Mapmaking in the AI age: How TomTom builds maps you can trust

It’s safe to say that we’re in the middle of a renAIssance. Over the past couple of years, with generative and agentic AI becoming mainstream, businesses around the globe have been in a race to integrate it into both their workflow and their output. But like any other new phenomenon, AI faces questions on the ethical implications of its widespread use.

Many of these questions revolve around whether AI should be used at all. At the same time, when used correctly, this technology can save lives — whether it’s by making autonomous vehicles safer or by making it easier to detect medical conditions. In areas where AI use is proving beneficial for society, its rapid adoption still raises important ethical considerations around transparency, privacy and accuracy.  
 
A closer look will reveal that many of these ethical concerns stem from how, and on what data, the AI models are trained. To ensure the quality and reliability of the model’s output, it’s important that this data is collected responsibly and transparently.  

Most widely used AI models have largely been trained by scraping the web and harvesting publicly available data, from news articles to social media posts. While this data is public, it might comprise sensitive or copyrighted information. On the surface, the resulting LLM-powered chatbots and image generators might seem almost magical in how naturally they can create information that previously didn’t seem to exist. But a critical eye will find that these creations are often AI slop — a mashed together byproduct of several people’s original, potentially unlicensed work, riddled with errors and inconsistencies. 

Having integrated AI into its mapmaking process years ago, TomTom takes great care to ensure every piece of data used in its map has a clear and trusted origin. Even its open-source contributions are carefully managed through intellectual property (IP) configuration tools to distinguish between commercial and open-license data.  
 
“At TomTom, we engineer location intelligence with precision and purpose. The datasets powering Orbis Maps are sourced through rigorous licensing frameworks, enriched by sensor-derived observations and validated through multi-layered quality analytics. Our platform doesn’t just reflect the world — it models it with integrity,” says Manuela Locarno Ajayi, TomTom’s SVP of Product Engineering. 
 
Putting the AI in mapmaking 
 
Even though the world is just waking up to AI, this technology has been around for much longer, especially in the realm of mapmaking. While digital mapmaking traditionally required humans to interpret data from satellite images and real-world observations, and then draw map features like roads and footpaths manually, AI — specifically Machine Learning (ML) and Computer Vision (CV) — has unlocked a new era of mapmaking, capable of processing mountains of data and turning observations into map edits, updates and features quickly and accurately.  

Mapmaking company TomTom was a frontrunner in transitioning the world from traditional, static maps to interactive, updatable maps. Powered by ML, CV and fueled by a wide variety of data, from sensors to satellite imagery and floating car data (FCD), TomTom has been able to create maps that update in real-time and reflect our rapidly changing world more closely than ever before. 
 
For a company built on data, the quality and accuracy of that data, and the maps made from it, have always been of utmost importance, even without the concerns raised by AI use. 
 
What it takes to make a good map 
 
This motivated TomTom to create Orbis Maps, which combines open data, proprietary data and advanced AI-driven validation for a wide range of commercial applications.  
 
“Data,” Michael Harrell, SVP of Software Engineering TomTom Maps, has said in the past. “It’s all about the data.” 
 
Today, digital maps are being used in more business models and applications than ever — from geotagging social media posts to routing food delivery drivers to calculating accurate insurance to finding the right location for urban infrastructure. To serve these use cases and more, developers and engineers need the best, most detailed map possible, which in turn requires a massive amount of data.  
 
Thankfully, there’s no shortage of location data in the world. Everything happens somewhere, and every change adds a new morsel of data to the vast amount of already existing location data. For TomTom, the challenge lies in sorting through all this and making sure only high-quality data is ingested into its map. The company does this by keeping its sources diverse and purposeful. 
 
Sources for Orbis Maps come in a variety of forms, including observations from automotive OEMs, vehicles, connected sensors, partners and open-data projects, like OpenStreetMap (OSM) and the Overture Maps Foundation, which TomTom co-founded.  
 
One of the most important of these are sensor-derived observations (SDOs), which are collected from sensors on vehicles, such as cameras, rain sensors, light sensors and forms of radar. This data is meticulously compared against TomTom’s 30 years of historical data and mobile mapping (MoMa) vehicle data to verify if road information is new, if it’s changed or even if it disappears. 
 
Other sources include points of interest (POIs) from a curated set of high-quality providers, addresses acquired country-by-country to ensure local accuracy, high-quality satellite imagery collected from legitimate partners and more.  

Data for TomTom's new map flows from various resources. Maintaing relationships with carmakers, tech companies and open-source communities uniquely positions TomTom to assimilate all this data.
Maintaining (data) integrity 
 
TomTom doesn’t subscribe to the 'grab-first-ask-questions-later' mentality of data scraping, nor does it use APIs without consent.  

With its maps powering over 600 million connected devices, TomTom has no shortage of high-quality data of trusted and traceable origin. The company’s various sources have verified provenance, ensuring that licensing terms and intellectual property rights are always respected. 
 
For example, the data collected from vehicles is anonymized to protect privacy. While TomTom knows there’s a vehicle driving down a certain street at a given time, the company has no idea what vehicle that is, where it starts or stops or where it’s going. 

Once collected, all the data is then put through transparent curation and validation processes, so users can trust its accuracy, consistency and fairness. And we can trust that AI models are getting the best input, to ensure the best output. 
 
Having built its data practices with responsibility and for compliance, TomTom adheres to frameworks like GDPR while working proactively on aligning with emerging AI regulations such as the EU AI Act. By continuously evaluating its systems and maintaining transparency, TomTom ensures its technology evolves responsibly — maintaining trust and setting a high standard for fair data use in AI. 
 
Tackling hallucinations: TomTom’s AI mindset 
 
A big concern with AI use is when AI hallucinates, or perceives patterns or objects that don't exist in reality — creating nonsensical, inaccurate outputs. If these hallucinations find their way onto widely-used maps, the consequences could be catastrophic. 
 
TomTom doesn’t take this responsibility lightly. Before it makes it to the map, all data undergoes rigorous regression checks to catch and correct errors before they reach customers. With the sheer amount of data that TomTom has, automating the processing of this bundle of rich source material can naturally speed things up. But when done negligently, this can also harm the reliability of the result. That’s why TomTom builds quality metrics into every stage of development.  
 
A team of data scientists and machine learning engineers labor to improve the algorithm and keep it in check, ensuring no errors make it onto the map. This collaboration between human and machine allows the mapmaker to get the most accurate, high-quality result in the least amount of time. Internal teams use Analytics-as-a-Service to make data-informed decisions. Data democratization is also prioritized to make information accessible and transparent across the company, enabling collective responsibility for quality.   

This carefully-designed approach helps TomTom employ AI as it’s meant to be used — as a magnifier that turns human strengths into superpowers, making them far more efficient and accurate mapmakers. 

The result is Orbis Maps — an open, collaborative map that stands out for its freshness, accuracy and reliability. 

A global, accurate, rich map 

Built thoughtfully by combining machine output and human oversight, Orbis Maps offers global reach and strategic value for customers and partners across industries, be it governments and public sector, automotive or insurtech.  

The map covers over 235 countries and territories, capturing over 1.5 billion buildings and more than 90 million kilometers of roads worldwide. These impressive numbers are made possible only by a hybrid ecosystem of proprietary, open and community data — far richer and more diverse than a model built on a single source, or several unvalidated sources. 

Mapmaking is by no means easy work, which explains why there are so few global mapmakers. Doing it well is even more difficult. But TomTom has been around the block and seen pretty much everything there is to see. The AI boom might be shaking up the world now, but TomTom had already put guardrails in place years in advance, when it started incorporating ML and CV into its processes. That’s what sets it apart from corporations rushing to win the AI race, sacrificing fairness and morals in the process. 
 
“In a time when others compromise on provenance, we build with accountability, ensuring our maps and services are not only accurate but ethically sound. That’s the foundation for trustworthy navigation and autonomous systems,” says Manuela. 

For TomTom, the concerns of sourcing data ethically, maintaining fair practices and ensuring the result is rich and accurate aren’t new. As a true mapmaker, the company is committed to preserving accuracy and truth — the foundational values of mapmaking. They’re baked into the company’s DNA, reflecting in everything it does and will continue to do.