Supporting AI standards at GitBook: What OKF, MCP and llms.txt tell us about the future of docs

Google recently introduced the Open Knowledge Format (OKF), a new open standard designed to make documentation easier for AI systems to understand and consume.

We recently published a deeper explainer on OKF, but the announcement also felt like a good opportunity to explain how we think about emerging AI standards at GitBook — and why we believe supporting them has become a foundational responsibility for any modern documentation platform.

I’m Steven Hall, Head of Engineering at GitBook, and a big part of my role is thinking about documentation as context for AI systems. OKF is the latest example of a broader shift that has been reshaping our industry for years.

AI-ready documentation is now table stakes

AI support is no longer a differentiator for documentation platforms. In 2026, it’s table stakes.

Documentation is increasingly consumed by agents and assistants before it ever reaches a human reader. Our recent research found that AI systems now account for more than half of documentation readership, turning docs into a critical source of context for automated workflows.

The best documentation teams already recognize this. They expect their platform to support emerging standards, expose structured content, and make knowledge accessible to both humans and machines.

“Your users are just telling their agents to build them something, but not mentioning your product. If you’re not in the agent’s frame of reference, you’re never going to come up.”
Dachary Carey, MondoDB
2026 State of Docs Report

The question is no longer whether documentation platforms should support AI workflows — it’s how quickly they can adapt as those workflows evolve.

How we approach AI standards at GitBook

The AI ecosystem moves quickly. New standards appear constantly, and it’s not always obvious which ones will last.

Our approach is simple: pay close attention, discuss developments early, and move quickly when something has clear value for customers.

That’s how we approached llms.txt and MCP support, and it’s how we’re thinking about OKF today.

What makes OKF particularly interesting is that much of its guidance already aligns with how GitBook content is structured. Markdown-native and folder-based, with structured metadata, reusable content, and machine-readable documentation — these have been core principles of the platform for years.

Often, supporting a new standard isn’t about reinventing documentation. It’s about ensuring the high-quality content customers are already creating can be surfaced in the formats modern AI systems expect.

How we approach AI features in GitBook

That same philosophy guides how we think about AI features we build into GitBook. We’re not interested in adding AI functionality for the sake of a launch announcement. Our goal is to improve documentation workflows in meaningful ways while keeping humans involved and in the loop.

That matters because documentation increasingly acts as a source of truth for AI systems. If agents are answering questions, generating summaries, or taking actions based on your docs, the underlying content has to be accurate, complete, and up to date.

AI can accelerate documentation work, but quality still depends on strong processes and human oversight.

We’re already working on what’s next

While standards are important, I'm even more interested in the workflows emerging around them.

Right now we’re actively exploring ways to give AI agents deeper access to GitBook so they can help create, customize, and publish documentation sites within the guardrails defined by a team. We’re also investing improving our proactive agent, to help you identify knowledge gaps, outdated content, and opportunities for improvement before those issues affect users.

At the same time, we’re looking at how teams can maintain consistency, quality, and brand voice as AI contributes more content to documentation workflows.

And of course, we’ll continue to support new, emerging standards quickly — we see that as a foundational responsibility of any modern documentation platform.

The next challenge is documentation quality

Supporting standards like OKF, MCP and llms.txt is increasingly a baseline expectation, and we’ll continue moving quickly whenever meaningful new standards emerge.

The more interesting challenge is what happens after that.

At GitBook, our focus is on helping teams create documentation that works equally well for humans and AI systems — improving speed and scale without losing the expertise, judgment, and editorial oversight that make documentation trustworthy in the first place.

That balance is where we believe the future of documentation will be built.

→ Research: AI agents are now the majority reader of your docs

→ Humans are still the real readers of your documentation

→ How to optimize your documentation for AI (without breaking it for humans)