*This is the seventh instalment of our series examining AI apps and services making waves in South Korea.*

Background

Liner began life as a web-highlighter: a tool that let users bookmark online content and mark up passages for later review. That modest ambition has since given way to something considerably more expansive. The Seoul-based start-up, led by chief executive Kim Jin-woo, now bills itself as an AI research partner used by more than ten million students and academics across 50 countries. Earlier this year, Fast Company ranked it second in the education category of its 2026 World's Most Innovative Companies list.

What it does

Liner's animating principle is straightforward: no answer without a source. Whatever the query, the system scours the web and academic literature, then returns a response accompanied by the original references that underpin it. Users can verify the evidence themselves, compare perspectives across multiple sources, and reach their own conclusions about what to trust.

The service also offers purpose-built modes. Liner Scholar is designed for researchers who need to navigate complex academic contexts; Liner Lite targets corporate users and aims to consolidate fragmented document workflows into a single interface.

Beyond search, Liner offers a range of practical tools. Upload a PDF or document and it will distil the key points without requiring the user to read the whole thing; users can also pose direct questions about the document's contents. Images can be uploaded for analysis, or for translation of embedded text. The platform has a dedicated investment-research function that analyses the reasons behind a stock's price movements and presents the findings in charts and comparison tables. The company claims its proprietary benchmark tests recorded an accuracy rate of 95.3% on OpenAI's SimpleQA metric—putting it ahead, it says, of ChatGPT's advanced models and Perplexity.

Liner is also pushing beyond search into what it calls "action-oriented agentic AI." A dedicated AI-transformation (AX) division now offers enterprises a full-service package covering workflow diagnosis, proof-of-concept development, agent deployment, and staff training. The company has also launched the Liner Model API, which it claims cuts large language model (LLM) token costs by more than 50% without sacrificing response quality. Its ambitions are not confined to South Korea: Liner already supplies its search engine to Humane One, the workplace AI platform of Humain, Saudi Arabia's state-backed AI company, signalling a growing footprint in the Middle East.

Strengths and limitations

The case for Liner is clearest among students and researchers. Users with an academic email address (ending in .ac.kr or .edu) receive a free month of the Pro plan, and even on the free tier, advanced search and academic mode are available without restriction. The source-citation structure is a genuine advantage for research tasks that demand verification.

The weaknesses, however, are real. Some users report that recent versions feel less capable than builds from a year or two ago. The mobile experience draws particular criticism: the context window—the volume of information the model can process in a single session—is reportedly too small to handle both a PDF upload and a detailed prompt simultaneously. There is also a more structural concern: critics warn against the service drifting towards becoming a mere "wrapper," stitching together outputs from various third-party LLMs without developing meaningful proprietary capability of its own. The free plan carries advertising, and the deep-research function is subject to credit limits once a user's allocation runs out.

Verdict

Liner is one of the few South Korean AI services to have expanded from a domestic base into global research markets and, more recently, into enterprise AI infrastructure in the Middle East. Its commitment to source-based answers is a sound strategy for building credibility as a research tool. But the more fundamental test—whether the model doing the reading is itself reliably good—remains the variable that will determine how competitive Liner can stay.

★★★☆☆ (3.5/5.0)

In a sentence "The citations are impeccable; the AI doing the citing still has something to prove."