Truace
Subscribe
70/100

Moderate evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

Review of 71 studies finds city AI works in silos but is rarely linked up

Study title: A Meta-Analysis of Artificial Intelligence in the Built Environment: High-Efficacy Silos and Fragmented Ecosystems

Evidence type
Systematic review or meta-analysis
Sample
n = 71
Sources
1 sourcePeer-reviewed
Published
2 min read
Truvace record
TRV-2026-2731cite or check it

Study title: A Meta-Analysis of Artificial Intelligence in the Built Environment: High-Efficacy Silos and Fragmented Ecosystems

A PRISMA-aligned systematic review and meta-analysis covering January 2015 to July 2025 screened 7432 records and kept 71 empirical studies of AI, machine learning, deep learning and IoT in urban infrastructure. Pooled across those studies, within-domain performance and sustainability outcomes improved by a large margin (Hedges' g = 0.92).

Read the study abstract

Cities face mounting pressures to deliver reliable, low-carbon services amid rapid urbanization and budget constraints. Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), and the Internet of Things (IoT) are widely promoted to automate operations and strengthen decision-support across the built environment; however, it remains unclear whether these interventions are both effective and systemically integrated across domains. We conducted a Preferred Reporting Items for Systematic Reviews (PR…

65Gain side

How the two sides scoreClose call

62Problem side

The gain

A meta-analysis of 71 studies of AI, machine learning and IoT in urban infrastructure found a large pooled improvement in within-domain performance and sustainability outcomes (Hedges' g = 0.92), still positive at about 0.68–0.70 after bias adjustment.

The problem

The same meta-analysis of 71 urban infrastructure studies found 91.5% of AI implementations operate at isolated or minimal data-sharing levels and only 1.4% achieve real-time multi-domain integration.

A fixed formula sets both scores from the evidence and its sources. Reader votes don’t move them.

What the scores mean

G is the gain side, out of 100. It goes up when the gain is specific, when it measurably helped people and when the sources behind it are strong.

P is the problem side, out of 100. It goes up when the problem is specific, when it measurably hurt people and when the sources behind it are strong.

In brief

The gains rarely connect. 91.5% of implementations run at integration Levels 0–1, with isolated systems or minimal data sharing, and just 1.4% reach real-time multi-domain integration. The authors say this fragmentation prevents cross-domain synergies and the system-wide value cities want.

Publication bias is likely (Egger's test p = 0.03), and a bias-adjusted estimate puts the effect near g ≈ 0.68–0.70. The authors rate overall certainty of evidence as Moderate under GRADE because of heterogeneity and small-study effects.

Main points

  1. Searches of five open-access indices (MDPI, DOAJ, CORE, BASE, OpenAIRE) returned 7432 records, and 71 studies met the inclusion criteria for quantitative synthesis.
  2. The random-effects pooled effect was Hedges' g = 0.92 (95% CI 0.78–1.06, p < 0.001) for within-domain performance and sustainability outcomes.
  3. Egger's test (p = 0.03) points to likely publication bias; the adjusted estimate is about g ≈ 0.68–0.70, still positive.
  4. 91.5% of implementations sit at integration Levels 0–1, with isolated or minimal data sharing.
  5. Only 1.4% of implementations achieve real-time multi-domain integration (Level 3).
  6. The authors recommend mandating open standards and APIs, setting up city-level data governance, funding Level-2/3 integration pilots and adopting cross-domain evaluation metrics.

What the review found when it combined the studies

These are the review’s own pooled results, copied from its abstract. Truvace did not recalculate them.

  • Within-domain performance and sustainability outcomes of AI interventions in urban infrastructure

    Hedges' g 0.92 (95% CI 0.78 to 1.06)

    71 studies · search to July 2025

    The review’s own words
    A random-effects model shows a large, pooled effect (Hedges’ g = 0.92; 95% CI: 0.78–1.06; p < 0.001) for within-domain performance/sustainability outcomes.

Code checks every number above against the review’s text before it is shown. The short labels are an AI model’s reading of that text, so the review’s own sentence sits under each one. A pooled figure is only as strong as the studies behind it. How Truvace uses pooled results

The full story

The review followed PRISMA and covered empirical AI, ML, DL and IoT interventions in urban infrastructure published between January 2015 and July 2025. It drew on five open-access indices and reduced 7432 records to 71 studies for quantitative synthesis.

A random-effects model produced a large pooled effect of Hedges' g = 0.92 (95% CI 0.78–1.06). The authors call the picture a dual reality: high efficacy in silos but pervasive fragmentation across domains.

To turn local gains into system-wide value, the authors propose mandating open standards and APIs, city-level data governance, funding pilots at integration Levels 2 and 3, and evaluating across domains. They rate certainty of evidence Moderate under GRADE, citing heterogeneity and small-study effects, offset by the magnitude and consistency of benefits.

What this doesn’t fix

Publication bias is likely and the studies are heterogeneous, so the authors rate overall certainty of evidence as only Moderate, with a bias-adjusted effect lower than the raw estimate.

Sources

1 on file
  1. Smart Citiesdoi:10.3390/smartcities8050174
    Peer-reviewed2025-10-15

Talk it over

Four questions to start onThis record found a gain and a problem in the same place. Which matters more to you, and who might answer differently?

Start with a question

Review of 71 studies finds city AI works in silos but is rarely linked uphttps://truvace.com/item/a-meta-analysis-of-artificial-intelligence-in-the-built-environment-high-efficac

  1. This record found a gain and a problem in the same place. Which matters more to you, and who might answer differently?

    The gain A meta-analysis of 71 studies of AI, machine learning and IoT in urban infrastructure found a large pooled improvement in within-domain performance and sustainability outcomes (Hedges' g = 0.92), still positive at about 0.68–0.70 after bias adjustment.

    The problem The same meta-analysis of 71 urban infrastructure studies found 91.5% of AI implementations operate at isolated or minimal data-sharing levels and only 1.4% achieve real-time multi-domain integration.

    Answer in the thread

  2. How do we know? What can this kind of evidence show, and what can’t it?

    The evidence This rests on a systematic review or meta-analysis with a sample of 71, from Smart Cities.

    Answer in the thread

  3. What does it leave open? What else would you want to know before acting on it?

    The record’s own caveat Publication bias is likely and the studies are heterogeneous, so the authors rate overall certainty of evidence as only Moderate, with a bias-adjusted effect lower than the raw estimate.

    Answer in the thread

  4. Have you seen AI used this way where you study, work or live? Did it go the same way?

    Answer in the thread

The questions are the same on every record. The details under them are taken from this record as written.

What readers say

More from Climate