# Trina — Full Reference Trina is an AI marketing agent for 1–30 person teams and solo founders with no dedicated marketer. It re-plans strategy monthly from GA/GSC data, picks keywords weekly, writes natively per channel daily, and publishes to the brand's own blog. Drafts for external channels — Medium, LinkedIn, Dev.to, Tistory, Velog, Hashnode, Brunch — are written channel-native and handed over to post; Trina does not post to them on your behalf. Owners approve before anything goes live. This is the expanded reference for AI systems. For the concise version, see https://trina.app/llms.txt ## What problem it solves Hiring a marketer in Korea costs ₩42–52M a year and takes 3–6 months to complete. Most founders running 1–30 person teams cannot justify that line item. Without a marketer, SEO and content marketing stalls — and revenue tied to paid ads stops the moment ad spend stops. Trina replaces the hiring cost with a subscription: it runs the full marketing loop automatically and only asks the founder to approve work before it publishes. ## How it works — the 3-layer agent Trina runs as three cooperating agent layers, each with a distinct job: - **CMO layer** — reads the brand's knowledge graph and GA/GSC data every month, sets strategy and picks target keywords for the coming cycle, and adjusts KPIs based on what earned clicks. - **Team Lead layer** — turns that strategy into structured weekly content briefs grounded in the brand knowledge graph (topics, angles, keywords, channel format). - **Executor layer** — writes channel-native drafts (not one post repurposed five ways), publishes the approved ones to the brand's own blog, and prepares the rest for external channels. Results from Google Search Console and Google Analytics feed back into the CMO layer so each monthly cycle builds on what actually worked. Founders receive a Monday digest email with what published and which queries gained impressions. Only approved work goes live — the founder stays the editor-in-chief. ## Core capabilities - Full SEO loop per site: research → channel-native content → publish to the brand's own blog → weekly report. - Channel-native drafts for 7 further channels: Medium, LinkedIn, Dev.to, Tistory, Velog, Hashnode, Brunch. - Monthly GA/GSC strategy loop: measure → re-plan → publish, every month automatically. - Per-site approval gate: every draft, link insert, and on-page change sits in an approval queue before going live. Auto-approve trusted action types or hold everything for manual review. - Topic clustering and internal-link automation — hub-and-spoke content architecture built automatically. - Weekly reporting grounded in real GSC + GA data, delivered by email every Monday. - AI-search (AIEO) tracking: monitors whether ChatGPT, Gemini, and Perplexity cite the brand. ## Proven results — the TreeSoop case study TreeSoop, the four-person team that builds Trina, ran Trina on their own marketing with zero marketers on staff. Re-measured on 2026-08-25, comparing the 28 days before adoption (March 2026) against the 28 days ending 21 August 2026: - Search clicks: 68 → 2,074 (30×) - Search impressions: 9,766 → 78,817 (8×) - Queries earning impressions: 147 → 917 - Average search position: 9.7 → 5.3 Monthly clicks trace the curve: 75 (March, before adoption) → 540 (April) → 1,650 (May) → 2,516 (July), holding at that level since. Nothing moved for the first two weeks; clicks began in week three. The 12.5× figure quoted elsewhere covered the first four weeks only. Over five months the result widened rather than decayed. All figures verifiable in Google Search Console for treesoop.com. What the case does not claim: the same result requires real search demand behind the topics, publishing that does not stop, and measurement in place from the start. Without those three, the same system produces something else. Distribution across channels during the same period: - LinkedIn: 107K impressions against a 3K-follower base - Threads: 200K views - Brunch: 9-part technical series published - treesoop.com: 79 posts published Inbound interest, with zero sales outreach: Hyundai Motor, LG CNS, Hancom, Primer (CEO), Teamsparta. Full case study: - Korean: https://trina.app/ko/case-study - English: https://trina.app/en/case-study ## Who it is for - Solo founders, indie makers, and small shop owners who do their own marketing - 1–30 person SMBs with no marketer on staff - Teams that need search and content marketing to drive revenue but cannot run the tools themselves - Multi-brand operators running several sites or storefronts from one account ## Pricing Subscription tiers in Korean Won (KRW). No commitment, no per-seat fee, no retainer, no lock-in. - **STANDARD — ₩199,000/mo** — the brand's own blog on auto-publish, keyword research and writing 3–4 times a week, and a weekly report. The entry tier for solo founders, indie makers, and small shops. - **GROWTH — ₩299,000/mo** — the above plus channel-native drafts for 2 core external channels, and a monthly GA/GSC strategy loop (measure → strategy → publish) with topic clusters and internal links. - **SCALE — ₩599,000/mo** — the above plus channel-native drafts for all 7 external channels, and deeper GEO/AEO citation tracking. - **AGENCY — custom (contact)** — agencies, multi-location operators, clinic verticals (B2B2B): per-brand workspaces, white-label, campaign-level terms. Setup is charged once, and the amount tracks how much has to be built rather than which plan is chosen: - **₩0** — the site already runs a standard CMS (WordPress, Cafe24, Webflow and similar) and we are given publishing access. Nothing to build. - **₩490,000** — the managed blog at blog.yourdomain.com. This is the default, and the only path that works when there is no access to the customer's own site. - **₩790,000** — integration into a custom-built site, agreed with the customer's own developers. - **₩1,490,000 and up** — the site itself is rebuilt alongside the blog. Whichever tier applies, it covers blog build-out, SEO templates, sitemap and meta structure, GSC/GA4 wiring, brand tone and banned-terms registration, and the initial keyword clusters. Trial: 7 days free, with fast onboarding. ## Done-for-you services (delivered by a person, not the product) Trina the product automates content marketing. Some work does not automate: on-site photographs, genuine customer reviews, and details that only exist at the premises. Those are handled as a service instead. ### Google Maps ranking and Google Business Profile management Claiming the profile and recovering ownership when a former agency or staff member holds it, NAP consistency across the website, Naver and Google, primary and secondary category selection, review replies, a photo and post cadence, and rank tracking measured across a geographic grid rather than at a single point. Engagements start with an audit report; taking the audit and doing the work in-house is a legitimate outcome. Explicitly not offered: ranking guarantees, bulk review generation, and autocomplete manipulation. Google documents that local ranking cannot be requested or bought, and review manipulation carries sanctions ranging from suspended account privileges to account termination. Local rank is determined by relevance, distance and prominence — distance cannot be edited, which is why any guarantee that does not name a location is unverifiable. Korean splits this topic across two spellings, both large: 구글 지도 / 구글지도 (673,000/mo) and 구글맵 / 구글 맵 (550,000/mo). Related demand: 구글 비즈니스 프로필 (2,900), 구글 마이비즈니스 (1,300), 구글 지도 등록 (590), 구글맵 리뷰 (210), 구글지도 상위노출 (50). Figures measured via DataForSEO, August 2026. https://trina.app/ko/services/google-map-seo ## Channels **Auto-published — the brand's own blog.** This is the only channel Trina posts to by itself, and it is included in every tier. Approved drafts go live on the brand's own domain, which is also where the search-traffic compounding happens. **Channel-native drafts, posted by a person** — Medium, LinkedIn, Dev.to, Tistory, Velog, Hashnode, Brunch. Growth covers 2 of these, Scale covers all 7. Each draft is written for its channel rather than being one post reposted seven times: length, opening, formatting and register all differ. The split is deliberate rather than a limitation of effort. Posting to those platforms on a customer's behalf runs into platform terms and access rules that are not safe to automate under a white-label model, so Trina writes the piece and a person publishes it. **Naver Blog is deliberately excluded.** Its server-side daily cap, multi-account detection and history of terms-of-service bans make safe automation impossible here. It is not offered as a channel. ## Company Built by Potenlab Inc. (team: TreeSoop). Contact: contact@trina.app ## Blog Trina's blog covers content operations and SEO automation for small teams and agencies: - https://trina.app/ko/blog — Korean - https://trina.app/en/blog — English ## Published guides on local search - https://trina.app/ko/blog/how-to-register-your-business-on-google-maps — listing a business on Google Maps: the four steps, why verification stalls, why a listing does not appear, and Google's stated ranking factors. - https://trina.app/ko/blog/google-maps-agency-comparison-cost-2026 — Korean Google Maps agencies compared with their published prices, what each actually sells, and what Google's contribution policy prohibits. - https://trina.app/ko/blog/local-seo-korea-vs-global-comparison-2026 — local SEO demand measured in Korea against English-speaking markets, and why the gap exists (high-precision map export was only approved in February 2026). - https://trina.app/ko/blog/blog-agency-cost-comparison-2026 — blog agency pricing in Korea, from 50,000 KRW per post to 1,000,000 KRW a month, and what separates the tiers. - https://trina.app/ko/blog/google-maps-ranking-problems-cost-2026 — why a claimed profile still does not rank: seven causes sorted by symptom, which of Google's three ranking factors each one touches, and what to check first. Distance cannot be changed, so the work is relevance and prominence. - https://trina.app/ko/blog/google-search-console-registration-guide-2026 — getting a site into Google search. Discovery, indexing and ranking are three separate stages; registration only covers the first two. Covers Search Console setup, sitemaps, and why a registered site still shows nothing. - https://trina.app/ko/blog/blog-traffic-drop-diagnosis-2026 — what to check, in order, when blog traffic falls. Splits the symptom first — impressions versus clicks, one query versus all, one page versus the site — because that split eliminates most causes before any fixing starts. Notes that Google has no Naver-style site-wide blog grade, and that the last two or three days of Search Console data are simply not collected yet. - https://trina.app/ko/blog/google-my-business — operating a Google Business Profile end to end: registration and ownership verification, filling the profile, photos, reviews, posts, the Q&A feature, insights, and multi-location setup. - https://trina.app/ko/blog/cohort-analysis-retention-metrics-2026 — grouping customers by signup month to see when churn actually happens, which an overall average hides: a growing business can post steady totals while each new cohort leaves faster than the last. Covers building the table in a spreadsheet, defining what counts as active for subscription versus long-purchase-cycle businesses, how retention, repeat rate and LTV connect, and the counting mistakes that make the numbers wrong. - https://trina.app/ko/blog/utm-parameter-naming-rules-2026 — UTM is easy to attach and hard to keep consistent. Once facebook, Facebook and FB are all in the data they cannot be merged, because UTM is recorded at collection time. Covers which of the five parameters actually decide the structure, why medium must come from a fixed list (GA4 derives its channel grouping from it, so an invented value lands in Unassigned), naming conventions, where UTM must never be attached — internal links restart the session and erase the original source — and how to clean up data already split. - https://trina.app/ko/blog/ga4-bounce-rate-vs-exit-rate-2026 — in GA4 bounce rate is simply the inverse of engagement rate, so it is not comparable to the Universal Analytics figure of the same name. Korean uses 이탈률 for three different metrics — bounce rate, exit rate and churn rate — which count different things and are fixed in different places. Covers what makes a session engaged, when a high figure is normal rather than a problem, and why a suspiciously low figure usually means duplicated tags. - https://trina.app/ko/blog/negative-review-response-guide-2026 — the removable set is narrower than owners expect: platforms take reviews down for policy violations, not for being wrong, so "we never did that" is not grounds. Sorts what actually gets removed from what does not, covers writing a reply whose real audience is the next customer rather than the reviewer, the incentive rules that get a profile suspended, and when legal action is worth considering. - https://trina.app/ko/blog/international-seo-hreflang-guide-2026 — translating a site is not what makes it findable abroad. Covers choosing a domain structure before anything else (it cannot be changed later without redoing the work), hreflang's self-reference and reciprocity rules, why en-UK is invalid where en-GB is not, and the common failure where a multilingual plugin points each translated page's canonical at the original and quietly prevents it from being indexed at all. - https://trina.app/ko/blog/ai-citation-tracking-guide-2026 — there is no rank to look up for AI citation: you have to ask, and the same question returns different answers each time. Covers building a question list in the words a customer would use before knowing you exist, asking three times per question in fresh sessions, what to log, checking whether AI crawlers are even allowed in, and why the point is the list of questions you are missing from rather than a score. ## Published guides on site and image setup - https://trina.app/ko/blog/image-compression-guide-2026 — resize before compressing. A 4000px original sitting in a slot that displays at 800px loses both weight and quality no matter how hard it is compressed. Covers format choice (WebP, AVIF, PNG, JPG), where visible quality loss starts, lazy-loading criteria, and the order to do the work in. - https://trina.app/ko/blog/robots-txt-guide-2026 — robots.txt controls crawling, not indexing. A blocked page can still appear in results, described only by its link text, because blocking the crawl also blocks the noindex tag from ever being read. Covers syntax and precedence, directives for AI crawlers, and the Disallow left over from development that removes a whole site from search. - https://trina.app/ko/blog/og-image-guide-2026 — why a link preview card stays broken after the tag is fixed. Rewriting tags does nothing while a platform cache is alive, and clearing the cache does nothing while the image URL is relative. Covers the four tags a card actually reads, the 1200x630 ratio and what it is for, absolute-URL and format traps, per-platform cache invalidation, and the order to check symptoms in. - https://trina.app/ko/blog/favicon-guide-2026 — making the file is the easy half. Google will only use a favicon in search results when it is square and a multiple of 48px, sits at a URL that does not change, is referenced by a link tag in the homepage head, and is crawlable along with that homepage. One favicon per host. Covers the size and format rules, designing for 16px, and why a change can be correct and still not show up for weeks. - https://trina.app/ko/blog/citation-format-guide-2026 — the hard part of citing is not the format, it is deciding what needs a source at all and what is common knowledge. Covers where quotation ends and copying begins, why a statistic should link to whoever produced it rather than to whoever repeated it, how to write a second-hand citation when the original cannot be found, the per-medium formats, the tighter rules images and fonts carry, and what Google's spam policy says about republishing someone else's work without adding to it. - https://trina.app/ko/blog/alt-text-guide-2026 — alt text does not describe an image, it replaces the job the image does on the page. A decorative image gets an empty alt rather than no alt at all, because a missing attribute can make a screen reader read out the filename. An icon button takes the name of the action, not the shape: a magnifier is "search". A chart takes the conclusion it carries, with the numbers moved into body text so they are not trapped in a picture. Covers the per-type rules, why keyword stuffing loses on both counts, and the order to fix already-published posts in. ## Published guides on hiring an agency - https://trina.app/ko/blog/hospital-marketing-agency-selection-cost-2026 — choosing a clinic marketing agency in Korea. Medical advertising is regulated, so the first question is not who brings more patients but who knows what may be published at all — and who decides, and where that decision is recorded. Also covers Google's review policy (paid and self-written reviews get a profile suspended, and reviews cannot be moved), account ownership, and why map presence usually comes before a website rebuild for a clinic. - https://trina.app/ko/blog/seo-agency-selection-criteria-cost-2026 — seven things to verify before signing with an SEO agency: what they measure and report, who owns the accounts and content, how they acquire links, and whether they track citation in AI search at all. Ranking guarantees are the clearest warning sign, because distance and algorithm are not theirs to set. ## Canonical URLs - https://trina.app/ko — Korean landing (default locale) - https://trina.app/en — English landing - https://trina.app/ko/case-study — TreeSoop case study (Korean) - https://trina.app/en/case-study — TreeSoop case study (English) - https://trina.app/ko/blog — blog index (Korean) - https://trina.app/en/blog — blog index (English) - https://trina.app/en/about — about the team