Google Introduces Open Knowledge Format

Google has announced the Open Knowledge Format (OKF), an open specification designed to help structure and exchange the information that AI systems rely on in order to carry out useful tasks. The idea behind the format is to make organisational knowledge easier for both humans and AI agents to access, understand and reuse in a consistent way.

The announcement highlights a growing challenge in modern AI development. While foundation models continue to improve, their effectiveness is still often limited by missing or incomplete context. Even though AI tools are now capable of writing code, summarising documents and analysing datasets, they still depend heavily on accurate and relevant background information to produce reliable outcomes.

The role of context in AI systems

One of the key issues OKF aims to address is how fragmented organisational knowledge tends to be in practice. In most companies, important information is spread across a wide range of systems. This can include internal documentation, data warehouses, code repositories, wikis, shared drives and other isolated storage platforms.

Because this knowledge is not stored in a unified or structured way, AI agents are often required to collect and interpret it from multiple sources before they can complete a task. This adds complexity and can reduce accuracy, especially when systems rely on incomplete or inconsistent information.

Google states that OKF is intended to solve this by creating a shared format that allows knowledge to move more easily between humans, AI agents, tools and organisations.

What the Open Knowledge Format is designed to do

At its core, OKF is a structured format for representing organisational knowledge in a way that can be used across different environments. It is designed to capture key concepts such as datasets, metrics, APIs, database tables and operational runbooks, and organise them into a consistent structure that can be understood by both machines and people.

Rather than being tied to a specific platform, OKF is intended to remain flexible and portable. This means the same structured knowledge can be used across different AI systems, tools and workflows without needing to be rewritten or reformatted.

Google has also provided early reference implementations to demonstrate how the format could be used in practice. One example is an enrichment agent that scans BigQuery datasets and automatically generates structured concept documents for tables and views. This process is then enhanced with additional information such as documentation references, schemas and data relationships.

Another example is a static HTML visualiser that converts OKF bundles into an interactive graph view. This allows structured knowledge to be explored visually in a single file without requiring any backend system or installation. Google has also released sample datasets, including GA4 e-commerce data, Stack Overflow content and Bitcoin datasets, to demonstrate how OKF might be applied in real-world scenarios.

Importantly, Google stresses that these examples are not strict requirements. Instead, they are intended as demonstrations to show possible implementations, while encouraging others to build their own tools and workflows around the format.

Who OKF is intended for

OKF is built around a producer and consumer model, where some users are responsible for creating and maintaining structured knowledge, while others consume it through AI systems and tools.

AI agents and systems

AI agents and large language models are the primary consumers of OKF. By providing structured and consistent context, the format helps these systems perform tasks more accurately and efficiently.

This includes a wide range of use cases such as coding assistants, data analysis tools, research agents, enterprise AI systems and automated workflows that rely on structured organisational knowledge.

Human contributors

Despite being designed with AI in mind, OKF remains fully readable and editable by humans. The format uses markdown files combined with YAML frontmatter, which allows people to work with it using standard tools without requiring specialised software.

This makes it relevant for a broad range of professionals, including AI developers, software engineers, data engineers, analytics teams, technical writers and business users involved in knowledge management.

Organisations and enterprise use

For organisations, OKF offers a way to bring together knowledge that is often scattered across multiple systems. This includes documentation platforms, metadata catalogues, code repositories and internal operational tools.

By standardising how this information is structured, organisations can potentially make it easier to share knowledge internally and integrate it into AI-powered systems and agents. This could be particularly useful for teams building internal AI assistants or managing large-scale data environments.

An open specification still in early development

Google has positioned OKF as an open standard rather than a finished product. The specification, reference implementations and sample bundles are publicly available on GitHub, allowing developers to explore and experiment with the format.

The company also makes clear that OKF is still in an early stage of development. Version 0.1 is described as a starting point that will evolve over time as more producers and consumers begin to use it and contribute feedback.

Google notes that releasing the format openly from the beginning is important for shaping a shared standard that can be widely adopted. The expectation is that both tools and use cases will expand significantly as the ecosystem develops.

Why this development matters

The introduction of OKF reflects a wider shift in how AI systems are being built and deployed. As AI agents become more capable and more widely used, the need for structured, machine-readable knowledge is becoming increasingly important.

Rather than relying on fragmented documentation or inconsistent data sources, OKF aims to create a standardised layer of knowledge that can be shared across tools and systems. This could help improve the reliability and accuracy of AI outputs, particularly in complex environments where context is critical.

It also highlights a growing focus on interoperability in AI development. Instead of isolated tools and proprietary formats, there is increasing interest in shared standards that allow different systems to work together more effectively.

Final thoughts

While still at an early stage, the Open Knowledge Format signals a clear direction of travel in AI infrastructure. The focus is shifting not only towards more powerful models, but also towards better structured knowledge that those models can use.

If widely adopted, OKF could play a role in reducing fragmentation across organisational data and improving how AI systems interact with real-world information. For now, it remains a developing standard, but one that reflects an important challenge in the next phase of AI adoption: ensuring that intelligence is supported by reliable, structured context.