About
We believe the results of what you do should shape what you decide next.
TRAIL Labs builds AI marketing software that helps brands reach the people who need them, and make better decisions from what their marketing actually does.We run three products, TRAIL Search, TRAIL Studio, and TRAIL Perform, and each works on its own.Founded in May 2026 in Seongdong-gu, Seoul, by researcher turned founder Fred Kim.
Why we started
Two ways of deciding: the lab and the field
I spent five years as a researcher at Seoul National University's graduate school of AI and completed my PhD coursework. Form a hypothesis, check the result, revise the earlier choice: that was simply how work got done.
Three years as CTO at marketing and ad agencies showed me a different way. There were rules of thumb about which ads a brand of a certain size should run, but why that choice fit that brand, and how the results would be judged, often went unexplained.
The change we want for brands and consumers
The work it takes for a brand to get noticed keeps growing. Content and ad creative for every platform, channels to run, a budget to split. The smaller a brand's budget and team, the heavier that load is before it ever reaches a customer.
Consumers should find what they need more easily, and brands should meet the customers who get what they offer. We build products that make that connection with less effort, and that let what you learn from doing accumulate into the next decision.

Fred Kim
Founder, TRAIL Labs
What we base our work on
We measure and make recommendations on three layers of evidence, not gut feel. We don't put a number on results until they're in.
| Layer | What it is | How the products use it |
|---|---|---|
| Measurement rules | Only questions that don't name the brand go into the share-of-voice denominator; unmeasured is kept separate from 0%; and a period where the question set changed is flagged as a composition change. | Every TRAIL Search result carries the engine, the question, and the time it was checked. The same 0% gets different recommendations for 'published but not showing up' and 'not published yet'. |
| Public research | Published papers such as the original GEO paper (KDD 2024), C-SEO Bench (NeurIPS 2025), SAGEO Arena, and FeatGEO (ACL 2026), plus a practical methodology that synthesizes them. | This is the basis for how recommendations are ordered and weighted. Where papers disagree, we note the conditions, and we cite paper figures only as results of the papers' own experiments. |
| Our own research | We are submitting our own method as a full paper to an international peer-reviewed conference in information retrieval. | We don't quote effect sizes until results are in, and we don't name the venue or the paper before acceptance. Our research on video knowledge indexing is at the planning stage. |
This site gets a regular TRAIL Search audit, and we fix it by the same rules.
Mission · Vision · Concept · Promise
TRAIL Labs runs on these four. How we work day to day is written up separately.
Mission
To bring what makes a brand unique to the people who need it, by easing the burden of marketing execution and taking the guesswork out of marketing decisions.
Vision
A world where any brand can reach its customers, whatever its budget or team size, and gets better at marketing the more it does.
Concept
Marketing software that connects discovery, content, and advertising, and feeds every result back into the next decision.
Our promise to customers
We explain why an action is worth taking, what to watch, and what to change based on the result.
The company and its products
TRAIL Labs is the company. TRAIL Search, TRAIL Studio, and TRAIL Perform are SaaS products run by TRAIL Labs. TRAIL Perform is the product for multi-channel ad operations, performance analysis, and budget allocation and execution.
Each product works on its own. You sign up for each separately; there is no shared account or combined billing. Each product answers a different question. Start with whichever one is urgent right now.
Why Now
Three shifts. One window.
Tech
The cost of AI generation dropped 10x in 24 months. SaaS built for small and mid-sized businesses is now viable.
Market
Single-purpose generation tools are saturating. We think the next decade belongs to products where discovery, production, and performance inform each other.
Policy
Korea's policies on SME digital transformation, K-commerce exports, and the AI ecosystem all point to the same gap.
The three shifts don't happen in sequence. They multiply.
Traction & Roadmap
What we've shipped. What's next.
Proof of execution
3
Products in operation: Search · Studio · Perform
1
MCP server that AI agents call directly
Naver
Engines measured (as implemented): ChatGPT · Claude · Perplexity · Gemini · Google AIO · Naver AI Briefing
2026-10
Our own method is under submission to an international peer-reviewed conference
TRAIL Studio MVP complete · TRAIL Search MCP server live
What's next
Now
Search · Studio live
Signup and billing connected
Next
Connecting the products
A path from discovery to execution to improvement
Policy alignment
Frequently asked questions
Three questions we hear most about the company.
Who built this?
It’s a one-person company. Is that sustainable?
Are accounts or data shared between the products?
Questions about products, pricing, and data are in the FAQ.
Let's talk
Whether it's about using a product, partnering, or investing, send us an email. We'll share the IR deck on request.
Last updated