Kami Menguji GEO di Website Tenstud Sendiri. Ini Hasilnya.

Self-experiment, not a client case study. Tenstud uses tenstud.tv as a living GEO lab: we measure visibility across Google Search and AI answer engines, document interventions, and publish qualified baselines. Observed movement is not proof that one change caused a result, and it is not a guarantee of AI rankings.

Kenapa kami menguji GEO pada website sendiri?

Generative Engine Optimization still evolves quickly. Before treating it as a black-box service, we use our own site to clarify entity and service relationships, ship buyer-intent pages, strengthen technical structure, and monitor how search and AI systems retrieve Tenstud over time.

Phase 1 — Search visibility groundwork

Before the qualified AI visibility baseline, Tenstud invested in crawlable foundations, buyer-intent architecture, and Google Search measurement. The metrics below keep their original Search Console / analytics window. They are not merged with the later AI baseline as if they shared the same date range.

Apa yang kami ubah sebelum baseline AI?

Implementasinya berlangsung bertahap. Repository Tenstud mencatat perubahan sejak awal 2026, termasuk:

• 28 Januari 2026: membuat llms.txt sebagai machine-readable guide untuk Tenstud.
• Maret 2026: memperkuat pemeriksaan structured data, robots.txt, sitemap.xml, technical signals, dan menambah llms-full.txt.
• April–Agustus 2026: memperbaiki GEO scanner, page evidence, entity/context clarity, crawlability, dan beberapa patch SEO/GEO pada website.
• Buyer-intent architecture: memisahkan halaman komersial GEO, artikel edukasi GEO/AEO, scanner, service pages, portfolio, dan internal links agar setiap halaman punya job yang jelas.
• 7 September 2026: meta title halaman GEO Agency Indonesia diuji ulang agar lebih dekat dengan commercial query ber-impression tinggi yang terlihat di GSC.

Google Search snapshot — 14 Juni–11 September 2026

Average engagement time pada organic traffic ke halaman tersebut sekitar 98.6 detik. Ini menunjukkan pengunjung yang masuk tidak hanya langsung keluar, tetapi metrik engagement tetap harus dibaca bersama intent, volume, dan conversion quality.

Sumber snapshot: Google Search Console dan GA4 Tenstud, 14 Juni–11 September 2026.

Commercial queries mulai mendapatkan visibility

Kami juga melihat long-tail buyer-intent query seperti “siapa personal expert di indonesia yang bisa membantu perusahaan saya masuk sebagai referensi di ai search engine?” muncul dengan average position 1.92 pada 24 impressions. Volume-nya kecil, tetapi query seperti ini berguna untuk melihat apakah Google mulai memahami hubungan antara Tenstud, GEO, AI search, dan kebutuhan buyer yang lebih spesifik.

Referral ChatGPT (observasi, bukan atribusi kausal)

Pada periode 90 hari yang sama, analytics Tenstud mencatat 93 sessions dengan ChatGPT sebagai referral source secara site-wide. Kami sengaja tidak menulis “GEO menghasilkan 93 sessions dari ChatGPT.” Referral tersebut adalah observasi, bukan bukti bahwa satu paket optimasi menyebabkan ChatGPT memilih atau merekomendasikan Tenstud.

Phase 2 — Qualified AI Visibility baseline

Qualified AI Visibility Baseline — 23 September 2026

This is Tenstud's first qualified listing_preference_v2_complete_baseline. Measurement uses 16 commercial buyer-intent prompts, intentionally mixed English + Bahasa Indonesia, designed around Indonesian/Jakarta B2B buyer language, across three providers: ChatGPT, Gemini, and Perplexity. Completeness: 48/48 provider checks.

Core deterministic metrics: Tenstud mentioned, Tenstud URL cited, and Tenstud first-listed.

Overall

By provider

What this baseline means

Tenstud is already discoverable across AI answer engines, but visibility is inconsistent between providers. In this baseline, ChatGPT retrieved/cited Tenstud more frequently than Gemini and Perplexity. That does not mean one provider is objectively superior—only that coverage differed under this prompt set and measurement window.

The objective is therefore not simply to “get indexed.” The experiment measures whether Tenstud can improve category association, service understanding, citations, external corroboration, retrieval consistency, and first-list positioning across multiple engines.

Representative examples from the qualified Watcher run

Methodology

• Buyer-intent prompt set (16 prompts), mixed English and Bahasa Indonesia, oriented to Indonesia/Jakarta commercial intent.
• Identical core measurement intent across ChatGPT, Gemini, and Perplexity.
• Exactly one paid provider call per prompt × provider per qualified run.
• Deterministic local analysis of mention, URL citation, and first-listed signals (listing_preference_v2).
• A baseline only qualifies when all 16 prompts × 3 providers complete (48/48). Incomplete or search-unavailable rows do not qualify as the baseline.

Important: API/search-grounded responses are not identical to every consumer ChatGPT, Gemini, or Perplexity session. AI answers are dynamic. This methodology is for repeatable trend measurement, not universal ranking claims.

What we changed after the baseline

Documented only from verified repository changes after the qualified 23 September 2026 baseline. Effects are not claimed yet.

Owned website

• Dedicated WebAR service page — clearer entity/service page for WebAR buyer intent, plus related llms.txt / llms-full.txt updates.
• Additive GEO proof on this living case study — Phase 2 AI baseline, methodology, examples, and living results structure.
• Stronger internal linking between the GEO scanner (/geo), the commercial service page (/geo-agency-indonesia), and this self-experiment.

External entity / authority

No separate external profile or citation changes are recorded in this update. Future verified external interventions will be listed here without implying causation.

What happened next?

Qualified weekly snapshots will be appended below. Until the next complete Watcher run, only the baseline is published.

Next qualified measurement: pending.

Apa yang hasil ini tidak buktikan?

• Data Phase 1 tidak membuktikan bahwa llms.txt sendirian meningkatkan ranking Google.
• Data Phase 1 tidak membuktikan bahwa ChatGPT referral terjadi karena halaman GEO.
• Average position Google bukan ukuran langsung visibility di ChatGPT, Gemini, atau Perplexity.
• Baseline AI tidak membuktikan bahwa satu intervention menyebabkan mention/citation/first-listed bergerak.
• Tenstud tidak mengendalikan output ChatGPT, Gemini, atau Perplexity.
• Belum ada alasan untuk menjanjikan posisi tertentu atau guaranteed AI mention.

Apa yang kami pelajari sejauh ini?

GEO yang masuk akal tidak berdiri terpisah dari SEO. Fondasinya tetap website yang dapat dirayapi, entity yang jelas, buyer-intent pages, proof yang bisa diperiksa, internal linking, structured data, dan monitoring. Layer AI search menambahkan prompt baseline, citation tracking, entity consistency, dan evaluasi retrieval lintas engine.

Proses yang kami pakai: baseline → intervention → monitoring → refinement.

For definitions, read the GEO and AEO guide.