AI may widen the gap between sustainability promises and action

A new systematic review warns that the impact of artificial intelligence (AI) on sustainability depends less on the technology itself and more on whether it changes real behaviour. Published in Sustainability, the study finds that AI can help individuals and organisations act on sustainability commitments, but it can also strengthen symbolic compliance, polished reporting and superficial environmental claims without producing deeper change.

The research paper, titled Behavioural Sustainability and Artificial Intelligence: A Multi-Level Systematic Review of the Intention–Behaviour Gap and Decoupling, reviews 48 papers on AI, behavioural sustainability, Environmental, Social and Governance performance, the intention-behaviour gap and sustainability decoupling, concluding that AI is not automatically a force for sustainable development. Instead, it operates as a conditional mechanism whose effects vary across individual, organisational, value chain and system-level contexts.

AI's sustainability promise faces a behavioural test

The study addresses one of the major problems in sustainability policy and business practice: the gap between what people, organisations and institutions say they intend to do and what they actually do. This intention-behaviour gap appears across green purchasing, energy use, transport, tourism, sustainable fashion, food choices and recycling. The review finds that AI can reduce this gap in some settings by making sustainable choices easier, more visible and more personalised. But it can also deepen the problem when it improves measurement, disclosure and optimisation without changing underlying behaviour.

The author frames the issue against the backdrop of rising interest in AI-driven sustainability initiatives, ESG measurement and progress toward the United Nations Sustainable Development Goals (SDGs). AI systems are now used in energy management, carbon tracking, smart homes, sustainable consumption tools, precision agriculture, environmental monitoring, sustainable marketing and corporate reporting. These technologies can process large datasets, identify patterns, automate decisions and provide real-time feedback. That makes them attractive to policymakers, businesses and consumers searching for faster routes to sustainability.

However, sustainability challenges are not only technical. They are also behavioural and institutional. Climate change, biodiversity loss, resource depletion and social inequality cannot be solved simply by adding better data systems if individuals, companies and public institutions do not change their practices. The study says the real question is whether AI helps translate sustainability commitments into sustained action, or whether it merely makes sustainability claims easier to monitor, package and promote.

The strongest evidence of AI's positive role appears at the individual level. AI-enabled tools can support sustainable behaviour by providing personalised feedback, decision support, nudges and real-time monitoring. Carbon footprint apps, recommendation systems, search tools, smart home assistants and sustainability profiling systems can help users understand the impact of their choices and act more consistently with their stated values. In domains linked to responsible consumption and climate action, AI can reduce information overload and help consumers move from intention to action.

The review points to several behavioural mechanisms. AI can influence attitudes by making environmental consequences more visible. It can shape perceived social norms by comparing user behaviour with peers or wider benchmarks. It can increase perceived behavioural control by making sustainable options easier to identify or select. In practical terms, this means AI can help people choose greener products, reduce household energy use, shift travel behaviour, support recycling and make more informed consumption decisions.

Notably, the study makes clear that these effects are not guaranteed. Some AI systems can weaken green decision-making. Recommender systems, for example, may reduce autonomous motivation if users feel choices are being pushed by an algorithm. AI-generated information can also create confusion or uncertainty, especially when sustainability claims are complex or inconsistent. If users do not trust the system, lack digital literacy or face structural barriers such as price, access or convenience, AI-generated prompts may fail to produce sustained behavioural change.

The review also finds that much of the available evidence is based on surveys, reported intentions or short-term behaviour rather than long-term observation. This limits the strength of claims that AI produces durable sustainability outcomes. The field, according to the study, is still stronger in showing that AI can influence attitudes and intentions than in proving that it delivers lasting behaviour change across time and contexts.

From green behaviour to greenwashing risk

While AI can improve workplace sustainability, reporting systems and ESG monitoring, it can also help organisations look greener without becoming greener. This is where the paper's focus on sustainability decoupling becomes central. Decoupling occurs when formal sustainability commitments, policies or reports are not matched by substantive changes in practice.

In business and institutional settings, AI is often used to improve disclosure, monitor performance, optimise operations and generate sustainability metrics. These tools may help organisations track emissions, resource use, supply chain performance and workplace practices. Used well, they can improve accountability and expose gaps between stated goals and actual operations. But the review finds that AI can also support selective disclosure, symbolic compliance and more sophisticated forms of greenwashing.

The risk is that organisations may use AI to enhance the visibility of sustainability claims rather than transform the behaviours behind them. Better dashboards, reports and ESG analytics can create an impression of progress even when business models, procurement practices, labour arrangements or resource use patterns remain largely unchanged. The study argues that improved measurement is not the same as meaningful change.

Investors, regulators and customers increasingly rely on sustainability data to assess organisational performance. AI can make that data more detailed, timely and persuasive. Yet if the underlying metrics are selective or if organisations optimise only what is easiest to report, AI may strengthen the appearance of accountability while leaving deeper environmental and social harms intact.

The review also warns about optimisation without transformation. AI can make energy systems, transport networks, logistics, buildings and production processes more efficient. But efficiency gains can be offset by rebound effects, where lower costs or improved performance lead to higher overall consumption. In such cases, AI may reduce waste per unit of activity while increasing total activity, weakening the net sustainability benefit.

The same risk appears in smart cities and infrastructure management. AI can optimise traffic, waste systems, buildings and energy use, but technical efficiency does not automatically change the behaviour of residents, companies or institutions. Without governance, incentives and public engagement, AI may make unsustainable systems operate more smoothly instead of helping society shift away from unsustainable patterns.

The study also identifies value chain and system-level blind spots. Much of the research focuses on individuals, while fewer studies examine how AI shapes behaviour across supply chains, institutions or entire socio-technical systems. This matters because sustainable development requires change across multiple levels. A consumer may receive AI-supported advice to buy more responsibly, but if supply chains remain opaque, corporate incentives favour volume growth and regulations are weak, individual-level improvements may remain limited.

The review uses an integrated theoretical framing to explain this complexity. The Theory of Planned Behaviour helps explain how AI affects attitudes, norms and perceived control. Socio-technical systems thinking shows that behaviour is shaped by infrastructure, regulation, markets and institutions. Institutional theory explains why organisations may adopt sustainability policies or AI tools for legitimacy without changing core practices. Together, these lenses show why AI's sustainability role is conditional, contested and deeply dependent on context.

Real progress depends on governance, trust and long-term change

The study proposes a multi-level model of AI-mediated behavioural sustainability, placing AI between sustainability intentions and actual outcomes. In this model, AI operates through decision support, nudging, automation, monitoring and accountability systems. These mechanisms can produce behavioural alignment when they help translate intentions into action. However, they can also produce sustainability decoupling when they support symbolic, superficial or reporting-driven outcomes.

The model identifies several conditions that determine which outcome is more likely.

• At the individual level, digital literacy, trust, environmental awareness and motivation shape whether people act on AI-generated information.
• At the organisational level, leadership commitment, incentive structures and integration into core strategy determine whether AI changes routines or simply improves reporting.
• At the technological level, transparency, usability and explainability affect whether users understand and trust AI systems.
• At the institutional level, regulation, governance quality and social norms shape whether AI is used for accountability or reputation management.

The study finds that AI's strongest alignment with the SDGs appears around SDG 12 on responsible consumption and production and SDG 13 on climate action. This is because many AI interventions are already focused on carbon tracking, consumption, waste reduction, energy efficiency and pro-environmental behaviour. The review also connects AI to SDG 9 on industry, innovation and infrastructure and SDG 11 on sustainable cities and communities, particularly through smart cities, infrastructure systems and organisational sustainability tools.

Additionally, the study warns that SDG alignment cannot be inferred from AI adoption alone. A company, city or institution may deploy AI in the name of sustainability without producing real behavioural or institutional change. AI contributes to sustainable development only when it helps close the gap between commitment and practice.

The review also raises concerns about geography and inequality. Evidence is drawn from several regions, including Asia, Europe, Africa and global studies, but many findings remain concentrated in better-studied or higher-capacity contexts. AI systems deployed in lower-income countries or infrastructure-constrained settings may face different barriers, including weak regulation, lower digital access and limited institutional capacity. The study says future research must treat geography as a key factor rather than assuming AI-enabled sustainability tools work the same everywhere.

The temporal dimension is another limitation. Most reviewed studies are recent, with the field expanding rapidly between 2020 and 2025. That means many findings capture a fast-moving moment rather than stable long-term outcomes. The study calls for more longitudinal research to test whether AI-supported behaviour change persists or fades once novelty, incentives or attention decline.

Policymakers should not treat AI as a shortcut to sustainability. Regulations and governance systems should require evidence that AI tools improve actual behaviour and institutional practice, not just reporting quality. Organisations should be asked to demonstrate operational change, not merely more detailed sustainability disclosure. Consumers and employees should be supported with transparent, usable tools that reduce real barriers to sustainable action.

For businesses, AI-enabled ESG systems may improve measurement, but if they are not tied to changes in incentives, procurement, operations and accountability, they may expose companies to accusations of more advanced greenwashing. The more powerful the reporting tool, the greater the expectation that reported progress reflects real change.