> For the complete documentation index, see [llms.txt](https://help.globalnaturewatch.org/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.globalnaturewatch.org/resources/product-status-and-updates.md). # Product Status & Updates **Global Nature Watch (GNW) Horizon is currently in preview** — an experimental release designed to test and improve AI capabilities for monitoring nature. While in preview, users may notice changes in assistant behavior, available data and application features as we iterate based on feedback and technical improvements. During this phase: * The **AI assistant** may behave inconsistently or change as we update underlying models. * **Features and datasets** may be added, updated or removed as we improve coverage and reliability. * **Results may contain errors or incomplete interpretations** — always verify with primary data sources before using results for analysis, reporting or decision-making. ## :new:Recent & Upcoming Changes **July 2026** **Global Forest Watch is now Global Nature Watch** Users landing on the [Global Forest Watch platform](https://www.globalforestwatch.org/) will see a new announcement banner and updated logo marking the transition to Global Nature Watch. Further updates to the platform will follow in the coming weeks. Within this transition, this Global Nature Watch has now become ***Global Nature Watch Horizon (GNW Horizon)***. This release marks the start of the rebrand transition and introduces a redesigned map interface alongside two new modes for running an area analysis. **Updated Data Catalogue and Area Tools panels** Both panels have been redesigned as part of the ongoing map interface refresh. The Data Catalogue makes it easier to browse and add datasets, and the Area Tools panel has been updated to match. These are visible to all users now. **Two ways to analyze an area (coming soon)** Following the introduction of one-click analysis, you will soon be able to choose how your analysis runs when you click an area of the map: * **Curated analysis** — returns a pre-built chart immediately, without using a query credit. Best for a fast answer to a standard question. * **AI-assisted analysis** — generates a more flexible, bespoke response using the assistant. Uses one query credit. This feature is currently in final testing and will be available to all users soon. **WRI and GFW blog sources now available** The GNW Horizon assistant can now draw on blog content from WRI and GNW (Formerly, Global Forest Watch) when answering broad or exploratory questions. When relevant blog posts are available, it will surface them as inline sources alongside its response, with links so you can read the original content directly. This works best for open-ended questions where you want context alongside data, for example, asking about deforestation trends in a region, or the relationship between fire and tree cover loss. The assistant will suggest follow-on analyses based on what the blog content covers. **June 2026** **One-click analysis, quality-of-life improvements, and accessibility updates** This release focuses on making analysis faster to reach, easier to act on, and more accessible across different workflows. **One-click analysis** You can now click directly on an area of the map to trigger an analysis - no prompt required. When you click, a nudge appears in the chat inviting you to analyze the selected area. Click it, and the assistant gets to work immediately. This is the fastest way to get from the map to an insight. This feature works alongside the existing conversational flow - you can still type a question at any point. **Export and continue your analysis** Insights can now be taken further in a few new ways: * Export as image - Download any chart as a PNG using the Download menu on the insight card. * Continue in your preferred tool - Use the "Continue in..." dropdown to export your data and context as a prompt, ready to open directly in Claude, ChatGPT, or Gemini. **Chart scaling** When chart data shows little variation across the y-axis, you can now toggle "Fit to data" to rescale the axis and make smaller differences easier to read. **Accessibility improvements** Area of interest borders on the map are now more visually distinct across both light and dark basemaps, making it easier to identify selected areas at a glance. You can now also press a stop button to interrupt the assistant while it is reasoning **A redesigned workspace, a smarter assistant, and new fires data** This release brings a significant update to how GNW Horizon looks and behaves. The map is now front and center, GNW Horizon has been rebuilt to handle more granular actions, and a new fires dataset has been added, along with tools to help you choose the right one. **A fresh look and feel** * **Map-first layout:** The workspace is now built around the map. The chat panel is more compact and glanceable, keeping the focus on where you are and what you are exploring. * **Redesigned cards:** Dataset, area, and insight cards have been updated for clarity, so you can see at a glance what GNW Horizon has selected and why. * **What's New panel:** A new in-app panel surfaces recent changes each time you open the application, so you are always up to date. Look for the notification in the header. * **Persistent preview badge:** The preview disclaimer has moved to the header, where it is always accessible. It now links directly to key Help Center pages so you can get answers without leaving the app. * **Ready to explore immediately:** Tree Cover Loss and GADM administrative boundaries are now active by default when you open GNW Horizon. Click any area and start asking questions straight away. **Smarter, more composable assistant** GNW Horizon has been rebuilt on a skills-based architecture. Rather than running a end-to-end analysis, it now performs individual actions and combines them when needed. * Navigate the map to a location: *Try: "Go to Japan."* * Add a specific dataset to the map: *Try: "Add tree cover loss to the map."* * Explain an insight already on screen: *Try: "What does this chart show?"* * Run a full analysis when that is what you need: *Try: "Analyze tree cover loss in Indonesia over the last 10 years."* This makes GNW Horizon more responsive to shorter, more directed requests. **Tree Cover Loss due to Fires (2001–2025)** Tree Cover Loss due to Fires is now available as a map layer and analyzable by GNW Horizon, covering annual data from 2001 to 2025. You can toggle it on in the layer browser, filter by year, and ask GNW Horizon questions about it directly. This dataset can be filtered by: * Tree cover threshold * Intact Forest Landscapes (IFL) * Primary forests **Dataset nudges** With three fire-related datasets now available - DIST Alerts, Tree Cover Loss, and Tree Cover Loss due to Fires - GNW Horizon will prompt you to choose the most relevant one when your question could be answered by more than one option. It explains what each dataset covers and when it is most useful, so you can make an informed choice before the analysis runs. *Try: "I'm interested in fires in Brazil" -* GNW Horizon *will offer options with a short explanation of each.* **May 2026** **A more collaborative assistant** The assistant now works more closely with you in conversation, giving you finer control over how it acts. You can tell it exactly what to do, and it holds your active area, dataset, and date range between turns so you can have a real back-and-forth without restating context every time. * Pull-only requests stop at the data: When you ask for raw data, the assistant retrieves it and stops, so you can decide what to do next. Try: "Pull DIST alerts for Bern, Switzerland, for the last 2 weeks." * Analysis requests go end-to-end: When you ask for a chart, insight, or analysis, the assistant continues through to a visualization with a short interpretation. Try: "Analyze tree cover loss in Canada, over the last 5 years." * Clearer routing for short requests: Dataset-only ("pick TCL by driver") and area-only ("zoom to Gabon") requests now stop after the matching step instead of escalating into a full analysis. Try: "Pick the tree cover loss by driver dataset" - the assistant selects it and waits. **The map as a workspace** The map is now where you and the assistant work together. Insights appear on the map where the analysis happened - pinned to the area in focus, rather than scrolling away in chat. A redesigned legend brings the assistant's choices into plain view: active layers, the area in focus, the dataset and date range in use, and the opacity of each layer. You can see at a glance what's driving each result, and adjust it directly. Chat becomes a cleaner narration of what the assistant did and why; the map is where you explore and compare. * Insights pinned to the map: Generated insights appear on the map alongside the area they describe, instead of in the chat thread. Navigate between recent insights with prev/next controls. Try: "Analyze tree cover loss in Pará since 2001" — the resulting chart appears on the map, not in chat. * The assistant's selections, made visible: The area in focus, the active dataset, and the current date range are all surfaced on the map and in the legend, so it's always clear what the assistant has chosen on your behalf. * A cleaner chat thread: Chat is now a narrative of what the assistant did and why. Dataset cards, area cards, and insights live in the workspace, not the chat — so the conversation stays scannable even after long sessions. **Intact Forest Landscapes dataset** GNW Horizon now hosts the **Intact Forest Landscapes dataset.** Ask GNW Horizon to filter TCL within Intact Forest Landscapes via the chat function. The results will then be visible on the map using 2025 data. **April 2026** **Tree Cover Loss 2025** The latest annual Tree Cover Loss data is now available in GNW Horizon, with two new ways to refine your queries: filtering for primary forest and setting a specific canopy threshold. Together these let you ask more precise questions about where, and what kind of, forest loss is happening. * **Latest Tree Cover Loss data:** Ask about the most recent year of forest loss alongside the historical record (2001 onwards). *Try: "How much tree cover loss occurred globally in 2025?"* * **Filter by Primary Forest:** Narrow any tree cover loss query to primary forest only, helping you separate loss in old-growth ecosystems from loss in younger or planted forest. *Try:* *"Show primary forest loss in Indonesia between 2002 and 2025."* * **Specific tree canopy threshold:** Set the minimum canopy density (e.g. 30%, 50%, 75%) used to define "forest" in your query, so results align with the methodology you or your audience expect. *Try:* "What was tree cover loss in the DRC in 2025 using a 75% canopy threshold?" **Improvements** * **Ask about multiple areas at once:** You can now include several locations in a single question e.g. *"Show deforestation in Brazil, DRC and Indonesia."* Results are returned together for easy comparison. * **Fewer errors, faster recovery:** The assistant now retries automatically when something goes wrong, reducing the chance of a query failing mid-analysis. * **Model change:** Migrated to Google Gemini Flash, improving response speed without sacrificing quality for more complex queries. **New features** * **Step-by-step AI reasoning now visible:** See how the AI reached its answer: the assistant now shows its reasoning steps as it works through your question. * **Improved legends and AOI handling:** *Map legends and area labels are now clearer and more consistent across queries.* * **Chart improvements:** General improvements to the chat interface for a smoother experience and clearer presentation * **Data accessibility:** Chart data can now be viewed in fullscreen, toggled between a chart and table view, and downloaded as a `.csv` file. **Fixes** * General improvements to interface speed and visual consistency. **December 2025** **Improvements** * **Model upgrade:** Migrated to [**Google Gemini Pro 3**](https://deepmind.google/models/gemini/pro/). * **Chart generation reliability:** The GNW Horizon Assistant now generates executable code when creating charts (rather than manipulating data directly), using a capability called CodeAct. **New features** * **“See how this was generated”:** Added UI component accessible from insights that opens a pane showing the full data transformation pipeline (**API call → intermediate transforms → final output**) supporting validation and **reproducible charts**. Raw analytics data can also be downloaded in `.csv` format. **Data** * **New dataset:** Added **Deforestation (sLUC) Emission Factors by Agricultural Crop** dataset with admin-unit coverage up to country level. **Fixes** * Multiple fixes to improve overall **accuracy**, **transparency** and reduce **bugs** across common workflows. **October 2025** * **Conversation thread sharing:** Users can now make their conversations public and share urls. * **Agent improvements:** Minor updates now to better recognize dataset-specific terminology and proactively surface relevant cautions when generating insights. ## :speech\_balloon:Feedback Your feedback directly helps shape the next version of GNW Horizon. Please tell us about your experience and share any suggestions: * [Complete the survey](https://surveys.hotjar.com/860def81-d4f2-4f8c-abee-339ebc3129f3), or * Email us at [landcarbonlab@wri.org](mailto:landcarbonlab@wri.org?subject=Global%20Nature%20Watch%20user%20feedback) ## :link:Related Resources * [Dataset library](broken://pages/wVRZOXsGSih9NnMcZDft) * [Models ](/get-started/models.md) * [For Developers](/technical-documentation/for-developers.md)