents make mechanisms that were once only theoretically attractive now feasible at scale.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

The broad adoption of AI agents will cause transformative downstream effects on the economy.

“The broad adoption of AI agents will cause transformative downstream effects on the economy.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

Determining the boundaries of autonomous action and deference is essential for preserving trust and ensuring that delegation improves welfare rather than creating new risks of overreach.

“Determining these boundaries is essential for preserving trust, ensuring that delegation improves welfare rather than creating new risks of overreach.”

Probably not a substantial claim: may not pass S4 (based).

Unlike human agents, AI agents do not derive utility from pecuniary compensation.

“Unlike human agents, they also do not derive utility from pecuniary compensation.”

• AssumptionRests on an assumption: AI agents do not derive utility from pecuniary compensation.

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Firms will design, integrate, and monetize agents, with outcomes hinging on whether agents operate within or across platforms.

“On the supply side, firms will design, integrate, and monetize agents, with outcomes hinging on whether agents operate within or across platforms.”

Classified as probably not a substantial claim. May combine two claims.

The core design problem is two-fold: eliciting the principal's preferences and ensuring the agent honors them (the alignment problem).

“The core design problem is thus two-fold: both eliciting those preferences and ensuring the agent honors them-together these are what computer scientists and philosophers refer to as the alignment problem (Bostrom 2014, Christian 2020).”

• cites Bostrom 2014Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Christian 2020Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S4 (based).

Agent adoption reflects derived demand: users trade off decision quality against effort reduction, with outcomes mediated by agent capability and task context.

“From the demand side, agent adoption reflects derived demand: users trade off decision quality against effort reduction, with outcomes mediated by agent capability and task context.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Agents must learn not only how to act, but also when to act autonomously and when to defer to their principals.

“Beyond preference elicitation, a central design challenge concerns meta-rationality: agents must learn not only how to act, but also when to act autonomously and when to defer to their principals.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Compared to humans, agents can persist through repeated failures at much lower marginal cost and continue monitoring tasks over longer windows to secure better outcomes.

“Compared to humans, agents can persist through repeated failures at much lower marginal cost and continue monitoring tasks over longer windows to secure better outcomes.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The use of AI agents will be driven by their design, which has both an engineering and an economic component.

“The use of AI agents will be driven by their design, which has both an engineering and an economic component.”

May combine two claims.

AI agents reduce transaction costs through direct task execution and through advisory services.

“This reduction occurs not only through direct task execution but also via advisory services.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

Over time, the progression of augmentation and substitution will lead to reorganization, as tasks are removed or redefined and workflows are increasingly structured around agent capabilities.

“Over time, this progression will lead to reorganization, as tasks are removed or redefined and workflows are increasingly structured around agent capabilities.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Differentiation will likely emerge, with some agents becoming known as particularly effective for specific applications.

“Differentiation will likely emerge, with some agents becoming known as particularly effective for specific applications.”

May combine two claims.

Agents create efficiency gains from lower search, communication, and contracting costs, but also introduce frictions such as congestion and price obfuscation.

“At the market level, agents create efficiency gains from lower search, communication, and contracting costs, but also introduce frictions such as congestion and price obfuscation.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

In markets that adapt, agents will initially augment humans by assisting with specific tasks, then, as their capabilities grow, will begin to substitute for tasks entirely, shifting human effort toward judgment, oversight, and relationship-focused activities.

“In markets that adapt, agents will initially augment humans by assisting with specific tasks, then, as their capabilities grow, will begin to substitute for tasks entirely, shifting human effort toward judgment, oversight, and relationship-focused activities.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

As adoption spreads from initial footholds, the dynamics of trust, evaluation, and inter-agent coordination will shape broader economic integration.

“As adoption spreads from these initial footholds, the dynamics of trust, evaluation, and inter-agent coordination will shape broader economic integration.”

May combine two claims.

By lowering the costs of preference elicitation, contract enforcement, and identity verification, agents expand the feasible set of market designs but also raise novel regulatory challenges.

“By lowering the costs of preference elicitation, contract enforcement, and identity verification, agents expand the feasible set of market designs but also raise novel regulatory challenges.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

Transformations in market structure will come from existing markets adapting to agent capabilities and from the emergence of entirely new, agent-first designs.

“Transformations in market structure will not only come from existing markets adapting to agent capabilities but also from the emergence of entirely new, agent-first designs.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Demand for reliable information about agent performance will be substantial, with benchmarks playing important roles alongside word-of-mouth recommendations.

“Demand for reliable information about agent performance will be substantial, with benchmarks playing important roles alongside word-of-mouth recommendations, even as consumers rightly worry about benchmark gaming.”

May combine two claims.

Control over foundation models and the terms of access directly shapes which firms can supply agents, how they differentiate, and how competition unfolds.

“Control over these models and the terms of access directly shapes which firms can supply agents, how they differentiate, and how competition unfolds.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

Agents need to be both rational and robust.

“Closely related is the need for agents to be both rational and robust.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

Novel, agent-first markets will be designed from the ground up with processes, interactions, and roles structured entirely around agent capabilities.

“Rather than evolving through incremental augmentation and substitution, they will be designed from the ground up with processes, interactions, and roles structured entirely around agent capabilities.”

Probably not a substantial claim: may not pass S4 (based).

Even if performance improves with compute, the returns to agent quality are unlikely to increase indefinitely.

“Even if performance improves with compute, the returns to agent quality are unlikely to increase indefinitely.”

Probably not a substantial claim: may not pass S4 (based).

Humans will deploy AI agents to make decisions of similar or potentially even lower quality, but at dramatically reduced cost and effort.

“second, to make decisions of similar or potentially even lower quality, but at dramatically reduced cost and effort”

May combine two claims.

Even when skilled execution remains human, the intermediation premium falls, creating demand for AI agents.

“Even when skilled execution remains human, the intermediation premium falls, creating demand for AI agents.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Ensuring that agents consistently deliver on capability, knowledge, and alignment poses substantial theoretical and practical challenges.

“Yet ensuring that agents consistently deliver on these characteristics poses substantial theoretical and practical challenges-ones that future research will need to address.”

May combine two claims.

AI agents operate in natural language and other open-ended mediums rather than within fixed input structures, making them far more flexible: any statement or request can, in principle, serve as input.

“AI agents, by contrast, operate in natural language and other open-ended mediums rather than within fixed input structures.This makes them far more flexible: any statement or request can, in principle, serve as input.”

May combine two claims.

In the scenario of diminishing returns, the agent may be offered for free and supported by advertising, or bundled with complementary goods and services.

“In this scenario, the agent may be offered for free and supported by advertising, or bundled with complementary goods (e.g., phones) and services (e.g., delivery).”

Probably not a substantial claim: may not pass S4 (based).

The activities that comprise transaction costs-learning prices, negotiating terms, writing contracts, and monitoring compliance-are precisely the types of tasks that AI agents can potentially perform at very low marginal cost.

“The activities that comprise transaction costs-learning prices, negotiating terms, writing contracts, and monitoring compliance-are precisely the types of tasks that AI agents can potentially perform at very low marginal cost.”

• AssumptionThe authors state this as an assumption: AI agents will have the ability to harness computational resources, communicate with other agents and humans, receive and send money, and access and interact with the Internet.

May combine two claims.

AI agents are not guaranteed to make markets more efficient or to improve other social objectives.

“That said, AI agents are not guaranteed to make markets more efficient or to improve other social objectives.”

Probably not a substantial claim: may not pass S4 (based).

AI agents are poised to transform digital markets by dramatically reducing transaction costs.

“AI agents-autonomous systems that perceive, reason, and act on behalf of human principals-are poised to transform digital markets by dramatically reducing transaction costs.”

Classified as probably not a substantial claim.

Consumers will want capable, knowledgeable, and faithful agents.

“In essence, they will want capable, knowledgeable, and faithful agents.”

Probably not a substantial claim: may not pass S4 (based).

Humans will deploy AI agents to optimize decisions that they would otherwise make sub-optimally due to information constraints or cognitive limitations.

“first, to optimize decisions that they would otherwise make sub-optimally due to information constraints or cognitive limitations”

May combine two claims.

AI agents face a very different cost and supply profile from human agents, which changes the pricing dynamics.

“AI agents, however, face a very different cost and supply profile, which changes the pricing dynamics.”

May combine two claims.

Key characteristics of markets where AI agents first gain traction include high-stakes transactions, large pools of potential counterparties, substantial effort required to evaluate options, information asymmetries that could be resolved through effort or experience, and experience asymmetries due to differences in transaction frequency.

“Key characteristics of these markets include high-stakes transactions, large pools of potential counterparties, substantial effort required to evaluate options, information asymmetries that could be resolved through effort or experience, and experience asymmetries due to differences in transaction frequency.”

Classified as probably not a substantial claim.

An AI agent must know the principal's preferences to act on their behalf.

“Clearly, an AI agent must know the principal's preferences to act on their behalf.”

Probably not a substantial claim: may not pass S4 (based).

Equilibrium suboptimality is particularly relevant in environments with externalities across agents, imperfect alignment, and asymmetric information.

“This is particularly relevant in environments with externalities across agents, imperfect alignment, and asymmetric information.”

Probably not a substantial claim: may not pass S4 (based).

As marginal gains diminish, the market dynamics begin to resemble those of other digital services such as search and social media.

“As marginal gains diminish, the market dynamics begin to resemble those of other digital services such as search and social media.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

This new set of feasible designs has the potential to improve consumer welfare and matching.

“This new set of feasible designs has the potential to improve consumer welfare and matching.”

May combine two claims.

AI agents will appear first in markets that already employ human agents or where large digital platforms were created to overcome matching frictions-such as LinkedIn, Upwork, Zillow, and Airbnb.

“A useful heuristic suggests AI agents will appear first in markets that already employ human agents or where large digital platforms were created to overcome matching frictions-such as LinkedIn, Upwork, Zillow, and Airbnb.”

Classified as probably not a substantial claim.

If agent quality depends on compute, prices could scale with the amount of compute required-potentially rising in proportion to the stakes of the transaction.

“In that case, prices could scale with the amount of compute required-potentially rising in proportion to the stakes of the transaction.”

Classified as probably not a substantial claim. May combine two claims.

The economics of building one's own foundation models versus using others' differs sharply, since training foundation models incurs high fixed costs, while operating agents incurs mostly variable costs.

“The economics of these differs sharply, since training foundation models incurs high fixed costs, while operating agents incurs mostly variable costs.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Combined identity, credentials, and reputation can enable more sophisticated market interactions, including targeted deals and negotiations over small purchases that would previously have been economically infeasible.

“Combined, identity, credentials, and reputation can enable more sophisticated market interactions, including targeted deals and negotiations over small purchases that would previously have been economically infeasible.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

These forces will decide whether agents become a foundation for welfare-enhancing prosperity or collapse into merely rent-redistributive outcomes.

“In the end, these forces will decide whether agents become a foundation for welfare-enhancing prosperity or collapse into merely rentredistributive outcomes.”

May combine two claims.

In e-commerce environments where agents enable users to consider all available options with complete information access, sophisticated agents would likely prove resistant to nudges and advertising that lacks informational or signaling value, fundamentally altering competitive dynamics.

“Consider a prototypical e-commerce market where agents enable users to consider all available options with complete information access.In such environments, sophisticated agents would likely prove resistant to nudges and advertising that lacks informational or signaling value, fundamentally altering competitive dynamics.”

May combine two claims.

Agents will be able to create and like social media posts, create artistic works, send direct messages, offer to buy items, and apply to jobs.

“Agents will be able to create and like social media posts, create artistic works, send direct messages, offer to buy items, and apply to jobs.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.
• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

As of September 2025, antitrust scrutiny of large AI firms has intensified, including investigations into Google's market practices.

“As of September 2025, antitrust scrutiny of large AI firms has intensified, including investigations into Google's market practices (U.S.Department of Justice 2025).”

• cites U.S.Department of Justice 2025Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Reducing the cost of applying to a job with a customized cover letter by deploying AI may flood employers with applications, imposing higher screening costs and making it harder to pick the right candidate due to congestion.

“For example, reducing the cost of applying to a job with a customized cover letter by deploying AI may flood employers with applications, imposing higher screening costs and making it harder to pick the right candidate due to congestion (Wiles and Horton 2025;Kessler 2025).”

• cites Wiles and Horton 2025Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Kessler 2025Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The bring-your-own-agent model reduces platform lock-in effects.

“Furthermore, the bring-your-own-agent model reduces platform lock-in effects.”

Probably not a substantial claim: may not pass S2 (checkable).

Taken together, AI agents enable new market designs that have previously unimaginable benefits.

“Taken together, AI agents enable new market designs that have previously unimaginable benefits.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.
• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Classified as probably not a substantial claim.

The two axes of ownership and specialization together yield four archetypes of AI agent supply.

“These two axes together yield four archetypes of AI agent supply, summarized in Table 2.”

Probably not a substantial claim: may not pass S2 (checkable).

Such market failures highlight the critical role that market design will play in ensuring positive-sum outcomes as AI agents become ubiquitous in economic transactions.

“Such market failures highlight the critical role that market design will play in ensuring positive-sum outcomes as AI agents become ubiquitous in economic transactions.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

More capable agents can emulate human browsing or operate on the principal's device to conduct certain actions.

“Yet more capable agents can emulate human browsing or operate on the principal's device to conduct certain actions.”

• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Overly restrictive privacy regulation may prevent agents from using private data for the purposes of better alignment.

“Alternatively, overly restrictive privacy regulation may prevent agents from using private data for the purposes of better alignment.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

A bowling-shoe agent is provided by the platform hosting the transaction; it enjoys deep integration, privileged signals, and low setup costs, but is less portable, may steer outcomes toward the platform's interests, and can contribute to platform lock-in.

“By contrast, a bowling-shoe agent is provided by the platform hosting the transaction.These agents enjoy deep integration, privileged signals, and low setup costs, but are less portable, may steer outcomes toward the platform's interests, and can contribute to platform lock-in.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Jailbreaking attacks allow users to circumvent safety measures and access system features in unintended ways, representing a persistent challenge for current chatbot technology.

“Jailbreaking attacks, for example, allow users to circumvent safety measures and access system features in unintended ways, representing a persistent challenge for current chatbot technology (Shen et al. 2024).”

• cites Shen et al. 2024 sourceNot checked yet

  Checked before — its earlier verdict is held back until a new check runs.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Information from usage traces like late-night browsing or vitamin purchases can be repurposed in other contexts, e.g., lower salary offers based on perceived pregnancy status or higher insurance premiums based on inferred health risks.

“This information can be repurposed in other contexts (e.g., lower salary offers based on perceived pregnancy status or higher insurance premiums based on inferred health risks).”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S4 (based).

Rational AI agents browsing on behalf of users are unlikely to be influenced by choice-architecture nudges.

“Rational AI agents browsing on behalf of users are unlikely to be influenced by these nudges.”

• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Probably not a substantial claim: may not pass S4 (based).

Most Internet activity will eventually originate from AI agents rather than humans.

“A fundamental challenge of implementing new market designs emerges from the reality that most Internet activity will eventually originate from AI agents rather than humans.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S4 (based).

Consumers will experience different types of agents along two key dimensions: ownership and specialization.

“In particular, we envision consumers experiencing different types of agents along two key dimensions: ownership, which concerns who has custody of the agent in a market transaction, and specialization, which reflects the breadth or narrowness of the agent's capabilities in a specific context.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Consumers' agents may strategically withhold certain information to secure more favorable prices, adding additional complexity to the market.

“At the same time, consumers' agents may strategically withhold certain information to secure more favorable prices, adding additional complexity to the market.”

Probably not a substantial claim: may not pass S4 (based).

AI agents will be able to effectively mimic humans in many situations.

“These agents will be able to effectively mimic humans in many situations.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Many economists initially predicted that the Internet would eliminate price dispersion, but this prediction largely failed to materialize, perhaps because human search costs suffer from Baumol's cost disease.

“Many economists initially predicted that the Internet would eliminate price dispersion in markets, reasoning that decreased search costs would create conditions similar to Bertrand competition.However, this prediction largely failed to materialize, perhaps because human search costs suffer from Baumol's cost disease-as productivity increases, so does the opportunity cost of time.”

May combine two claims.

These mechanisms highlight substantial efficiency gains but also introduce new risks and distributional uncertainties, leaving overall welfare effects ambiguous in some domains.

“These mechanisms highlight substantial efficiency gains but also introduce new risks and distributional uncertainties, leaving overall welfare effects ambiguous in some domains.”

Classified as probably not a substantial claim. May combine two claims.

The bring-your-own-agent model reduces computational and hosting expenses for platforms while minimizing liability and simplifying API maintenance, but it sacrifices usage insights, opportunities for optimizing the user experience, and potentially profitable opportunities to steer consumers to the platform's preferred options.

“The bring-your-own-agent model reduces computational and hosting expenses for platforms while minimizing liability and simplifying API maintenance, but it sacrifices usage insights, opportunities for optimizing the user experience, and potentially profitable opportunities to steer consumers to the platform's preferred options.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Both sides can precommit their agents to pose all legally permissible questions without penalty, with agents disclosing only relevant responses.

“Both sides can precommit their agents to pose all legally permissible questions without penalty, with agents disclosing only relevant responses.”

• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Training and adaptation on sensitive user data, often without explicit consent, heightens privacy risks and raises questions about data ownership and control, especially in employment, housing, and other high-stakes domains.

“Training and adaptation on sensitive user data, often without explicit consent, heightens these risks and raises questions about data ownership and control, especially in employment, housing, and other highstakes domains.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

A model trained based on Bob's social media posts may memorize information about Bob, and when Amy writes a prompt relating to Bob, information about Bob may surface in unpredictable and potentially damaging ways.

“For example, a model trained based on Bob's social media posts may memorize information about Bob.When Amy writes a prompt relating to Bob, information about Bob may surface in unpredictable and potentially damaging ways.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

A bring-your-own agent carries the user's instructions and data across sites, and neither platform sees or dictates its programming beyond what it explicitly shares.

“This is a bring-your-own agent: it carries the user's instructions and data across sites, and neither platform sees or dictates its programming beyond what it explicitly shares.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Vertical agents specialize in a narrow domain or even a specific platform workflow, trading breadth for depth at the cost of portability and reuse.

“Vertical agents, on the other hand, specialize in a narrow domain (e.g., tax filing, job search, travel) or even a specific platform workflow, trading breadth for depth (i.e., stronger performance, richer integrations, and tailored guardrails) at the cost of portability and reuse.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Existing data regulations will need to be adapted in response to generative AI and automated decision-making.

“Due to these concerns, we expect that existing data regulations (California Consumer Privacy Act 2018; General Data Protection Regulation 2018) will need to be adapted in response to generative AI and automated decision-making.”

• cites California Consumer Privacy Act 2018; General Data Protection Regulation 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Over time, technical evolution may also sharpen the distinction between agent-oriented and person-oriented interfaces, with websites offering parallel access: streamlined, data-rich endpoints optimized for agents alongside traditional interfaces for humans.

“Over time, technical evolution may also sharpen the distinction between agent-oriented and person-oriented interfaces, with websites offering parallel access: streamlined, data-rich endpoints optimized for agents alongside traditional interfaces for humans.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.
• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

May combine two claims.

Even if the principal prefers timely resolution, agents can initiate negotiations earlier, hold many concurrent negotiations, and continue engaging in these for a longer period of time.

“As a result, even if the principal prefers timely resolution, agents can initiate negotiations earlier, hold many concurrent negotiations, and continue engaging in these for a longer period of time (e.g., begin negotiating summer-2027 rentals in January 2026).”

Probably not a substantial claim: may not pass S2 (checkable).

New approaches to digital accountability, potentially including cryptographic proofs, digital IDs, or decentralized verification platforms, are necessitated to ensure system safety and reliability.

“This necessitates new approaches to digital accountability, potentially including cryptographic proofs, digital IDs, or decentralized verification platforms to ensure system safety and reliability.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Classified as probably not a substantial claim. May combine two claims.

Platform-provided agents enable companies to maintain control over the user experience, benefit from their own R&D investments, and potentially steer users toward preferred options, but require substantial hosting and computation resources while potentially suppressing consumer demand due to alignment concerns.

“Platform-provided agents enable companies to maintain control over the user experience, benefit from their own R&D investments, and potentially steer users toward preferred options, but require substantial hosting and computation resources while potentially suppressing consumer demand due to alignment concerns.”

Probably not a substantial claim: may not pass S2 (checkable).

The ultimate impact of AI agents will depend critically on collective choices adopted regarding agent design, market structures, and regulatory frameworks.

“The ultimate impact will depend critically on collective choices adopted regarding agent design, market structures, and regulatory frameworks.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Proof-of-personhood solutions can prevent sybil attacks and can preserve privacy, but require a larger restructuring of existing systems and mass adoption to be truly transformative.

“Proof of humanhood solutions can prevent sybil attacks and can preserve privacy, but require a larger restructuring of existing systems and mass adoption to be truly transformative.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S2 (checkable).

The new traffic and cost profile may require entirely new, agent-first surfaces-authenticated, rate-limited APIs with machine-readable pricing and consent signals-rather than human-oriented pages.

“The new traffic and cost profile may require entirely new, agent-first surfaces-authenticated, rate-limited APIs with machine-readable pricing and consent signals-rather than human-oriented pages.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.
• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

New conventions will emerge, akin to the robots.txt standard, that define how agents interact with platforms and what activities are permitted or prohibited.

“We expect new conventions to emerge, akin to the robots.txtstandard, that define how agents interact with platforms and what activities are permitted or prohibited.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Classified as probably not a substantial claim.

The same forces that make agents attractive-their tireless persistence, computational superiority, and negligible marginal costs-also threaten to overwhelm existing market structures.

“Yet the same forces that make agents attractivetheir tireless persistence, computational superiority, and negligible marginal costs-also threaten to overwhelm existing market structures.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.
• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

The fundamental tension between user autonomy and platform control mirrors broader debates in digital markets about the optimal balance between personalization, cost, and market power.

“This fundamental tension between user autonomy and platform control mirrors broader debates in digital markets about the optimal balance between personalization, cost, and market power.”

May combine two claims.

By rapidly collecting, analyzing, and comparing product data across markets, agents can mitigate search frictions and enhance allocative efficiency.

“By rapidly collecting, analyzing, and comparing product data across markets, agents can mitigate these frictions and enhance allocative efficiency.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

There are likely to be two types of AI agents along specialization: horizontal and vertical.

“We posit that there are likely to be two types of AI agents along this dimension: horizontal and vertical.”

Probably not a substantial claim: may not pass S2 (checkable).

This capability extends beyond labor markets to any matching context where the complexity of optimal allocation exceeds human computational capacity.

“This capability extends beyond labor markets to any matching context where the complexity of optimal allocation exceeds human computational capacity.”

• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

May combine two claims.

When products are differentiated, lower search costs can paradoxically lead to higher prices and greater dispersion.

“When products are differentiated, for example, lower search costs can paradoxically lead to higher prices and greater dispersion (Ellison and Ellison 2018;Kaye 2024).”

• cites Ellison and Ellison 2018 sourceNot checked yet

  Checked before — its earlier verdict is held back until a new check runs.

• cites Kaye 2024Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

The choice between bring-your-own-agent and platform-provided agent models presents trade-offs from the consumer perspective.

“The choice between bring-your-own-agent and platform-provided agent models presents trade-offs from the consumer perspective.”

May combine two claims.

This enhanced informational access enables personalized pricing strategies that can improve price discovery and market efficiency, lowering the deadweight loss previously caused by information asymmetry.

“This enhanced informational access enables personalized pricing strategies that can improve price discovery and market efficiency, lowering the deadweight loss previously caused by information asymmetry.”

May combine two claims.

As agentic browsing becomes ubiquitous, overall traffic volumes are likely to surge, leaving website owners to bear these costs.

“However, as agentic browsing becomes ubiquitous, overall traffic volumes are likely to surge, leaving website owners to bear these costs.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S4 (based).

AI agents may be much less influenced by puffery and selective information disclosure in ads than humans.

“AI agents may be much less influenced by these types of ads than humans.”

• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

Few consumers will literally create their own models, so they will rely on firms to supply them.

“Presumably, few consumers will literally create their own models so they will rely on firms to supply them.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Bowling-shoe agents may sacrifice perfect alignment with individual needs, either due to explicit self-preferencing or due to simply not considering options offered on other platforms.

“But bowling-shoe agents may also sacrifice perfect alignment with individual needs, either due to explicit self-preferencing or due to simply not considering options offered on other platforms.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

While the capabilities of AI agents are new, their transformative impact is ultimately governed by fundamental economic forces such as supply and demand.

“While the capabilities of AI agents are new, their transformative impact is ultimately governed by the fundamental economic forces such as supply and demand.”

May combine two claims.

While personalized pricing may benefit consumers by closely matching products to individual preferences, it may also lead to inequitable outcomes if firms exploit consumer-specific price elasticities.

“While personalized pricing may benefit consumers by closely matching products to individual preferences, it may also lead to inequitable outcomes if firms exploit consumer-specific price elasticities.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Economists have the tools to play a key role in this transformation.

“Economists have the tools to play a key role in this transformation, but whether they choose to do so remains to be seen.”

Probably not a substantial claim: may not pass S2 (checkable).

As agents proliferate, platforms will need to adjust.

“As agents proliferate, platforms will need to adjust.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S3 (load-bearing).

Providers may experiment with more sophisticated models, including tiered pricing or two-part tariffs.

“Providers may also experiment with more sophisticated models, including tiered pricing-where a limited free version serves price-sensitive users while a premium plan with full functionality cross-subsidizes it-or two-part tariffs, in which users pay an upfront subscription plus per-prompt or per-token fees.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

From a consumer's perspective, the agentic interaction becomes less inspectable, highlighting transparency and trust as critical challenges that will shape adoption.

“From a consumer's perspective, however, the agentic interaction becomes less inspectable-highlighting transparency and trust as the critical challenges that will shape adoption.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

These mechanisms may effectively separate sensitive inquiries from signaling concerns, enhancing transparency and efficiency across sensitive contexts.

“These mechanisms may effectively separate sensitive inquiries from signaling concerns, enhancing transparency and efficiency across sensitive contexts.”

• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Platform-provided agents eliminate technical complexity and may be better if they are trained on platform specific data.

“Conversely, platform-provided agents eliminate technical complexity and may be better if they are trained on platform specific data.”

May combine two claims.

The deferred acceptance algorithm guarantees stability and strategyproofness for proposing parties.

“Consider the deferred acceptance algorithm (Gale and Shapley 1962), which guarantees stability and strategyproofness for proposing parties.”

• cites Gale and Shapley 1962 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

AI will serve as a filter between humans and the broader digital environment.

“AI will serve as a filter between humans and the broader digital environment.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S4 (based).

By delegating sensitive questions to AI agents, parties can credibly commit to privacy-preserving interactions.

“By delegating sensitive questions to AI agents, parties can credibly commit to privacy-preserving interactions.”

• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

The capacity of AI agents to dramatically reduce transaction costs as automated intermediaries could unlock new forms of market participation, enable previously infeasible mechanisms, and push allocative efficiency closer to competitive ideals.

“The capacity of AI agents to dramatically reduce transaction costs as automated intermediaries could unlock new forms of market participation, enable previously infeasible mechanisms, and push allocative efficiency closer to competitive ideals.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.
• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Horizontal agents are generalists that can span many tasks and platforms, leveraging a single memory/preferences layer across markets and arbitraging options when beneficial, but they may lack domain-specific tooling or compliance features.

“Horizontal agents are generalists that can span many tasks and platforms, leveraging a single memory/preferences layer across markets and arbitraging options when beneficial, but they may lack domain-specific tooling or compliance features.”

May combine two claims.

Most matching platforms rely on far simpler recommendation systems that present all participants with the same best options, often leaving outcomes unstable or leaving many participants unmatched.

“As a result, most matching platforms (e.g., Upwork, Tinder) rely on far simpler recommendation systems that present all participants with the same”

Classified as probably not a substantial claim.

This filtering function poses challenges to existing digital business models that rely on bundling content with advertising or engagement-driven recommendations.

“This filtering function poses challenges to existing digital business models that rely on bundling content with advertising or engagement-driven recommendations.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Interfaces made primarily for agents may be useful because of their speed, and the ability of the interface provider to tailor information to the agent.

“Interfaces that are made primarily for agents may be useful because of their speed, and the ability of the interface provider to tailor information to the agent.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

AI agents could review incoming information streams and selectively transmit content based on user preferences and actual utility, substantially reducing exposure to irrelevant material and ultimately steering digital platforms toward prioritizing user preferences over attention capture.

“AI agents could review incoming information streams and selectively transmit content based on user preferences and actual utility, substantially reducing exposure to irrelevant material and ultimately steering digital platforms toward prioritizing user preferences over attention capture.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.
• AssumptionRests on an assumption: AI agents have superior computational abilities relative to humans.

Probably not a substantial claim: may not pass S3 (load-bearing).

Over time, there may be a shift to alternative platform monetization models, such as subscription services, which have less dependence on advertising.

“Over time, we may see a shift to alternative platform monetization models, such as subscription services, which have less dependence on advertising.”

• AssumptionRests on an assumption: AI agents will proliferate in economic transactions.