Digital Ghosts: How Ukraine Uses AI to Predict Russian Movements
- Matthew Parish
- May 31
- 6 min read

In the fog of war victory often belongs not to the side with more tanks, but to the side with better data. In Ukraine, where Russian forces have spent three years trying to overwhelm cities, borders and civilians, the battlefield is increasingly shaped not just by drones and artillery — but by algorithms.
Welcome to the world of predictive warfare, where Ukraine’s defence increasingly relies up on artificial intelligence (AI) to anticipate Russian troop movements, artillery strikes and cyberattacks. Through a fusion of satellite imagery, open-source intelligence, "deep learning" (a form of enhanced machine learning based on neural networks, in which progressively more complex patterns are derived from large data sets) and combat feedback, Ukraine is building what soldiers now call “digital ghosts” — anticipatory models that help them see the enemy before it arrives.
AI at the Front: Seeing What Isn’t There — Yet
One of the defining features of AI-driven defence is its ability to synthesise enormous amounts of disparate data and forecast what human analysts might miss. Ukraine’s military, working with local tech startups and Western partners, has developed AI systems that:
Process satellite imagery to identify new troop build-ups or camouflage patterns.
Analyse social media posts and Telegram traffic to geolocate Russian units.
Use pattern recognition from previous attacks to anticipate the timing and target of future missile strikes.
Map vehicle movement using thermal and drone imagery, even in fog or under tree cover.
Some systems are semi-autonomous, flagging anomalies in movement — a sudden spike in fuel convoys, for instance — that may precede an offensive. Others run simulations using battlefield reports, weather data, and topographical inputs to suggest the most likely path of attack.
“We are creating models that learn from every day of the war,” said an official from Ukraine’s Ministry of Digital Transformation. “The system remembers what we survive — and then helps prevent it.”
Ghosts in the Code: Predicting Artillery and Drone Attacks
Perhaps the most transformative application of AI has been in counter-battery warfare — predicting the location of Russian artillery based on acoustic data, drone visuals and heat signatures.
Before a human analyst can even triangulate a Russian firing position, AI tools like GIS Arta (see postscript below) — a Ukrainian-developed artillery intelligence platform — can propose coordinates for a return strike. The result is faster, more accurate fire with fewer rounds.
In drone warfare, algorithms have been trained to analyse incoming FPV (first-person-view) drone attack patterns, recognise flight signatures and predict which unit might be targeted next. These tools alert frontline units and sometimes even pre-programme evasive manoeuvres into Ukrainian drone formations.
Powered by Data — And Volunteers
Unlike traditional militaries, Ukraine’s AI advantage stems in part from its civilian technology base. Much of the data used to train predictive models comes from open sources — TikTok videos, satellite providers like Planet Labs and Telegram groups monitored by civilian volunteers.
In a Kyiv basement, data scientists once employed in advertising technology now fine-tune machine learning models to predict troop movements. Meanwhile battlefield data is relayed from front-line units via encrypted platforms, reviewed by analysts, and then piped into the growing neural networks used in targeting, logistics and defence planning.
“Our code is written by people who have skin in the game,” says one developer in Lviv. “Every line is about keeping someone alive tomorrow.”
Cyber Defence and AI: Anticipating Digital Assaults
AI’s battlefield use isn’t limited to physical movement. Ukraine also uses machine learning in cyber defence, detecting phishing campaigns, malware signatures and infrastructure attacks.
In partnership with NATO and private companies, Ukraine’s security agencies use AI-driven tools to:
Monitor for behavioural anomalies across networks.
Flag spear-phishing attempts on civil servants or military personnel.
Simulate enemy tactics to test vulnerabilities in critical infrastructure.
These digital systems — trained on years of Russian hybrid warfare patterns — now form part of a predictive firewall that anticipates not just brute force attacks, but psychological and digital incursions.
Challenges and Limitations
Despite successes, predictive warfare is not flawless. Russian forces have grown more cautious about leaving digital footprints. Satellite coverage is sometimes delayed by cloud cover or prioritisation. AI models require constant retraining as battlefield dynamics evolve.
And as Ukraine’s systems grow more effective, so too do Russia’s efforts to deceive them — using decoys, signal spoofing, and fake data trails to mislead algorithms.
Moreover AI remains a tool, not a guarantee. Its value lies in guiding decisions — not making them in isolation. Ukrainian commanders have stressed that predictive models augment human judgment, but do not replace it.
Toward a Smarter War — and a Smarter Peace
Ukraine’s use of AI is not just about staying alive today. It’s about learning at speed. Every engagement teaches the algorithms. Every strike and survival loops back into the system, strengthening tomorrow’s defences.
In the long term Ukraine’s innovations may reshape not only how wars are fought, but how peace is maintained — through data-aware border monitoring, real-time disinformation tracking, and autonomous emergency response to future threats.
As one Ukrainian officer put it:
“Russia fights with weight. We fight with speed and memory.”
The ghosts that Ukraine’s data conjures are not spectres of defeat — they are early warnings, spectral maps of the next move. And they may yet prove to be the most potent weapons of all.
---
What is GIS Arta?
GIS Arta (Geographic Information System for Artillery) is a Ukrainian-developed command and control platformthat enables real-time targeting, coordination and artillery strike optimisation. In essence, it functions as a digital fire control system, integrating multiple battlefield inputs to deliver fast and accurate artillery responses.
Though often compared to NATO-standard systems like AFATDS (Advanced Field Artillery Tactical Data System), GIS Arta is lighter, faster and tailored to Ukraine’s warfighting environment — making it ideal for a conflict where rapid mobility and decentralized response are essential.
Key Capabilities of GIS Arta
1. Real-Time Target Acquisition
GIS Arta fuses data from a range of sources to identify and geolocate Russian assets, including:
UAV (drone) feeds
Forward observer reports via mobile apps
Thermal and infrared imagery
Signals intelligence
Acoustic sensors
Public and commercial satellite imagery
This fusion allows users rapidly to pinpoint Russian positions and generate precise coordinates — often within minutes of sighting.
2. Automated Fire Mission Generation
Once a target is confirmed, GIS Arta can:
Calculate firing solutions based on gun type, shell, and distance
Account for terrain, weather and counter-battery risks
Automatically assign the most suitable artillery unit nearby
Relay firing data directly to gunners via secure channels
Loop time from detection to fire can be as short as 1–2 minutes — a critical edge in counter-battery duels.
3. Distributed Access and Frontline Use
GIS Arta is designed for modular use. Soldiers can access the platform from:
Tablets or "ruggedised" laptops
Mobile phones (in stripped-down versions)
Battlefield command posts
Drones acting as data relays
This makes the system decentralised, so even platoon-level operators can participate in fire coordination, reducing the burden on top-down command.
4. Integration with Civilian Tech
One of GIS Arta’s unique strengths is its ability to interface with commercial technologies, including:
Starlink terminals for battlefield internet
DJI drones and other consumer UAVs
Crowdsourced apps (e.g. military spotters using smartphone GPS tagging)
This allows Ukraine to run high-tech operations without a traditional military-industrial backbone.
How AI Supports GIS Arta
While GIS Arta is not purely an AI system, it relies increasingly on machine learning for:
Anomaly detection in incoming drone or vehicle movement
Target prioritisation (e.g. recognizing tanks versus trucks in thermal scans)
Predictive location modeling, especially for mobile rocket systems or artillery batteries
Recent updates (as of late 2024) have included AI-supported terrain analysis — helping predict likely enemy movement paths based on previous patterns.
Impact on the Battlefield
GIS Arta is credited with:
Dramatically improving counter-battery accuracy — enabling rapid suppression of Russian artillery.
Reducing shell wastage — with fewer rounds needed per target.
Enhancing coordination between artillery and drone teams — creating a true “sensor-to-shooter” chain.
Allowing smaller artillery units to operate autonomously — especially valuable under conditions of communications jamming.
A Ukrainian artillery officer told Radio Liberty:
“With GIS Arta, we no longer wait for orders. We observe, we confirm, we fire — in minutes.”
Origin and Evolution
Originally developed by a team of Ukrainian software engineers and military advisors after 2014, GIS Arta has been under constant iteration since 2022.
The system is not public domain but is widely deployed across the Armed Forces, often supported by local IT volunteers and battlefield developers.
Unlike NATO systems that require extensive training, GIS Arta is designed to be intuitive and usable by minimally trained personnel under pressure.
Why It Matters Globally
GIS Arta shows how a non-NATO military — with limited resources — can use smart software and open-source tools to match or even exceed the capabilities of larger, more traditional armies.
It has inspired international interest, including among countries facing asymmetric threats. GIS Arta is now seen as a model for agile military digitization.
Final Note
GIS Arta is not a “magic bullet.” It depends on stable connectivity, good inputs, and human judgment. But in a war where speed, dispersion and precision matter more than mass, it represents one of Ukraine’s most effective force multipliers.