Parul Dhull

Parul Dhull

Auto-generated profile, based on publicly available information

Parul Dhull

Department of Computer Science, College of Dairy Science and Technology, Sri Karan Narendra Agriculture University, India

Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
10 September 2026 in Current Agriculture Research Journal
Verified
Journal Article
Open access
SSSURESH KUMAR SHARMA
SURESH KUMAR SHARMA
PDParul Dhull
Parul Dhull
Agricultural extension in most developing countries suffers from a persistent ‘last-mile’ problem: too few trained personnel, uneven coverage of remote and arid regions, and advice that rarely matches the specific agro-climatic conditions of an individual farm. Generative artificial intelligence (GenAI), and in particular large language model (LLM)-based conversational chatbots, has recently been put forward as a scalable complement to human extension networks. Drawing on a structured narrative synthesis of 30 peer-reviewed, pre-print, and institutional sources identified through systematic database and grey-literature searches, this review synthesises the fast-growing literature on GenAI and chatbot-based advisory in agriculture, tracing its evolution from rule-based expert systems and SMS/IVR advisory to retrieval-augmented, multilingual, voice-first conversational agents. It examines the enabling technologies — large language models, retrieval-augmented generation, and multimodal, multilingual natural language processing — and reviews global and Indian deployments, including Digital Green’s Farmer.Chat, IFPRI’s Generative AI for Agriculture (GAIA) initiative, and India’s newly launched Bharat-VISTAAR platform, inaugurated in Jaipur, Rajasthan, in February 2026. Evidence on farmer trust, adoption, and socio-technical barriers is critically evaluated, with particular attention to the semi-arid, drought- and locust-prone agro-ecology of Rajasthan and to unresolved challenges around low-resource Indic and dialectal language coverage, data sovereignty, and human-oversight requirements. Concretely, the reviewed evidence shows automatic-speech-recognition error rates for Indian languages climbing sharply outside Hindi, multilingual large language models measurably more prone to factual error in low-resource Indic languages than in English, and a 2026 producer survey in which just under half of respondents report weekly use of general-purpose AI tools while only about a quarter say they trust its agricultural recommendations without independently cross-checking them; a large-scale randomised evaluation of non-GenAI digital advisory in Odisha, by contrast, found only modest average yield gains that were markedly larger for farmers facing weather shocks — a reminder that platform scale and demonstrated impact do not automatically track one another. The review identifies persistent gaps — thin longitudinal impact evidence, negligible coverage of regional dialects such as Marwari and Mewari, unresolved liability questions for erroneous advice, and weak integration with local Krishi Vigyan Kendra (KVK) networks — and sets out a research and implementation agenda for context-specific, human-in-the-loop GenAI advisory suited to arid and semi-arid smallholder agriculture.