> Markdown version of [/videos/1323-martin-o-hanlon-make-llms-make-sense-with-graphrag](https://www.wearedevelopers.com/videos/1323-martin-o-hanlon-make-llms-make-sense-with-graphrag). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Martin O'Hanlon - Make LLMs make sense with GraphRAG

Martin O'Hanlon argues vector embeddings alone cannot stop AI hallucinations. Discover how GraphRAG acts as your LLM's right brain, injecting deterministic truth into unstructured workflows.

- **Speakers:** Martin O'hanlon
- **Event:** WeAreDevelopers LIVE
- **Published:** April 15, 2025
- **Duration:** 24:42
- **URL:** https://www.wearedevelopers.com/videos/1323-martin-o-hanlon-make-llms-make-sense-with-graphrag

## Summary

Large language models excel at creative text generation and conversational reasoning but are highly prone to fabricating information when confronted with domain-specific knowledge gaps. This phenomenon, known as AI hallucination, occurs because models generate responses based on probabilistic historical relationships rather than deterministic logic or private data structures. To build robust and reliable applications, software engineers deploy Retrieval-Augmented Generation (RAG) to inject factual context directly into the prompt stream. While traditional vector embeddings effectively handle fuzzy matching across unstructured text deposits, they frequently falter when systems require precise, fact-based answers tied to specific properties, totals, or entity intersections. Transitioning to GraphRAG bridging architectures solves this specific fallback by treating knowledge graphs as an enterprise-grade truth engine. By mapping data meticulously as interconnected nodes, relational labels, and distinct properties, graph databases like Neo4j provide a missing contextual hemisphere for AI workflows. The AI framework effectively serves as the imaginative "left brain" of reasoning, while the knowledge graph functions as the "right brain" providing exact enrichment and structural context. This dynamic, bidirectional relationship enables models to parse unformatted text into rigid schemas using tools like the LLM Graph Builder, ensuring that downstream systems query specialized internal networks with absolute precision and radically diminished hallucination rates.

**Keywords:** GraphRAG architecture, retrieval-augmented generation, AI hallucination mitigation, knowledge graph extraction, neo4j graph databases, vector embedding limitations, generative AI workflows, entity relationship mapping, LLM context enrichment, node and edge properties, unstructured data parsing, deterministic factual retrieval, graph data modeling, semantic text connections

## Chapters

1. **Hallucinations and factual inaccuracies in generative language models** (00:00) — Generative models frequently fabricate information and present it as undisputed truth when lacking access to private or updated data.
1. **Demonstrating language model hallucinations with custom system prompts** (02:35) — Injecting structured localized data into system prompts constrains conversational outputs and meaningfully reduces model hallucinations.
1. **Enhancing model accuracy with retrieval augmented generation patterns** (07:19) — Pairing language features with robust data retrieval architectures seamlessly grounds generated text in verifiable domain facts.
1. **Structuring connected data using graph database architectural fundamentals** (09:25) — Graph databases optimally organize complex ecosystems using systematically interconnected nodes and explicit relationship properties.
1. **Limitations of vector embeddings for precise factual querying** (13:46) — While vector embeddings streamline fuzzy semantic searches, they consistently struggle to reliably resolve strict arithmetic or fact-oriented constraints.
1. **Integrating knowledge graphs to provide verified application context** (17:48) — Mapping unstructured enterprise information into explicit factual networks empowers language applications to correctly evaluate nuanced logical conditions.
1. **Implementing graph generation and automated contextual querying methodologies** (23:08) — Automated extraction utilities effectively synthesize unstructured text into traversable databases to continuously supply contextual data pipelines.

## Related Moments

- [Enhancing language models with graph retrieval augmented generation](https://www.wearedevelopers.com/videos/1311-graphs-and-rags-everywhere-but-what-are-they-andreas-kollegger-neo4j) (from "Graphs and RAGs Everywhere... But What Are They? - Andreas Kollegger - Neo4j")
- [Introduction to generative AI and knowledge graphs](https://www.wearedevelopers.com/videos/1154-large-language-models-knowledge-graphs) (from "Large Language Models ❤️ Knowledge Graphs")
- [Enhancing language models with retrieval-augmented generation](https://www.wearedevelopers.com/videos/844-enter-the-brave-new-world-of-genai-with-vector-search) (from "Enter the Brave New World of GenAI with Vector Search")
- [Unlocking generative AI capabilities using knowledge graphs](https://www.wearedevelopers.com/videos/100318-context-graphs-for-explainable-decision-aware-ai-agents) (from "Context Graphs for Explainable, Decision-Aware AI Agents")
- [Mitigating artificial intelligence hallucinations with constraints and context](https://www.wearedevelopers.com/videos/994-mastering-ai-driven-problem-solving-in-engineering-with-observability) (from "Mastering AI-Driven Problem Solving in Engineering with Observability")
- [Overcoming language model challenges using retrieval-augmented generation](https://www.wearedevelopers.com/videos/1249-building-blocks-of-rag-from-understanding-to-implementation) (from "Building Blocks of RAG: From Understanding to Implementation")

## Related Articles

- [Graph and AI Trends 2026: Why Is AI Running but Not Yet Delivering?](https://www.wearedevelopers.com/magazine/680-graph-and-ai-trends-2026-why-is-ai-running-but-not-yet-delivering)
- [Everything a Developer Needs to Know About MCP with Neo4j](https://www.wearedevelopers.com/magazine/604-everything-a-developer-needs-to-know-about-mcp-with-neo4j)
- [ChatGPT on AI Hallucinations: Can It Fix Its Own Mistakes?](https://www.wearedevelopers.com/magazine/566-chatgpt-on-ai-hallucinations-can-it-fix-its-own-mistakes)
- [WWC24 Talk - Scott Hanselman - AI: Superhero or Supervillain?](https://www.wearedevelopers.com/magazine/469-wwc24-talk-scott-hanselman-ai-superhero-or-supervillain)

## Related Jobs

- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**
- [AI & Machine Learning Engineer (all genders)](https://www.wearedevelopers.com/jobs/48217-ai-machine-learning-engineer-all-genders) at **msg**
- [Data Scientist](https://www.wearedevelopers.com/jobs/ext/1351648-data-scientist) at **Almedia**
- [Staff Developer Advocate, GitHub Security Lab](https://www.wearedevelopers.com/jobs/ext/1921051-staff-developer-advocate-github-security-lab) at **GitHub**
- [Head of AI Applications](https://www.wearedevelopers.com/jobs/ext/1456210-head-of-ai-applications) at **ZEISS Group**
- [Senior AI Agent Software Engineer (Go, Python) (m/f/x)](https://www.wearedevelopers.com/jobs/48277-senior-ai-agent-software-engineer-go-python-m-f-x) at **Dynatrace**