Every Keenious search returns a set of publications. The search map shows that set as a navigable map. Each publication is a single dot, dots sit close together when the publications are about similar things, and color groups them into research areas. It is the same set of results as the list, shown a different way, so you can see the overall shape of the results before reading them one by one.

The first time you open a map, a short walkthrough introduces it. You can reopen that walkthrough at any time from the ? button at the bottom left corner on the map.

Where a dot sits is decided by meaning. Every title, abstract and publication context is stored as an embedding, a representation of its meaning, and the same embeddings that match your search also place each result on the map. Dots close together are close in meaning, and a dot that sits between two areas is somewhere between them in topic.

The underlying space of meaning has many dimensions, and the map flattens it to two, or three, while keeping things that are close in meaning close on the screen.

Three things about a dot carry information: its color, its size, and its brightness.

The dots are grouped by color into research areas, which are clusters of publications whose meanings sit close together. Each area is given a short name by a language model, chosen to describe what sets that group apart from the rest of your results rather than the query as a whole. The names are generated fresh for each search rather than drawn from a fixed list, so the same publication can appear under a different name in another search. This is the same grouping that appears in the list, and How Search Works describes it in more detail.

Area names appear on the map, and the same areas show as colored chips below it. The chips act as a legend and as filters, and the colors stay consistent across the map, the chips, the result cards, and links in Chat. Selecting an area focuses the map on it, so the other areas dim, the map zooms to fit, and the list narrows to those publications. Selecting the area again clears the focus. A search builds up to fifteen areas, and more of them appear for larger result sets.

You move around the map the way you would any map. Drag to pan, and zoom with a scroll, a trackpad pinch, the plus and minus buttons, or a double-click. As you zoom in, the dots begin to show their titles. Hovering over a dot brings up a small card with its rank, year, citation count, title, authors, venue, and the start of its abstract, and clicking a dot opens the publication. Papers you have already opened keep a dashed ring, so you can see at a glance what you have looked at.

When the map sits above the list, it stays still until you click into it, so scrolling the page does not get caught in it. A hint reading "Click to zoom & pan" tells you when this is the case, and clicking outside the map releases it again. A Reset control returns you to the full view and clears any area focus. You can hide the map to read the list on its own and show it again from the results bar, or expand it to fill the screen.

The map and the list stay in step. Whatever is visible in the map viewport is what the list shows, so panning or zooming the map narrows the list to match, with a count that reads "Viewing N out of M". Hovering a dot highlights its row in the list, and hovering a row rings its dot on the map. Selecting a research area filters both at once. To see everything again, reset the view or clear the selected area.

The map is flat by default. The View control switches it to 3D, where you drag to orbit and hold Shift, or use the right mouse button, to pan. The 3D map is the flat layout with a depth axis added, so the research areas and the positions are the same in both. It is the same map seen from another angle.

By default, color on the map means research area. The Display options menu can color the dots by something else instead, and each choice comes with its own legend:

Changing the coloring affects only how the dots look. The layout and the research areas stay as they were. Where the results carry no value for the attribute you pick, the map shows "no data".

The AI in Chat searches the same way. When you ask it something, it runs a Keenious search on your behalf, the same search described here, and every search comes with a map. The results the AI finds are laid out on a map you can open and explore like any other, and you can hide the map whenever you would rather only read.

One thing to keep in mind: the AI does not know which part of the map you are looking at. Panning, zooming, or focusing a research area changes your own view, but it does not tell the AI where you are. If you want it to work with a particular research area or a particular set of papers, name them in your message.

Download as PNG saves the current view of the map as an image, with your query as the title. Every search also has a permanent link, and opening it later rebuilds the same map from the same results.

Does the map show all of my results? Yes. Every publication in the result set has a dot. The set has a fixed size, 300 by default and adjustable up to 1,000, so choosing a larger size puts more publications on the map. See The Result Set.

Why did the areas or their names change when I searched again? The areas are worked out from the results, so changing the query, the filters, or the result size recomputes them. The names can also come out differently between two runs of the same search.

What do the dot size and brightness mean? Size is how often a publication has been cited, and brightness is how relevant it is to your search. See Reading the Dots.

Is the 3D map different from the 2D one? No. It has the same results, the same research areas, and the same positions, with a depth axis added.

Can I search by clicking on the map? Clicking focuses and filters the results you already have. It does not start a new search. To change what appears on the map, change the search.