Our Algorithm Is Open Source — Here's Why

When we started building AgentReady, one of the first decisions we faced was whether to keep the scoring algorithm proprietary or make it open source. Every existing web metric — Domain Authority, Domain Rating, Authority Score — is a black box. The methodology is loosely described but never published. The code is closed. The weights are secret.

We understand why those companies made that choice. Proprietary algorithms are defensible. They are harder to reverse-engineer. They create vendor lock-in. From a pure business perspective, keeping it closed makes sense.

But we chose a different path. The AgentReady scoring algorithm is fully open source, published on GitHub under the MIT license. Anyone can read it, run it, fork it, and contribute to it. Here is why.

The fundamental problem with black-box metrics is trust. When nobody can verify how a score is calculated, every number is taken on faith. Agencies use Domain Authority in client decks knowing that no one — including them — can validate the methodology. It works because enough people agree to trust the number, not because the number has been verified.

We think AI readiness is too important for faith-based measurement. Businesses will make real decisions based on their scores — investing in protocol adoption, restructuring content, changing technical infrastructure. Those decisions should be based on a methodology that can be audited, questioned, and improved.

Open source is how you earn trust. When a developer, an agency, or a researcher questions a score, they can read the code, understand the logic, and determine whether the methodology is sound. That transparency makes the metric more credible, not less.

No single team, no matter how talented, can build the best possible AI readiness scoring algorithm alone. The landscape is moving too fast. New AI platforms launch monthly. New protocols emerge quarterly. The signals that matter today may not be the signals that matter in six months.

Open-sourcing the algorithm invites contributions from the people who are closest to the problem. SEO professionals who work with hundreds of sites across different industries. Developers who build the tools and frameworks that power the web. Researchers who study how AI systems discover and cite content. AI engineers who understand the internals of retrieval-augmented generation.

Since publishing the code, we have already received pull requests that improved our schema validation logic, added detection for new AI crawler user agents, and proposed weight adjustments based on independent citation analysis. Each of these contributions made the algorithm better than we could have made it alone.

Let us be precise about what is open source and what is not.

Open source: The complete scoring algorithm — all 8 factor calculations, weight distributions, grade thresholds, and bonus logic. The signal detection code that checks for schema markup, AI protocols, bot access rules, and content structure. The test suite that validates scoring accuracy. And the documentation that explains every decision.

Not open source: The AgentReady platform infrastructure — the scanning pipeline, the web application, the database layer, the API, and the monitoring system. These are proprietary because they represent the service layer, not the methodology. You can run the scoring algorithm on your own infrastructure, but the hosted platform is our business.

Contributing to the AgentReady scoring algorithm is straightforward. The repository is hosted on GitHub and we welcome contributions of all sizes.

Bug fixes — If you find a case where the scoring produces an incorrect or unexpected result, open an issue with a reproducible example. If you can fix it, submit a pull request.

New signal detectors — If you have identified a signal that correlates with AI citation outcomes and is not currently measured, propose it. Include the signal definition, detection logic, and any supporting data.

Weight adjustments — If you have data suggesting a factor should be weighted differently, open a discussion with your methodology and evidence. We evaluate weight proposals quarterly.

Documentation — Improvements to the methodology docs, code comments, or contribution guidelines are always welcome.

Every accepted contribution is credited in the changelog. Major contributors are listed in the project README. We believe the people who help build the standard should be recognized for it.

By open-sourcing the scoring algorithm, we are making a bet: that AI readiness measurement will become more valuable, not less, when the methodology is shared.

We want agencies to embed AgentReady scores in their audits. We want CMS platforms to integrate AI readiness checks into their dashboards. We want competing tools to build on the same foundation so that the industry converges on a shared standard rather than fragmenting into incompatible proprietary metrics.

The web needs a common language for AI readiness — just as it needed a common language for mobile readiness, page speed, and security. Open source is how common languages emerge. We are putting our methodology out there and inviting the industry to build on it, challenge it, and make it better.

If we are right, the AgentReady scoring framework will become the foundation that the entire industry uses to measure and communicate AI readiness. If we are wrong, at least we will have contributed something useful to the conversation. Either way, the web wins.