Felo Search Teardown — Japan's AI Search Insurgent Reaches 1.2M Users Without Touching English-First Markets

$10M Series A. 300K daily queries. 34% Japan traffic dominance. Built in 12 months.

Let me tell you what makes Felo interesting, and it is not the multilingual search angle you keep reading about in press releases.

The interesting part is that a Tokyo startup — founded in July 2024, so barely a year old at Series A — found a seam in the AI search market that Perplexity, with 900× more monthly traffic, had no incentive to close. That seam is not "non-English speakers want AI search." That's true but trivially obvious. The seam is that knowledge workers in Japan, Korea, and Taiwan operate in a trilingual information environment daily — they need Japanese regulatory filings, Korean VC announcements, and English academic papers synthesized in a single research session — and every English-first AI search tool forces them to context-switch constantly.

Felo Search was built to collapse that context-switch. Whether they pull it off at scale is a different question. But the playbook they used to get from zero to $10M of institutional backing in 12 months is genuinely worth dissecting.

The Origin Story: Sparticle Pivots Into Consumer Search

First, the corporate structure, because it matters for understanding how Felo moved fast with limited capital.

Felo is technically the consumer-facing brand of Sparticle Inc., a Tokyo company co-founded by Jeffery Jin (also known as 金田達也, CEO) and Tang Jinsong. Sparticle existed prior to Felo — it was doing generative AI tooling and raised approximately ¥500 million ($3.5M USD) in a pre-Series A round from Silicon Valley's Wisemont Capital in October 2023. That pre-Series A was for the Sparticle AI platform, not Felo specifically.

Then something changed between late 2023 and mid-2024. The team watched Perplexity's trajectory — $73M ARR run rate by end of 2024, 100M+ monthly users globally — and identified the precise failure mode of every English-first AI search tool in Asian markets: they treat multilingual search as a translation feature bolted on afterward, not as a core retrieval architecture decision.

Sparticle relaunched as Felo Inc. (Ｆｅｌｏ株式会社) in July 2024, with Felo Search as the flagship product. This was not a pivot from scratch. The pre-existing team, engineering infrastructure, and Sparticle's Japanese LLM expertise gave Felo roughly 18 months of runway before they needed to prove product-market fit externally.

This matters for any builder studying the playbook: the capital efficiency story only makes sense because Felo was not starting cold. They had an operating company, a technical team already working on multilingual AI, and $3.5M in the bank from 2023. The "built in 12 months" headline is real but needs that context.

The Growth Timeline: Quarter by Quarter

Q3 2024 (July launch): Felo Search launches publicly. Core features at launch: multilingual search with citation, AI mind map visualization, AI presentation generator (Slides). The team makes a deliberate decision not to compete on the English-language head terms where Perplexity, Kagi, and YouChat already dominate. Instead they push distribution through Japanese tech media, X/Twitter Japanese creator accounts, and line-by-line SEO for Japanese-language queries. The product immediately resonates in Japan — there is a non-trivial community of Japanese researchers and business analysts who are genuinely underserved by English-first tools.

Q4 2024 (October–December): Word of mouth in Japan accelerates. ProductHunt engagement and coverage by Japanese tech outlets (ITmedia, TechCrunch Japan) drive early user registrations. The Chrome extension launches, which matters because it plugs Felo into the existing browser workflow without requiring users to change their starting URL. Chrome Web Store installs become a meaningful distribution vector for a browser-native audience.

Q1 2025 (January 8, 2025): Felo Search Agent launches — this is the critical product expansion. The team reframes the product from "AI search" to "Search 3.0": the agent autonomously plans research tasks, pulls from multiple language sources, and generates deliverables (reports, mind maps, presentations) without the user managing the workflow step by step. This is Felo's answer to the "Perplexity with Deep Research" problem. They are not trying to beat Perplexity on English-language recall quality. They are making the output layer — the thing you do with search results — dramatically better for Asian knowledge workers.

Q2 2025 (April–June): Enterprise product roadmap announced. "Felo Enterprise" targets corporate analysts, IR teams, and research functions that need cross-market intelligence synthesis across Japanese, Korean, and English sources simultaneously. The specific use case being pushed is financial research: Japanese institutional investors tracking Korean equities, Korean VCs monitoring Japanese M&A, Taiwan-based businesses navigating mainland China regulatory filings. These are real workflows with real budget owners.

July 2025: Series A closes. ¥1.5 billion ($10M USD) led by Peak XV Partners (formerly Sequoia Capital India & Southeast Asia) and Mirae Asset Venture Investment. This investor combination is telling. Peak XV's domain is South and Southeast Asia; Mirae Asset is a major Korean financial institution with deep cross-Asia distribution networks. Neither of these investors would back a Japan-only play. They are betting on a pan-Asian multilingual AI search layer, not a local Japanese tool.

By Series A close: 1.2 million registered users, 300,000+ daily queries, Japan at 34% of traffic, Taiwan at 17.65%, South Korea at 5.76%.

Current state (May 2026): felo.ai is clocking approximately 270,000 monthly active visitors (SimilarWeb/Semrush estimates — note the distinction between registered users and monthly active). Traffic has grown 20%+ month-on-month recently, with an unusually large jump in paid search spend (+7,456% from near-zero baseline, suggesting the team is now paying to acquire users at scale post-Series A). Japan leads AI tool adoption globally with 214% annual growth per SimilarWeb 2025 data, so Felo is surfing a genuine macro tailwind.

The Product Wedge: What Felo Actually Built

Let's separate the real technical differentiation from the marketing narrative.

Real differentiation: multilingual semantic retrieval architecture. Felo's claim is that you can query in English and receive answers synthesized from Japanese, Chinese, Korean, Spanish, and 40+ other language sources simultaneously, without manually switching languages. Whether this is powered by cross-lingual embeddings, real-time machine translation at the retrieval layer, or some combination is not publicly disclosed. The team references "RPA technology" for scraping social platforms (Reddit, X/Twitter) in real time, which is a different system from the semantic search engine. The honest read is that Felo has built a Japanese-language semantic search engine that is genuinely better than Perplexity at surfacing Japanese-language sources, layered on top of standard LLM synthesis, and they have wrapped it in a multilingual UX that makes the experience feel seamless for non-English queries.

Marketing claim, examine carefully: "revolutionary multilingual AI." The "multilingual" framing in Felo's press materials is heavily worked. The actual quality claim most credibly holds for Japanese, Chinese (Simplified and Traditional), and Korean. The "40+ languages" number appears in marketing copy but independent reviewers note that quality drops significantly outside the core CJK plus English cluster. This is important for anyone considering this as an indie wedge: the addressable high-quality-retrieval market is East Asian + English, not "the whole non-English world."

Genuinely strong: the output layer. Where Felo clearly outcompetes Perplexity for its target user is in what happens after the search. The mind map visualization of research results, one-click AI slide generation, and structured research report export are native features, not add-ons. For a Japanese business analyst who needs to walk into a meeting with a formatted competitive landscape document, Felo's pipeline — search → structure → present — compresses what is normally a 3-hour workflow into 20 minutes. This is the actual value proposition Felo has discovered, and the Search Agent launch in January 2025 doubled down on it correctly.

The Perplexity gap Felo occupies: Perplexity's product decisions over 2024-2025 consistently prioritized English-language research quality, Pro features for Western enterprise (enterprise SSO, Slack integration, API access), and US media distribution. Their Asia go-to-market is an afterthought. Felo did not try to build a better Perplexity. They built a Perplexity-shaped product with a completely different distribution engine (Japanese media, Asian VC networks, East Asian creator communities) and a product-market fit tuned for trilingual knowledge workers that Perplexity's product team has no organizational incentive to serve.

The Playbook: Three Replicable Moves

Move 1: Locale-specific SEO before product maturity.

Felo's traffic distribution — 34% Japan, 17% Taiwan, 5% Korea — was not accidental. They aggressively indexed for Japanese-language search terms (AI検索, AI検索エンジン, 多言語検索 etc.) from day one. Most English-first AI tools completely neglect non-English SEO, which means there is almost no competition for high-intent Japanese queries about AI productivity. This is a pattern any builder targeting a non-English market can replicate immediately: the content and SEO gap in non-English AI tool discovery is enormous, and the conversion rate from Japanese/Korean/Chinese organic search is high because these users have been underserved for so long.

Move 2: The PPT/mind map hook as viral acquisition.

Felo's slide generator and mind map feature are not just value-adds — they are inherently shareable. When a user generates a presentation from a Felo research session and shares it with colleagues, those colleagues see the Felo branding on the output. This is a classic productivity tool virality loop: the artifact travels further than the user. Notion did this with database templates. Canva did it with designs. Felo is doing it with AI-generated research presentations, and it is working particularly well in Japan and Korea where formal presentation culture in business settings creates high demand for polished slide decks.

Move 3: Anchor on a market moment (Japan's AI surge).

Japan's AI adoption grew 214% in 2025, the fastest of any major market globally. Felo's timing was not entirely intentional — they launched in July 2024 when the Japan AI wave was already building — but their product was perfectly positioned to ride it. The lesson for builders: a tool positioned for an underserved market + a macro tailwind is a rare combination. Felo is benefiting from both. The compounding effect is significant: organic Japanese media coverage, government attention to domestic AI champions, and enterprise buyers looking for Japan-headquartered AI vendors (for data compliance and contractual comfort reasons) all created distribution channels that a US-headquartered competitor cannot easily replicate.

Investor Logic: Why Peak XV and Mirae Asset Said Yes

The $10M Series A is not a bet on Felo being a better Perplexity globally. The investor thesis is narrower and more credible: Felo can become the default AI research and search layer for East Asian knowledge workers, an addressable market of hundreds of millions of people who are systematically underserved by English-first AI infrastructure.

Peak XV's portfolio expertise in emerging market distribution (Zomato, Meesho, CRED) suggests they see Felo as a similar story: a product that has no real distribution or cultural competition from US incumbents, and that benefits from a local brand advantage that is hard to replicate from the outside. Mirae Asset brings Korean financial sector distribution, which is directly relevant to Felo's enterprise product targeting cross-Asia financial intelligence workflows.

The capital deployment plan — enterprise product buildout, Korea and Taiwan expansion, R&D in autonomous agent technology — is coherent. The risk is whether the enterprise market actually closes at the price points Felo needs, and whether Perplexity or Google eventually decides the Asian market is worth a localized product investment.

What Could Break the Story

Three failure modes are worth naming honestly:

1. Google's inevitable move. Google's dominance in Japanese search (roughly 80% market share in Japan, versus Google's ~90% global share) means Google understands the Japanese search user better than anyone. When Google fully deploys its AI Overviews (formerly SGE) with Japanese-language optimization, the casual Japanese AI search user will encounter it in their existing Google workflow. Felo's defense is the power-user research workflow — the mind map, agent synthesis, enterprise report generation — but if Google's AI Overviews absorb the "quick multilingual answer" use case, Felo's free-tier user base gets squeezed.

2. The "multilingual" claim eroding under scrutiny. Felo's brand is built on multilingual quality. If independent benchmarking emerges showing that Felo's Japanese/Korean/Chinese retrieval quality is only marginally better than Perplexity with a translation layer, the differentiation story collapses for sophisticated users. The current favorable reviews are from early adopters who are naturally biased toward the product that speaks their language first. Enterprise buyers will run their own evaluations.

3. Free-to-pro conversion remains opaque. With 1.2M registered users and 300K daily queries, Felo's free tier is clearly working. The Pro plan at $14.99/month (or ¥2,099/month) gives 300 professional searches per day versus 5 on free. The question is what fraction of those 1.2M users are paying. My estimate: under 10K paying subscribers based on the MRR tier classification, which implies under $150K monthly recurring revenue. That is a thin enterprise base for a company now carrying $10M in capital and Series A expectations. The July 2025 shift from "usage count" to a unified "credits" system signals the team is working through monetization architecture challenges.

The Indie Wedge Verdict

If you are a solo builder or small team watching Felo, the replicable pattern is narrower than it looks. The surface-level lesson is "build AI search for non-English markets." The actual lesson is more specific: find a vertical where the research-to-deliverable workflow is broken for a linguistically specific professional audience, and build an integrated pipeline that replaces the entire workflow, not just the search step.

Felo could have built a multilingual answer engine and stopped there. They would have been a Perplexity clone with slightly better Japanese recall. Instead they built the downstream output layer — PPT, mind map, report — which created the viral artifact, the enterprise value proposition, and the workflow lock-in simultaneously.

The indie wedge version of this is highly domain-specific: think legal research for Japanese/Korean firms with cross-border practice, or financial compliance research for Taiwan-incorporated companies operating in Japan. The multilingual retrieval architecture is complex, but the "search → structured output" pipeline for a narrow professional domain is buildable with modern LLM APIs and considerably less capital than Felo raised.

On the Mar 2026 Google core update exposure: Felo is fairly well defended. Their content is primarily user-generated research sessions, not bulk-produced article content. Their SEO appears to target tool discovery queries rather than information queries, which is less exposed to Google's Information Gain filter. The risk is more on the product side (Google AI Overviews) than the SEO side.

My verdict: Felo found a real wedge, executed the early growth playbook correctly, and raised institutional money with the right investors for Asian distribution. The hard part starts now — converting a large free user base into enterprise contracts at the ARR levels that justify a $10M Series A in a market where Google is not standing still. The playbook got them here. What gets them to Series B is a different skill set entirely.

---

Felo Search was founded July 2024 by Sparticle Inc. (CEO: Jeffery Jin). Series A: ¥1.5B ($10M USD), July 2025, led by Peak XV Partners and Mirae Asset Venture Investment. 1.2M registered users, 300K daily queries, 34% Japan traffic share. Pro plan: $14.99/month. Primary markets: Japan, Taiwan, South Korea.