Migaku Review: strong product, weak speed (65/100)

Field	Value
Domain	migaku.com
Category	Immersion-based language learning with contextual media tools and flashcards
Pricing	Unknown in the supplied site facts; the profile records a subscription model with free trial and lifetime option
Crawl	39 pages on 2026-10-11

Migaku is legible as an immersion-first language-learning platform for serious, self-directed learners. First impressions score 86/100 and audience messaging 82/100 because the site makes the workflow concrete: choose native content, click words in websites or subtitles, see context, and turn discoveries into cards. Accessibility has strong sampled foundations at 78/100, alongside verified contrast, alt-text and semantic issues.

Performance is the defining weakness. It scores 28/100: mobile lab LCP reached 34.1 seconds, while real-user homepage LCP was 3.4 seconds and INP was 376 ms. Design execution scores 58/100, held back by token drift, contrast failures, missing image sizing metadata and mobile layout risk. Usability is 58/100 because desktop paths are short but mobile loading creates conversion friction.

Writing quality is fair at 68/100, decision support 61/100, review-content integrity 70/100, self-serve help 72/100 and technical SEO 72/100. The public surface is useful, but proof and plan comparison remain thinner than the product story.

Migaku presents a clear immersion-first proposition for serious, self-directed language learners. The homepage shows a concrete workflow: click words in websites and subtitles, see explanations in context, and create flashcards from content you already care about. The team page adds credibility through language-learning experience, including JLPT N1 experience and multilingual staff. Pricing metadata identifies monthly, yearly, and lifetime options plus a free trial. The remaining positioning gap is competitive framing: Google Suggest places Migaku beside Language Reactor, LingQ, Anki, Yomichan, and Duolingo. State where contextual immersion is distinct from those alternatives, then support that distinction with a concise comparison.

A motivated learner can understand Migaku’s core workflow: choose native content, click or tap words, inspect context, and turn discoveries into cards. Navigation is goal-oriented, with Download and Pricing alongside language-specific Learn links such as Learn Japanese and Learn Chinese. The team page signals an audience of experienced learners and power users rather than absolute beginners. The unresolved decision is where to start and what Migaku replaces. A concise first-session path should explain when to choose the browser extension, app, or language course, then show how pricing, download, and signup relate. Transparent comparison would reduce the work of deciding whether Migaku complements or replaces other study tools.

Migaku’s information architecture is easy to enter, but mobile loading can block the decision path. The homepage presents a specific promise, explains learning through streaming content and subtitles, and reaches Pricing in one click; the main language pages are also at depth 1. On mobile, however, PageSpeed Insights measured 34.1 s LCP, 30.4 s FCP, 36.6 s TTI, and a 12,046 KiB payload. Download, Pricing, Log in, Get started, and Sign up split the next action across several labels. Prioritize above-the-fold mobile rendering, then clarify one primary conversion path and make the relationship between pricing, installation, and signup explicit. Promote Getting started from FAQ as a second usability fix.

Migaku has strong sampled accessibility foundations, but verified semantic and visual issues still need repair. Available DOM facts show landmarks, English language metadata, a skip link, and no unlabeled inputs. The homepage also has six empty alt attributes, while the accessibility audit reports failed color contrast and SVG alternative-text checks. A CSS review found three outline:none or outline:0 occurrences alongside five :focus and two :focus-visible rules; equivalence is not established. Audit every outline suppression, add concise alternatives to informative images, intentionally retain empty alternatives only for decorative imagery, and measure failing color pairs against WCAG AA. Reducing the 12,046 KiB payload should also improve interaction for users facing heavy rendering work.

Migaku has a recognisable illustration-led presentation, but the implementation lacks a compact, dependable visual system. The crawl inventory reports 142 colors, 59 font sizes, 43 radii, 145 spacing values, and 15 font families, a pattern consistent with token drift. PSI flags color-contrast failures on mobile and desktop; the homepage outline contains two h1 elements; and images fail the explicit-sizing audit despite measured CLS of 0.004 mobile and 0.015 desktop. Mobile PSI also reports 34.1 s LCP, 1,539 ms TBT, and a 12,046 KiB payload. Establish semantic tokens, repair contrast and heading structure, add intrinsic image dimensions or aspect ratios, and defer non-critical code before polishing component details.

Performance is Migaku’s defining weakness: the mobile lab reached LCP in 34.1 s, while real-user homepage LCP was 3,387 ms and INP was 376 ms. The lab transferred 12,046 KiB; the homepage network log recorded 310 requests, 99 blocking requests, and 17,246 KiB of traffic. Listed third-party bytes include Facebook at 821 KiB, Google Tag Manager at 513 KiB, TikTok at 156 KiB, and Klaviyo at 127 KiB. The changelog reached 3,856 ms TBT and 6,826 ms of JavaScript boot-up time, while the Chinese page weighed 5,065 KiB. Defer non-essential scripts, set route budgets, remove unused code, optimize media, and validate the true LCP element after each change.

Migaku’s strongest writing is concrete: the homepage names Netflix, Japanese, a three-month goal, and the Chrome extension, then explains how favorite websites become learning materials. The Netflix article offers a practical sequence but opens with the duplicated phrase “Ahh, yes. Here you are are.” The pricing page has no H1, while the homepage exposes a second H1 and malformed extracted headings such as “What the heckheck...”. The grammar page and changelog each contain a 22,000-word-scale paragraph in the extracted sample: 22,406 words and 22,565 words respectively. Correct the opener, keep one H1 per page, remove demo text from the heading tree, and segment long visible prose into shorter paragraphs, lists, and task-oriented subheadings.

The pricing surface is easy to locate, but the supplied evidence shows a handoff rather than a rich comparison experience. The FAQ pricing answer confirms that Migaku offers several subscription plans and directs readers to the main pricing page for current details and promotions. It also sets a clear expectation that business plans are handled case by case through support@migaku.com. This answers basic availability questions but leaves plan differences outside the immediate decision context. Add a compact summary of plan differences and a stable starting price to the FAQ answer, while keeping the live Pricing page as the detailed follow-up. Also explain how Pricing, Get started, Download, and business contact fit together.

Migaku’s public proof is consistent with its product story, but it is primarily promotional. Language-specific pages explain learning through Netflix and books, show extension and mobile-app imagery, repeat Get started and No credit card required, and include testimonials attributed to the App Store and Discord. The FAQ pricing surface provides a direct business-contact route. This consistency is useful: the workflow, calls to action, and testimonial pattern reinforce the same offer. The supplied crawl does not establish independent outcome metrics, customer case studies, or dated review freshness, so confidence would improve with evidence beyond testimonials. Keep the shared landing-page template, but add language-specific examples or comparison cues and label testimonial source and context consistently.

The sampled public surface appears stable, with one clear status-code defect. A guaranteed-nonexistent URL returned HTTP 200 while presenting a 404 title, allowing an error page to compete with real content and blurring search status signals. The crawl found no sitewide noindex, robots block, client-side metadata conflict, or broad rendering failure. Coverage remains partial: 39 pages were crawled while the sitemap lists 3,817 URLs, and GSC is not connected, so traffic magnitude and index coverage cannot be measured. Return HTTP 404 for unknown routes, then connect GSC to establish indexed-page, click, and impression baselines. The homepage also reports 78 images without alt text, an avoidable secondary risk.

Migaku has a meaningful public self-serve surface. FAQ sections cover account, getting started, goals, pricing, features, and courses; goal pages answer tasks such as making subtitles interactive and creating media-rich flashcards; and /llms.txt provides a structured product overview with links to core surfaces. The main weakness is findability. Label the entry Help or Guides, put Getting Started prominently on the landing page, organize answers around tasks and troubleshooting, and add search or topic filtering. Freshness and full coverage remain only partly assessable because the crawl is partial and FAQ freshness metadata was not supplied.

Migaku is broadly crawlable and indexable across the sampled public surface: HTTPS pages return 200, use self-referencing canonicals, and are included in a sitemap. Robots.txt permits public content and major AI crawlers while disallowing admin and asset paths, and raw-to-rendered comparisons show near parity on sampled marketing and editorial templates. The highest-priority defect is the guaranteed-nonexistent probe, which returned HTTP 200 with a 404 title. The homepage also reports 78 images without alt text; /pricing has the bare title “Pricing” and no H1; and several language pages expose duplicate H1 elements. The status-code finding warrants the detailed fix in Risk & stability; here, focus on the 39-page sample against the 3,817-URL sitemap, the 78 missing alt attributes, the /pricing title and H1, and duplicate H1 normalization.

Migaku is a credible fit for serious self-directed language learners who want to use native video and web content as study material. Its strongest asset is specificity: the site explains a practical workflow, gives learners direct routes to Download and Pricing, and supports the product with FAQ coverage and a structured /llms.txt overview.

The first fix is mobile performance: the 34.1-second lab LCP, 3.4-second real-user LCP and 376 ms INP make loading the clearest conversion risk. Then resolve contrast and image-treatment issues and improve plan comparisons. At 65/100, Migaku has a strong product explanation and useful support surface, but speed and decision evidence keep the experience below its potential.

This 13-dimension review uses the supplied crawl, public dimension score table and sampled accessibility, performance, content and technical evidence. The crawl covered 39 pages on 2026-10-11. Dimension 19 · Editorial QA of content was excluded because its source was missing. GSC access was unavailable.

Read How SiteList scores for the review method and evidence standards. Pricing remains unverified in the supplied site facts, although the site profile records a subscription model with free trial and lifetime option.