Who's using ArcGIS AI capabilities in their Research?
The title above is actually not a question to you, but if you'd like to get in contact and tell us (highered@esri.com) about your research that'd be great 🙂 The aim here is to highlight a small sample of the great work out there, from researchers who are applying ArcGIS AI capabilities!
AI and ArcGIS AI inside ArcGIS
Our Esri Education team is fortunate to be able to support and work with about 11,000 universities around the world, in addition to thousands more schools, museums, libraries, and other informal education or capacity development organizations. We are always on the lookout for educators who are applying ArcGIS in novel manners and we are constantly surprised at what is being done out there. Who's teaching what? That is not so easy for us to follow, because although everyone likes open science, open data, etc. not everyone shares their curriculum materials. We'd love to hear from educators before or while they are creating new learning materials rather than bump into them years afterward.
Publish or Perish
On the other hand, Research uses of ArcGIS, and specifically AI tools within ArcGIS, are much easier to learn about. Researchers publish! If it's not published, it didn't happen. You may have had experiences where researchers can't provide research details because they are embargoed pending an article's publication. Thanks to Google Scholar (and other online tools) we can easily have a look at what has been published.
Try this at home.
Take a quick look at ArcGIS + AI, and filter on Since 2025 and you will find roughly 16000 publications, including a few misclassifications but most are quite relevant.
In my case (your list may vary) the first publication is a review of a recent Esri Press book on the subject 🙂 Then there are several review articles but as I scrolled down I found an interesting article applying ArcGIS AI technology to Greek archaeology: "A Deep Learning Method for the Automated Mapping of Archaeological Structures from Geospatial Data: ..."
Fylaktos and colleagues were searching for ancient structures on the island of Delos, so they used Maxar imagery and applied the Mask R-CNN deep learning model within ArcGIS Pro. That model is described in this and other ArcGIS documentation.
The Greek researchers were happy with the model accuracy (93%) and recommend the procedure for replication by other researchers...which is a central goal of science: standing on the shoulders of others, not reinventing the wheel.
Road Engineering
Ok, so that was an archaeology use case. I found another paper where road engineering researchers (at University of Maryland) state in their final Discussion section: "There is a growing need for landslide risk monitoring for areas that have not been identified before. When integrated into GIS platforms such as ArcGIS Pro 3.2.0, the digital twin model can become an active component of transportation management systems, providing ongoing updates and risk mitigation strategies for state and federal transportation agencies. The toolbox in ArcGIS Pro can be customized to meet the specific requirements of users. Machine learning packages can now be embedded in ArcGIS Pro 3.2.0 [38] to train ML models and make predictions tailored to users’ needs." (You can often tell when the actual research was done by the version of ArcGIS they report.)
In that paper, "AI-Powered Digital Twin Technology for Highway System Slope Stability Risk Monitoring", the researchers used a whole host of AI techniques integrated into ArcGIS and obtained new information that either they could not have found or in much less time.
These researchers chose their own combination of analysis tools and AI methodologies: there's not just one way to do things. What's important to note is that they, and an increasing number of researchers, are choosing to integrate their AI, modeling, and related research tasks inside or alongside ArcGIS. Not AI and ArcGIS, but rather AI inside ArcGIS.
Geocoders
We could go on and on from the corpus of thousands of papers using AI with ArcGIS, but here's one last example that is quite different: geographic text analysis. ArcGIS is not synonymous with Pro, of course, and has many other facets within the system including a global geocoder that can be accessed from various software clients across the web. In the article entitled "Toponym resolution leveraging lightweight and open-source large language models and geo-knowledge" researchers in Germany tested several LLMs (back in 2023) for their ability to disambiguate placenames/toponyms. They instructed the LLMs to access and study data from GeoNames, Nominatim, and the ArcGIS geocoder.
These researchers applied the low-rank adaption LoRA model and, interestingly, now 3 years later, you can combine the SAM (segment anything model) released by Meta, with LoRA to issue natural language queries asking the combined model to find and extract features in an image. This is shown in a free online tutorial that I recommend you try out.
Search on...
Give it a try and search Google Scholar for ArcGIS + your favorite research terms. You might be surprised about what people are doing out there, and maybe you'll reach out to authors and even make a few new friends.
Please keep an eye out for the creative integrations of AI models and frameworks, within the ArcGIS workflow. That's where the real power is, and where research, government, and industry specialists are finding new efficiency and productivity. And again, if you include the "ArcGIS" keyword or tag in your articles we'll more easily find and promote them.