Back to case studies
    Success story

    Case Study: How an Online Language Courses App Grew AI Citations by 70% in 6 Months

    Nov 12, 2026 · 7 min read

    Nov 12, 2026 · @Martech Team

    An online language courses app went from 410 to 697 monthly AI citations in six months, a 70% increase. AI assistants now name it in 41% of answers about learning a language online, up from 12%. None of this came from new ads or a bigger blog. It came from giving AI models the facts they needed to compare the app with bigger, better-known rivals.

    At a glance

    BusinessOnline language courses app with live tutors and self-paced lessons
    Languages9 languages taught, website and app in 5 interface languages
    Learners~4,500 active learners, mostly in Europe
    Prompts tracked60, in 3 languages, across 4 AI models
    Time frame6 months
    Headline result+70% citations (410 → 697 per month)

    The client

    The app teaches 9 languages through a mix of short self-paced lessons and 25-minute live sessions with certified tutors. Its learners are mostly adults: people moving abroad for work, students preparing for exams such as IELTS or DELE, and professionals who need a language for their job.

    It had a loyal user base and a 4.7 app store rating. Its marketing team of three ran paid social, an email programme and a blog in English. AI visibility had never been measured.

    The challenge: losing the comparison before it started

    Language learning is one of the most crowded categories AI assistants answer about. When a learner asks which app to use, the AI usually lists three to five options with one line each. The two largest free apps took a place in almost every answer.

    The client did appear sometimes, but in the wrong way. In the first month of tracking, AI answers described it as "a tutoring marketplace", quoted a price that was two years out of date, and in one model claimed it did not offer Spanish. When it was cited, the source was usually a third-party listicle, not its own website.

    Three problems sat underneath:

    • No comparable facts. Course pages said "learn faster" but never said how long a level takes, what it costs per month or which exam it prepares for. AI models compare on facts, so they skipped the app.
    • English only. A third of learners searched in German, Spanish or Polish, but the site's useful content existed only in English.
    • Outdated third-party pages. The most cited pages about the app were review articles written in 2023 and 2024.

    Step 1: Track the questions learners actually ask

    Learners rarely ask an AI for "a language app". They ask about a goal and a deadline. We pulled questions from support tickets, onboarding surveys and tutor notes, and turned them into 60 prompts. Each prompt was tracked in English, German and Spanish.

    Learner goalPromptsExample prompt
    Exam preparation14Best online course to pass DELE B2 in 3 months
    Moving abroad12How to learn German fast before relocating to Berlin
    Speaking practice10Affordable app with live conversation practice in Italian
    Work and business9Business English course with certificate for my CV
    App comparisons9Language app with real tutors vs free apps, which is better?
    Kids and teens6Online Spanish lessons for a 12-year-old

    Brand-name prompts were left out on purpose. The growth had to come from learners who did not yet know the app.

    Step 2: Find where the AI gets its answers

    After three weeks of runs, the sources export showed which pages AI models relied on for language-learning questions.

    Source typeShare of citationsWhat we found
    "Best language apps" listicles31%Client in 3 of 11 lists, all with old prices
    Forums and Q&A communities21%Learners praised the tutors, but threads were 2+ years old
    Brand websites (all apps)19%Client's pages cited only 6 times; no clear facts to quote
    Video platforms (via transcripts)12%No client videos with spoken comparisons or level guides
    Exam boards and education sites11%Official exam pages linked to rivals' prep courses, not the client
    App store listings6%Description focused on design, not on levels, tutors or exams

    The lesson: the AI was not ignoring the app. It just could not find a recent, factual page to quote when comparing it.

    Step 3: Fix, publish and earn

    The work ran in three overlapping phases. The marketing team did most of it, with tutors writing the expert content.

    Months 1 and 2: make the facts quotable

    • One fact sheet per course. Every course page now opens with the same short block: level range (A1 to C1), hours per level, live sessions per month, price per month and which exams it prepares for.
    • Updated pricing everywhere. Prices and plans were corrected on the site, app store listings and 22 partner and review pages that the team could edit or request changes to.
    • Structured data. Course, FAQ and review markup was added so models and search engines read the same facts.

    Months 3 and 4: content AI models can compare

    • Honest comparison pages. Six pages compare the app with free apps and tutor marketplaces, including where the alternatives are the better choice. These became the most cited pages on the site.
    • "How long does it take" guides. Tutors wrote guides per language and level, such as hours from A2 to B1 in German, based on the app's own anonymised learner data.
    • Five interface languages. The fact sheets, comparison pages and guides were translated and adapted into German, Spanish, French and Polish, not machine-translated word for word.

    Months 5 and 6: earn new third-party sources

    • Original data. A short report on 4,000 learners showed how often learners study and which habits predict passing an exam. Two education outlets and 9 language blogs cited it.
    • Listicle refresh. The team sent updated facts and a test account to the authors of the 11 most cited listicles. Seven updated their articles; four added the app for the first time.
    • Tutor voices in communities. Tutors answered learner questions in forums under their own names, with clear disclosure. They gave advice first and mentioned the app only where relevant.
    • Short video explainers. Twelve videos with spoken level guides gave transcript-reading models new material to cite.

    Results after 6 months

    AI citations rose 70% in 6 months, from 410 to 697 monthly citations
    PeriodMonthly citations
    Start410
    Month 1428
    Month 2465
    Month 3520
    Month 4585
    Month 5645
    Month 6697

    AnswerRank runs · 60 prompts in 3 languages across 4 AI models · monthly totals

    Months 1 and 2 fixed what the AI already had wrong, so growth was small. The jump came in months 3 and 4, when comparison pages and level guides gave models facts to quote. Months 5 and 6 added new third-party sources from the learner data report and refreshed listicles.

    ResultStartAfter 6 months
    Share of AI answers that name the app12%41%
    Prompts where the app is mentioned at least once11 of 6034 of 60
    Citations of the app's own website per month6148
    AI models that recommend the app2 of 44 of 4
    Non-English prompts that mention the app3 of 4022 of 40
    Monthly trial sign-ups from AI tools~90~610

    Sign-ups from AI tools were measured through referral data and a "How did you hear about us?" question at sign-up. Learners who came from an AI answer converted to paid plans at a higher rate than paid-social sign-ups, because they arrived already comparing on price and level.

    Takeaways for edtech and language apps

    Give the AI numbers to compare. Hours per level, price per month and exam coverage are what models quote. Marketing claims are not.

    Track in the languages your learners use. Prompts in German and Spanish produced different sources and different rivals than English ones.

    Be honest in comparisons. Pages that admitted when a free app was the better fit were cited far more than pages that only praised the product.

    Your own data is a source. Anonymised learner statistics turned into a report that others cited, and the AI followed.

    Old reviews keep talking. Refreshing the listicles the AI already trusted was faster than waiting for new ones to rank.

    In the client's words

    *"We thought AI assistants simply preferred the big free apps. It turned out they just didn't know our facts. Once our prices, levels and exam results were easy to find and compare, we started showing up next to the names everyone knows."*

    — *Head of Marketing, online language courses app*

    See how your brand appears in AI answers

    AnswerRank monitors your visibility across ChatGPT, Gemini, Perplexity and Claude — and builds the content that earns it.