--- title: JudgeGPT Experiment Boosts Case Closures in Pakistan Courts description: A large-scale field trial in Pakistan tested JudgeGPT, a custom GPT-4-based tool combining retrieval-augmented generation with a database of nearly 130,000... canonical: https://daily.dev/posts/judgegpt-experiment-boosts-case-closures-in-pakistan-courts-itarczxdc twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: JudgeGPT Experiment Boosts Case Closures in Pakistan Courts | daily.dev og:description: A large-scale field trial in Pakistan tested JudgeGPT, a custom GPT-4-based tool combining retrieval-augmented generation with a database of nearly 130,000... og:url: https://daily.dev/posts/judgegpt-experiment-boosts-case-closures-in-pakistan-courts-itarczxdc og:image: https://api.daily.dev/og/posts/iTArCzxdc.png og:image:alt: JudgeGPT Experiment Boosts Case Closures in Pakistan Courts og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # JudgeGPT Experiment Boosts Case Closures in Pakistan Courts **[IEEE Spectrum](https://daily.dev/sources/ieeespectrum)** · 6 min read · 0 upvotes · 0 comments ## Summary A large-scale field trial in Pakistan tested JudgeGPT, a custom GPT-4-based tool combining retrieval-augmented generation with a database of nearly 130,000 Pakistani judicial opinions and 943 statutes, to help judges with legal research and drafting. Offered to 1,559 trial judges starting in 2024, the tool boosted case resolutions by 6.3% in trained districts with no drop in judgement quality, and appeal rates fell slightly. Training proved critical: judges who received six 90-minute sessions on LLM limitations and hallucination risks used the tool far more persistently (56 logins, 212 prompts) than those given generic training (10 logins, 25 prompts) or none at all. Economists estimate $38.50 saved in judicial costs per dollar spent. However, about a fifth of prompts showed signs of substantial AI delegation, and researchers didn't report hallucination rates, raising questions about the quality of justice versus efficiency. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Did using an AI tool actually help judges resolve more court cases in a real study? Yes, a nationwide trial in Pakistan found trained judges using JudgeGPT, a GPT-4 tool built with retrieval-augmented generation over 128,292 judicial opinions and 943 statutes, resolved 6.3% more cases with no measurable drop in judgement quality. Appeal rates also fell slightly. Trained judges resolved 38.5 more cases per month than baseline, an estimated $38.50 saved in judicial costs per dollar spent on the tool. _Anyone evaluating AI tools for high-stakes professional work can track results like these on daily.dev._ ### Why did training matter so much for judges adopting an AI legal research tool? Training determined whether judges kept using the tool at all. Judges who completed six 90-minute sessions covering how LLMs work, bias, hallucination risks, and output verification logged in 56 times and sent 212 prompts on average, versus just 10 logins and 25 prompts for those given only generic tech training, and untrained judges typically abandoned the tool after about a month. _Teams rolling out AI assistants internally follow adoption patterns like this on daily.dev._ ### How does JudgeGPT prevent AI hallucinations in legal research? JudgeGPT uses retrieval-augmented generation, attaching GPT-4 to a searchable database of 128,292 Pakistani judicial opinions and 943 statutes so responses include footnoted links to actual cases and laws rather than relying on the model's unverified memory. Co-author Elliott Ash notes the fix for hallucinations isn't more intelligent models but connecting models to a verifiable search tool, though exact hallucination rates were not reported. _Developers building RAG pipelines for high-stakes domains compare grounding strategies on daily.dev._ ## Similar posts on daily.dev - [GPT-5 bests human judges in legal smack down](https://daily.dev/posts/gpt-5-bests-human-judges-in-legal-smack-down-vdupltziq) · The Register · 1 upvotes · 0 comments - [South African lawyers learn hard lesson in AI fiction](https://daily.dev/posts/south-african-lawyers-learn-hard-lesson-in-ai-fiction-r6xvbfy4r) · TechCentral · 0 upvotes · 0 comments - [A Better Way To Build LLM-as-a-Judge Pipelines](https://daily.dev/posts/a-better-way-to-build-llm-as-a-judge-pipelines-59gkauzir) · Daily Dose of Data Science \| Avi Chawla \| Substack · 0 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#rag](https://daily.dev/tags/rag), [#gpt](https://daily.dev/tags/gpt) [View this post on daily.dev](https://daily.dev/posts/judgegpt-experiment-boosts-case-closures-in-pakistan-courts-itarczxdc) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"JudgeGPT Experiment Boosts Case Closures in Pakistan Courts","url":"https://daily.dev/posts/judgegpt-experiment-boosts-case-closures-in-pakistan-courts-itarczxdc","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/judgegpt-experiment-boosts-case-closures-in-pakistan-courts-itarczxdc"},"datePublished":"2026-08-12T11:03:25.462Z","dateModified":"2026-08-12T11:03:52.998Z","description":"A large-scale field trial in Pakistan tested JudgeGPT, a custom GPT-4-based tool combining retrieval-augmented generation with a database of nearly 130,000...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/b9d6246f69b0124fee84f174e63f2f05?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/b9d6246f69b0124fee84f174e63f2f05?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"IEEE Spectrum","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"IEEE Spectrum","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/b123f5e8420e492ab36f6adb30c1793a","url":"https://daily.dev/sources/ieeespectrum"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/judgegpt-experiment-boosts-case-closures-in-pakistan-courts-itarczxdc","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,rag,gpt","timeRequired":"PT6M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"IEEE Spectrum","item":"https://daily.dev/sources/ieeespectrum"},{"@type":"ListItem","position":3,"name":"JudgeGPT Experiment Boosts Case Closures in Pakistan Courts"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/judgegpt-experiment-boosts-case-closures-in-pakistan-courts-itarczxdc#faq","mainEntity":[{"@type":"Question","name":"Did using an AI tool actually help judges resolve more court cases in a real study?","acceptedAnswer":{"@type":"Answer","text":"Yes, a nationwide trial in Pakistan found trained judges using JudgeGPT, a GPT-4 tool built with retrieval-augmented generation over 128,292 judicial opinions and 943 statutes, resolved 6.3% more cases with no measurable drop in judgement quality. Appeal rates also fell slightly. Trained judges resolved 38.5 more cases per month than baseline, an estimated $38.50 saved in judicial costs per dollar spent on the tool. Anyone evaluating AI tools for high-stakes professional work can track results like these on daily.dev."}},{"@type":"Question","name":"Why did training matter so much for judges adopting an AI legal research tool?","acceptedAnswer":{"@type":"Answer","text":"Training determined whether judges kept using the tool at all. Judges who completed six 90-minute sessions covering how LLMs work, bias, hallucination risks, and output verification logged in 56 times and sent 212 prompts on average, versus just 10 logins and 25 prompts for those given only generic tech training, and untrained judges typically abandoned the tool after about a month. Teams rolling out AI assistants internally follow adoption patterns like this on daily.dev."}},{"@type":"Question","name":"How does JudgeGPT prevent AI hallucinations in legal research?","acceptedAnswer":{"@type":"Answer","text":"JudgeGPT uses retrieval-augmented generation, attaching GPT-4 to a searchable database of 128,292 Pakistani judicial opinions and 943 statutes so responses include footnoted links to actual cases and laws rather than relying on the model's unverified memory. Co-author Elliott Ash notes the fix for hallucinations isn't more intelligent models but connecting models to a verifiable search tool, though exact hallucination rates were not reported. Developers building RAG pipelines for high-stakes domains compare grounding strategies on daily.dev."}}]} ```