Comparison · 2026
Natural-language map generation tools.
An honest, same-aperture comparison of six tools that turn data into maps. We compare ShowMeOnMap, Felt, Datawrapper, Kepler.gl, Mapbox Studio, and CARTO on the dimensions that actually decide which tool fits your problem.
Published 2026-05-21 · Updated 2026-10-08
“Natural-language map generation” is the category of tools that take a plain-English question (wildfire risk in California, population density across the US, the world’s major airports) and produce a finished map. The category is new — it only became practical once large language models could plan a multi-step data fetch and pick the right visualization in one pass.
Several established mapping tools have added AI features but kept a manual layer editor as the primary surface. We include them in this comparison because they are the realistic alternatives a reader will be weighing — but we are honest about which tools are answering the question “can I describe a map and get it built?” versus the question “can I build a map myself with AI assistance?”.
At a glance.
| Tool | NL interface | Primary purpose | Pricing | Free tier | Data sources | Open source |
|---|---|---|---|---|---|---|
| ShowMeOnMap | Native | Natural-language map generation | One-time prepaid credit packs ($5 / $15 / $50) | Limited | Both | No |
| Felt | Partial | Collaborative web mapping and editing | Freemium with paid Pro and team plans | Yes | Bring your own | No |
| Datawrapper | None | Publication-ready charts and choropleth maps | Freemium with paid Custom plan | Yes | Bring your own | No |
| Kepler.gl | None | Large-scale geospatial data exploration | Free, open source (MIT) | Yes | Bring your own | Yes |
| Mapbox Studio | None | Custom basemap design and tile publishing | Freemium with usage-based pricing | Yes | Built-in catalog | No |
| CARTO | Partial | Enterprise GIS analytics on warehouse data | Enterprise sales, no published self-serve pricing | Limited | Bring your own | No |
Tool by tool.
ShowMeOnMap
AI assistant that turns a plain-English question into a finished, cited map.
ShowMeOnMap takes a natural-language question and selects the right dataset from its built-in catalog of live public data, then renders the result on a MapLibre + deck.gl 3D globe with cartographically correct ramps, legends, and source citations. Pricing is one-time prepaid credits (1 credit per standard query, 5 per agent query); credits never expire. You can also bring your own data — upload CSV, Excel (.xlsx), GeoJSON, TopoJSON, KML, KMZ, GPX, zipped Shapefile (.zip), or paste a list of addresses to geocode — and uploads render free without an account; AI analysis of an upload costs 3 credits. Best fit: you want a finished map from a question, or a quick cited map of your own list, without wiring up data plumbing yourself.
Best for: Data-curious people who want a finished map without learning a GIS tool.
Felt
Collaborative web GIS for drawing, uploading, and styling layers in a browser.
Felt is a collaborative web-based GIS where you upload your own data (GeoJSON, KML, shapefiles, CSV with lat/lon, raster) and style or draw on it with a polished editor. Felt has shipped AI-assisted authoring features, but the primary surface remains a manual layer editor optimized for team collaboration. Best fit: a small team that has its own data and needs to draw, annotate, and share map layers with shared editing.
Best for: Teams who need to draw, annotate, and share map layers together.
Datawrapper
Chart and map publishing tool for newsrooms and analysts.
Datawrapper is a publication tool for charts, choropleth maps, locator maps, and symbol maps. You paste a CSV, choose a map type, configure colors and legends through a polished form-based UI, and embed the result. It does not have a natural-language interface; it earns its reputation on cartographic quality, accessibility, and clean default styling. Best fit: you have a CSV ready and need a publishable chart or map for an article or report.
Best for: Journalists and analysts publishing a chart or map inside an article.
Kepler.gl
Open-source GPU-accelerated geospatial analysis tool from the Uber team.
Kepler.gl is an open-source web application built on deck.gl that visualizes large geospatial datasets — point clouds, trips, hexbins, H3 grids, heatmaps — entirely in the browser using GPU rendering. It has no natural-language interface; you configure each layer through a UI panel. Best fit: you have a large CSV or GeoJSON and you want to explore it interactively on your own machine without sending the data to a third-party service.
Best for: Analysts exploring millions of points or trip records locally.
Mapbox Studio
Visual style editor for vector basemaps and custom map tiles.
Mapbox Studio is a visual editor for designing custom vector basemaps and uploading your own tilesets. You pick a base style, restyle every layer (roads, water, labels, terrain) with a designer UI, and publish a tile URL your app can consume. It is not a tool for generating a map from a question; it is the upstream tool for designing the basemap a map sits on. Best fit: you have an app that already uses Mapbox GL or MapLibre and you want a custom-branded basemap.
Best for: Designers and developers who need a branded basemap for an app.
CARTO
Enterprise spatial analytics platform on top of cloud data warehouses.
CARTO is an enterprise spatial-analytics platform that runs spatial SQL and visualizations directly on top of a cloud data warehouse (Snowflake, BigQuery, Redshift, Databricks). Recent releases include an AI assistant for SQL generation. Best fit: an enterprise team that already has spatial data in a warehouse and needs a governed analytics environment rather than a standalone map tool.
Best for: Enterprises running spatial analysis directly on Snowflake / BigQuery / Redshift.
What about ChatGPT and Claude?
ChatGPT Maps finds places; ShowMeOnMap maps data. OpenAI’s ChatGPT Maps is a place-discovery feature — its own page calls it “a new way to discover places” and offers recommendations, news, and weather for what is around you. It is the right tool for finding somewhere to eat or planning a route through a city. ShowMeOnMap answers a different question: where is something happening, and how much of it — live earthquakes, wildfire perimeters, census tracts, World Bank indicators, transit, or your own spreadsheet — with the source cited on the map.
An assistant can draw a map; it rarely cites one. Ask ChatGPT or Claude for a map and it may draw one, embed one, or write the code for one in the chat. That works for a sketch. What you get depends on that conversation: the assistant may fetch a live dataset, or fill in coordinates and values from its own memory when a fetch fails, and the result usually has no per-layer source and no page you can share. On ShowMeOnMap every coordinate comes from a cited source or a gazetteer lookup and every value from a cited source, never from the model’s memory, and every map has a legend, a source line, and a shareable link.
You can use both. ShowMeOnMap runs inside those assistants too: add it to Claude as a connector, or to ChatGPT or any MCP client, and the assistant builds a cited ShowMeOnMap map instead of a sketch. Connect your assistant.
Which one should I pick?
I want to ask a question and get a finished, cited map without learning a GIS tool.
Pick ShowMeOnMap
I have my own data and need a publishable chart or map inside an article.
Pick Datawrapper
A small team needs to draw, annotate, and edit map layers together.
Pick Felt
I need to explore millions of points or trip records locally on my own machine.
Pick Kepler.gl
I have an app and want a custom-branded basemap.
Pick Mapbox Studio
My enterprise runs spatial analysis on warehouse data (Snowflake / BigQuery / Redshift).
Pick CARTO
I want to find places near me or plan a trip.
Pick ChatGPT Maps
I want a quick sketch map inside a chat and don’t need sources.
Pick Your AI assistant
Frequently asked.
Is ShowMeOnMap the same as ChatGPT Maps?
No. ChatGPT Maps is OpenAI’s place-discovery feature: its own page describes it as “a new way to discover places,” answering questions about what is around you such as place recommendations, news, and weather. ShowMeOnMap maps data rather than places to visit — earthquakes, wildfires, census and World Bank indicators, transit, your own files — fetched live from public sources and cited on every map. Use ChatGPT Maps to find a café or plan a trip; use ShowMeOnMap to see where something is happening and how much of it there is.
Can’t ChatGPT or Claude just make the map for me?
For a quick one-off sketch, often yes: general assistants can draw a map, embed one, or write map code in the conversation. What they produce depends on that one chat — the assistant may fetch a live source or fill in coordinates and values from its own memory, the map usually carries no citation per layer, and there is no saved, shareable page. On ShowMeOnMap every coordinate comes from a cited source or a gazetteer lookup and every value from a cited source, never from the model’s memory; every map has a legend, a source line, and a shareable link. It also works inside those assistants: connect it to Claude, ChatGPT, or any MCP client and the assistant builds a ShowMeOnMap map for you.
Which tool has a true natural-language interface?
As of 2026, ShowMeOnMap is the tool in this comparison whose primary interface is natural language — you type a question and the system selects the data, picks the visualization, and renders the map. Felt and CARTO have shipped AI-assisted authoring features (layer suggestions, SQL generation), but the primary surface for both is a manual layer editor. Datawrapper, Kepler.gl, and Mapbox Studio do not have natural-language interfaces.
Which tool should I pick if I already have my own data?
If you already have your own GeoJSON, CSV, shapefile, or warehouse table, the best fit depends on what you want to do with it. For publication-ready static maps inside an article, pick Datawrapper. For collaborative drawing and editing, pick Felt. For large-scale point or trip exploration, pick Kepler.gl. For enterprise spatial analytics on warehouse data, pick CARTO. To get a quick, shareable map of your own file or address list without learning a GIS tool, ShowMeOnMap takes uploads (CSV, Excel (.xlsx), GeoJSON, TopoJSON, KML, KMZ, GPX, zipped Shapefile (.zip)) and pasted address lists, which it geocodes and cites; its distinctive strength is the opposite case — you describe what you want and it fetches the public data for you.
Which tool is open source?
Kepler.gl is open source (MIT) and self-hostable. MapLibre GL JS, which ShowMeOnMap and most modern web mapping tools build on, is also open source. Felt, Datawrapper, Mapbox Studio, CARTO, and ShowMeOnMap are commercial products.
Which tool is cheapest for an individual user?
Kepler.gl is free and self-hostable. Datawrapper and Felt have free tiers suitable for individual use. ShowMeOnMap lets logged-out visitors make up to 3 maps a day free, then sells one-time credit packs from $5 USD for 100 credits with no subscription. Mapbox Studio is free until you exceed the included tile-request quota. CARTO does not publish self-serve pricing and is positioned for enterprise budgets.
Where does each tool fetch its data from?
ShowMeOnMap fetches data live at query time from a built-in catalog of authoritative public datasets, so you do not need to bring data — but you can: it also maps your own uploads (CSV, Excel (.xlsx), GeoJSON, TopoJSON, KML, KMZ, GPX, zipped Shapefile (.zip)) and pasted address lists. Felt, Datawrapper, and Kepler.gl require you to upload or paste your own data. Mapbox Studio combines Mapbox-published vector tiles with your own tilesets. CARTO connects to your own cloud data warehouse.
Disclosure: ShowMeOnMap is the publisher of this comparison. We described our own tool the same way we described the others — same dimensions, same format, no superlatives, no claims about market share or “best in class.” Every competitor description is based on public information from the tool’s own website and documentation as of 2026-05-21; ChatGPT Maps is described from its own page (chatgpt.com/maps) as of 2026-10-08. Pricing structures and AI feature sets change frequently; check each vendor’s site for the current state before making a decision.
Found an inaccuracy? Open the composer and tell us — corrections welcome.