Google Introduces Five Trust Signals

Google has released version 0.2 of its Open Knowledge Format (OKF), introducing five new trust signals designed to help developers, AI systems and applications better assess the reliability of structured information. The update aims to improve transparency by making it easier to understand where information comes from, how trustworthy it is, whether it is still accurate, and how it was produced.

According to Google, these additions allow anyone using OKF bundles to make more informed decisions when working with AI-generated or structured knowledge.

Five New Trust Signals

The latest update introduces five core trust features:

Provenance
Trust
Freshness
Lifecycle
Attested Computation

Together, these signals help answer five important questions about any piece of information:

Where did it come from?
How reliable is it?
Is it still accurate?
Is it the latest version?
Was it calculated correctly?

The goal is to improve confidence in the data without assigning a single trust score, allowing different systems to evaluate the information using their own criteria.

Provenance Tracks Information Sources

A new Sources field has been added to record where information originates.

This may include external websites, internal documents, datasets or project files used to create a concept. Additional details such as the author, modification date and usage frequency can also be included.

Rather than deciding whether a source is trustworthy, Google simply records these signals, allowing applications and AI systems to make their own judgement based on the available evidence.

This provides greater transparency by showing exactly what information was used to build a particular knowledge bundle.

Generated and Verified Fields Improve Trust

Google has also introduced two new fields that distinguish between who created information and who confirmed its accuracy.

The Generated field records the individual or AI system responsible for producing the content, while the Verified field records independent reviewers who have checked its accuracy.

This allows applications to identify whether information has only been generated automatically or has also undergone human verification.

As a result, organisations can apply different levels of trust depending on their own quality requirements.

Freshness and Lifecycle Indicators

Keeping information up to date is another focus of the update.

The new Status field identifies whether a concept is currently a draft, stable or deprecated, while the Stale After field specifies when information should be reviewed again before being relied upon.

These additions help prevent outdated information from being used in future work while still preserving historical records where needed.

For organisations managing large volumes of structured knowledge, these indicators provide a clearer way to monitor content over time.

Attested Computation Verifies Calculations

One of the biggest additions in OKF version 0.2 is a new concept called Attested Computation.

Instead of simply storing a calculated value, this feature records the approved method used to produce that result. It can also verify that the calculation followed the correct process rather than relying on an alternative method created by an AI model or another system.

This is particularly useful for financial reporting, business analytics and other situations where consistent calculations are essential.

By documenting the approved calculation method alongside the final result, organisations can improve accountability and reduce the risk of inaccurate outputs.

Why the Update Matters

As AI becomes increasingly involved in creating and analysing information, transparency has become more important than ever.

Google’s latest Open Knowledge Format update focuses on helping users understand not only what information is presented, but also where it came from, whether it has been reviewed, how current it is and how any figures were calculated.

Rather than relying on a single trust score, Google provides the underlying evidence so developers and organisations can evaluate information using their own standards.

Version 0.2 has now been published alongside updated documentation in Google’s GitHub repository, giving developers immediate access to the new trust features as they continue building AI-powered applications and knowledge systems.