University of Bath researchers have created an AI system that assesses a building’s embodied carbon from simple text descriptions, enabling architects to make more sustainable choices early in the design process without needing detailed specifications.
Researchers at the University of Bath have unveiled an artificial intelligence tool that estimates a building’s embodied carbon from plain-text descriptions, aiming to give designers actionable sustainability insight at the moment when choices have the greatest influence on lifecycle emissions. According to the University of Bath, the system uses machine learning and natural language processing to convert brief notes about materials, dimensions and use into rapid, probabilistic estimates of embodied carbon, and refines those estimates as design descriptions become more detailed.
The team says the tool is intended to guide early-stage decisions rather than produce definitive material inventories. “Maximising sustainability before specifications are locked in.” Professor David Coley, an author on the research, told PBC Today: “Architects designing sustainable buildings often lack the level of detail they need to analyse the materials and construction processes involved in the final build. Traditional tools often address embodied carbon via material mass accounting methods. However, these tools rely on precise material breakdowns, accurate quantities and specialist engineering knowledge. Our tool would allow designers to quickly assess and refine their designs to maximise sustainability long before formal specifications are locked in.”
In trials reported by the university, 43 building professionals used the prototype on live projects. One illustrative case concerned a high-end glass-and-masonry building near Exeter, where iterative use of the model prompted adjustments to glazing, wall build-ups and insulation that reduced predicted embodied carbon without requiring contractors to produce full quantity schedules. The researchers trained the model on a synthetically generated dataset representing roughly 150,000 buildings to overcome the scarcity of public embodied-carbon records, and say the system can be retrained on higher-fidelity, real-world datasets as those become available.
The developers report linguistic robustness: the natural language pipeline identified primary materials such as steel, concrete and timber correctly about 80% of the time, produced stable outputs for varied descriptions of the same asset, and successfully ranked existing buildings by embodied-carbon intensity. Bath’s announcement highlights potential secondary benefits, including automated suggestions to improve daylighting, thermal comfort and acoustics, and an educational role in exposing students to carbon-conscious design without specialist inputs.
Industry reaction has been positive, the university states, with users praising the ease of integration into established workflows. The project team also acknowledges limitations: synthetic training data and the probabilistic nature of predictions mean the tool is a decision-support aid rather than a compliance-level calculator.
Experts working in adjacent fields say the approach aligns with broader trends in applying interpretable machine learning to building performance. According to a study published on ScienceDirect, models that explicitly account for building function, compactness and window-to-wall ratio can reliably predict energy use and cooling demand under changing climate conditions, emphasising the value of including form and facade parameters in early-stage assessments. Research from the University of Waterloo reinforces the point that targeted interventions, such as addressing heat loss at junctions and around windows, can deliver substantial operational energy and comfort gains, suggesting early-stage carbon estimates that flag high-risk assemblies could lead to both embodied- and operational-carbon reductions.
City-scale mapping work reported via EurekAlert demonstrates another complementary capability: models that estimate operational emissions across hundreds of thousands of buildings reveal pronounced spatial variation driven by urban form, building age and socioeconomic patterns. Taken together, these strands imply that rapid, text-driven embodied-carbon estimates could be combined with operational-emissions mapping and targeted retrofit diagnostics to inform more holistic decarbonisation strategies across design, refurbishment and masterplanning activities.
For practitioners and decarbonisation leads, the Bath tool promises a practical mechanism to surface trade-offs and low-regret changes early in the process, when material choices, envelope strategies and massing are still fluid. The university notes further development work will focus on retraining with real project data, improving material-recognition accuracy and exploring integration with existing BIM and lifecycle-assessment workflows. The project therefore represents a step toward making embodied-carbon considerations accessible to teams without specialist analysts, while leaving open the need for higher-fidelity follow-up analysis as projects progress toward specification and procurement.
- https://www.pbctoday.co.uk/news/digital-construction-news/construction-technology-news/new-ai-tool-predicts-building-emissions-from-simple-descriptions/159834/ – Please view link – unable to able to access data
- https://www.bath.ac.uk/announcements/pioneering-ai-tool-predicts-building-emissions-from-simple-text-descriptions – Researchers at the University of Bath have developed an AI tool that predicts the carbon footprint of buildings from simple text descriptions. This tool uses machine learning and natural language processing to transform descriptions of materials, dimensions, and usage into estimates of embodied carbon, providing architects with instant feedback on sustainability at the earliest design stages. As designs evolve, the tool refines its estimates, guiding conversations towards more sustainable building practices. The tool also suggests improvements to environmental conditions, such as increased natural light and enhanced thermal comfort, and could play a role in architectural education by promoting sustainability awareness without requiring advanced knowledge.
- https://www.pbctoday.co.uk/news/digital-construction-news/construction-technology-news/new-ai-tool-predicts-building-emissions-from-simple-descriptions/159834/ – A pioneering AI tool developed by the University of Bath predicts the carbon footprint of buildings from simple text descriptions at the earliest stages of design. The tool uses machine learning and natural language processing to transform descriptions of materials, dimensions, and usage into credible estimates of embodied carbon—the carbon emissions associated with materials and construction throughout the building’s lifecycle. This provides architects with instant feedback on the sustainability of their plans right at the start of the design process, allowing for informed decision-making to maximise sustainability before formal specifications are locked in.
- https://www.sciencedirect.com/science/article/pii/S0360132325008960 – This study presents an interpretable machine learning approach to predict building energy consumption in the context of climate change. By considering various building functions and the impacts of climate change, the model offers accurate predictions of cooling energy demand. The research highlights the importance of including factors such as cooling degree hours, cooling coefficient of performance, lighting gain, building compactness, and window-to-wall ratio in energy predictions. The findings suggest that climate change is expected to have a growing positive impact on cooling demand in Haikou, China, underscoring the need for adaptive building designs.
- https://www.sciencedaily.com/releases/2024/05/240515122854.htm – Researchers at the University of Waterloo have developed a new method to identify major heat loss regions in multi-unit residential buildings, leading to significant energy savings. The team identified 28 major heat loss regions, with the most severe at wall intersections and around windows. By addressing 70% of these regions, a potential energy savings of 25% is expected. This method aims to improve building energy use and comfort by targeting specific areas responsible for heat loss, thereby enhancing energy efficiency and occupant comfort.
- https://www.eurekalert.org/news-releases/1095603 – A model developed by researchers maps building emissions to support fairer climate policies. The model estimates operational carbon emissions of individual buildings at the scale of entire cities, providing insights into how emissions are distributed within urban areas. Applied to data mapping over half a million buildings in five cities—Singapore, Melbourne, New York City (Manhattan), Seattle, and Washington DC—the model explained up to 78% of the variation in emissions. The results revealed significant differences in emissions distribution and identified key factors influencing building energy use, including urban form, planning history, and income levels.
- https://www.bath.ac.uk/announcements/how-to-predict-climate-change-from-the-comfort-of-your-home/ – The Climate Predictor, developed by the University of Bath, is a Python-based program designed to run on a Raspberry Pi, allowing users to predict climate change from the comfort of their homes. While less sophisticated than models used by the Intergovernmental Panel on Climate Change (IPCC), the program enables users to explore the effects of increased carbon dioxide on atmospheric temperatures. Users can adjust the amount of CO2 released over a given period and observe the resulting temperature changes, providing insight into the climate crisis and the impact of human activities on climate change.
Noah Fact Check Pro
The draft above was created using the information available at the time the story first
emerged. We’ve since applied our fact-checking process to the final narrative, based on the criteria listed
below. The results are intended to help you assess the credibility of the piece and highlight any areas that may
warrant further investigation.
Freshness check
Score:
10
Notes:
The article was published on 3 March 2026, and the University of Bath’s press release was issued on 2 March 2026. No earlier publications or recycled content were found. The narrative appears original and timely.
Quotes check
Score:
9
Notes:
Direct quotes from Professor David Coley are consistent across sources. However, the exact wording of the quote in the press release differs slightly from the article, which may indicate paraphrasing. The press release states: “Our tool would allow designers to quickly assess and refine their designs to maximise sustainability long before formal specifications are locked in.” The article reports: “Our tool would allow designers to quickly assess and refine their designs to maximise sustainability long before formal specifications are locked in.” The minor difference in wording does not significantly affect the meaning.
Source reliability
Score:
8
Notes:
The primary source is the University of Bath’s official press release, which is a reputable and authoritative source. The article is published on PBC Today, a niche publication focusing on construction and building services. While PBC Today is a specialised source, it is not as widely recognised as major news organisations. The article appears to be a direct summary of the press release, with minimal additional reporting.
Plausibility check
Score:
9
Notes:
The claims about the AI tool’s capabilities align with current trends in applying machine learning to building design for sustainability. Similar tools, such as the CARE Tool, exist and serve comparable purposes. The reported accuracy of the tool in identifying materials and predicting carbon emissions is plausible given advancements in AI and natural language processing. No contradictory information was found in other reputable sources.
Overall assessment
Verdict (FAIL, OPEN, PASS): PASS
Confidence (LOW, MEDIUM, HIGH): MEDIUM
Summary:
The article presents a timely and original report on the University of Bath’s new AI tool for predicting building emissions. While the primary source is the university’s press release, the information aligns with current trends in sustainable building design and is plausible. However, the lack of independent verification from other reputable sources and the reliance on a single source reduce the overall confidence in the article’s accuracy. Editors should consider seeking additional independent verification before publication.
