NegotiateCOP Gets a Major Update Ahead of COP31

Practical usefulness of NegotiateCOP

The updated app brings significant improvements to its core features. Users can now explore the UNFCCC document corpus through two complementary approaches: document search and an AI-powered chat assistant.

Document search combines two search modes. Exact matching through keyword search helps users find specific terms and phrases in documents, while Smart Search uses semantic similarity to identify relevant passages and documents, even when using different terms. For example, a search for "deforestation" can also identify documents that address "forest loss". Additional filters help narrow down results to relevant documents, countries, negotiation sessions, and thematic areas.

For party submissions, users can explore documents grouped by calls for submissions, thematic areas, and submission types. Each document can be analyzed individually, with AI-generated summaries and key discussion points. The updated position comparison feature also allows users to examine how different submissions relate to each other and identify areas of alignment. Thematic clusters are taken from the documents ("matters"), which are then compared to the strength of their language. In addition to the direct comparison of two submissions, the positions of all other submitted documents under the respective call for submissions can also be checked in an overview view directly after alignment.

As a second main component, we proudly introduce the new AI Chat Assistant. Users can ask questions in natural language and explore relevant information in the document corpus without having to know exactly which document or terminology to search for. The assistant uses agentic retrieval-augmented generation (RAG) to anchor its answers in official UNFCCC documents. The user's question is understood, the agent decides which filters, database, chunks are relevant to the question and gives a generative answer based on the data previously selected. The sources are included as in-line citations, so that users have a direct reference to the original document.

Back to our roots with an AI-powered chat

The return of the chat assistant and in general a semantic search is particularly pleasing for us as GIZ Data Lab. This feature inspired the original NegotiateAI prototype, which we developed in 2025 as part of the UN negotiations for a global plastics treaty.

To see this initial approach now being embedded in the much broader and more complex context of international climate negotiations, is a great achievement for NegotiateCOP. We already wanted to release this feature last year for COP30, where we clearly have seen the potential of this approach. At the same time, extensive testing involving actual negotiators has revealed weaknesses that we were unable to fully anticipate during development, which is why we have decided to disable the chat for the time being.

As for the other features, we were genuinely surprised by the level of interest and positive feedback the app received at COP30.. But the experience also gave us a much clearer picture of what needed to be improved. We identified numerous minor bugs, inconsistencies, and features that didn't work as intended (for example, the filter functions were sometimes illogical nested and didn't condition each other, so suggested filter selection generated a completely different output). Feedback from negotiators and analysts helped us understand where the platform was useful, where it was failing, and what made it more relevant to their day-to-day work.

The latest update addresses many of these problems and significantly improves the usability and reliability of the platform, also in terms of the user interface in a more modern look. It brings back the chat to the home page and extends the semantic search and chat functions that formed the starting point of our work.

A joint effort between government and development partners

We are particularly pleased to have completed this major update in time for COP31. This would not have been possible without the commitment of all stakeholders.

In addition to PD's – consultants to the public sector – developers who took lead in developing the core and the complete architecture of the app, our by now long-standing developer partners from &effect have joined the process to bring in valuable experience and technical expertise regarding the chatbot development (as they used to do with NegotiateAI).

But foremost, we would like to thank the data labs of the German Federal Ministry for Economic Cooperation and Development (BMZ), the German Federal Foreign Office (AA) and the German Federal Ministry for the Environment, Climate Action, Nature Conservation and Nuclear Safety (BMUKN). Their tremendous commitment and constant engagement have been crucial in bringing NegotiateCOP to a new level of maturity.

We have found the cooperation within this ministerial and interdisciplinary team to be exceptionally productive. Bringing together different institutional perspectives and areas of expertise has enabled us to develop a tool that can be used to meet a common need, both for negotiators of smaller, under-staffed delegations and by all other negotiators, including German negotiators.

NegotiateCOP at COP31

We are now curious to see how the updated features perform in practice. Ultimately, the value of NegotiateCOP depends on whether it helps people navigate the increasing complexity of international climate negotiations and deal more effectively with the enormous amount of information available.

Johannes Jacob and Robin Nowok will be on site during COP31 in Antalya to present the platform at several side events and walk-in sessions at the German Pavilion, share experiences with negotiators and analysts, and gather feedback for future developments.

So should you be attending COP31 or know colleagues and organizations that might be interested in exploring NegotiateCOP, we would be very happy to get in touch. You can always reach out to us.

We look forward to continuing the conversation in Antalya!