We added NotebookLM and DeepSeek to the AI referrer taxonomy

Two more AI platforms are now in the default taxonomy for CQI Referrer Attribution, bringing the total to 19.

This is a short post, but the reason behind it is worth explaining because it illustrates something that is easy to miss about AI referrer traffic.

NotebookLM

Google's NotebookLM is a research tool that lets users upload documents, articles, and web pages, then ask questions across all of them. When a NotebookLM user cites a source and clicks through to the original, the referrer domain is `notebooklm.google.com`.

NotebookLM traffic is qualitatively different from ChatGPT or Perplexity traffic. The user has already read or processed your content within the tool. They are returning to the source for a specific reason - to verify something, to read the full context, or to share it. It is a high-intent click from someone who has already engaged with your material.

If you are not separating this from general Google referral traffic, you are not seeing it.

DeepSeek

DeepSeek's web interface sends `chat.deepseek.com` as a referrer for traffic it generates. It is a large and growing platform, particularly in technical and developer communities. It merits its own row in the taxonomy rather than being absorbed into Referral.

Why the taxonomy is editable

The 19 platforms in the default taxonomy are the ones with confirmed referrer domains and meaningful traffic volumes at the time of each release. They are not the complete list of AI tools that exist, and they will not remain current indefinitely.

New platforms launch. Existing platforms change their referrer behaviour. Some platforms route traffic through intermediate URLs that obscure the original source. The taxonomy needs to change faster than a plugin can realistically ship updates.

For that reason, the taxonomy is fully editable from within the plugin. Add an entry, paste in the domain, set a display name, activate it. It takes thirty seconds. You do not need to wait for a plugin update to start attributing traffic from a new platform.

You can also export the full taxonomy as JSON, edit it in bulk, and re-import it. If you have multiple sites, you can maintain a master taxonomy file and apply it consistently across all of them.

What we track and what we cannot

We track AI referrer traffic that arrives with a referrer header identifying an AI platform domain. We cannot track AI traffic that arrives with no referrer header, which is a meaningful fraction depending on the platform and the user's path.

That fraction is not our data to have - it is not anyone's. The referrer header is the signal available, and we classify everything it carries. What it does not carry remains in Direct for everyone.

The honest position: if you are seeing AI Referrer traffic in your reports using a referrer-based tool, that is real traffic from real users who clicked through from an AI platform. It is a floor, not a ceiling. The actual volume of AI-driven interest in your content may be higher. But the attributed traffic is accurate.

*CQI Referrer Attribution v1.6.0 added NotebookLM and DeepSeek to the default AI taxonomy, now covering 19 platforms. The taxonomy is editable, exportable as JSON, and can be extended without a plugin update.*