There is a structural break happening in the consumption of investment research.

For decades, the model was stable. Research was produced, distributed, and consumed in formats that preserved its economic and intellectual integrity. A report was read, a model was reviewed, a call was held.

Attribution was clear, entitlements were enforced, and – critically – feedback loops existed. Producers of research could see who consumed it, how often, and with what impact.

That model is now breaking.

Assessing the Challenges: From Research Documents to Structured Data

As large language models become embedded into investment workflows, research is no longer being “read” in the traditional sense. It is being ingested, distilled, summarized, decomposed, and recombined.

Increasingly, AI systems are not navigating to research – they are querying it directly, extracting signals and generating outputs that are integrated into decision-making processes.

In that transition, three foundational pillars of the research ecosystem are lost almost instantly.

1. Attribution Fades.

The analyst, the franchise, and the originating institution are often no longer visible in the output that ultimately reaches the portfolio manager.

2. Entitlements Weaken.

Once research is absorbed into internal AI systems, traditional controls over access become difficult to enforce in any meaningful way.

3. The Feedback Loop Deteriorates.

Producers of research lose visibility into how their content is used, which elements carry value, and how that usage translates into commercial outcomes.

This is not a marginal shift. It is more profound than the unbundling introduced under MiFID II. That regulatory change altered the economics of research. AI-driven consumption alters its structure.

What is emerging is a world in which the buy side increasingly operates “black box” systems – AI environments that ingest research, extract insights, and surface outputs with limited transparency to the original producers. Even where formal agreements exist, they tend to cover narrow use cases, while broader, less visible usage continues to expand.

At the same time, the sell-side is responding with pragmatism. Firms are experimenting, engaging selectively, and acknowledging that AI usage is becoming embedded in investment workflows. But there is not yet a clearly defined framework for attribution, entitlement preservation, or measurement.

This Creates a Rare Moment of Leverage for the Sell Side.

For the first time in decades, sell-side research departments have the ability to influence the terms under which their content is consumed.

AI systems require a constant flow of high-quality, domain-specific written content to remain relevant. That makes differentiated research – both current and historical – foundational to how these systems operate.

The question is no longer whether research will be consumed by AI. It already is. The question is whether that consumption will occur in a way that preserves the integrity of the research ecosystem.

The future consumption layer for investment research will not be:

• Email

• PDF

• Traditional portals

It will be AI systems interacting directly with research through structured interfaces.

To support that future, the industry will need a new consumption framework – one that is designed for machine interaction rather than human reading.

Such a framework must ensure that research can be accessed in a structured way, that entitlements are preserved as content moves through AI systems, and that attribution remains attached to outputs derived from that research. It must also provide a way to understand how research is being used in this new, more fragmented mode of consumption.

The most immediate and sensitive issue is attribution.

As AI systems generate summaries and insights, the connection to the originating source becomes attenuated. Without a mechanism to preserve that linkage, the economic and intellectual value of research is at risk of being separated from its producers.

What is required is a model in which attribution persists as content is transformed – where derived outputs remain anchored to their source, and where that connection is visible and reliable.

Closely related is the question of entitlement enforcement.

In an AI-driven environment, access control cannot be limited to the point at which a document is opened. It must extend to how content is accessed, queried, and incorporated into downstream outputs. This implies a shift toward more granular, system-level enforcement of permissions.

Finally, the industry must rethink measurement.

Traditional readership metrics were designed for a world in which consumption was discrete and observable. AI-driven usage is continuous, partial, and often indirect. Understanding value in this context requires new forms of visibility into how research contributes to the generation of insights.

In this environment, infrastructure becomes central.

A platform that sits at the intersection of research creation and distribution is uniquely positioned to support this transition – embedding attribution, preserving entitlements, and restoring visibility into consumption as research moves into AI systems.

The opportunity is to evolve from a distribution mechanism into a governance layer for research in the age of AI.

If the industry does not define how this new model should work, it will be defined implicitly by the behavior of AI systems and the incentives of those who build them.

If it does, it can preserve attribution, maintain control over access, and reestablish a feedback loop between those who produce research and those who rely on it.

The window to act is now.