AI Search untuk Agency: Jangan Cuma Optimasi Website Sendiri
Ada satu situasi yang agak awkward.
Sebuah agency menawarkan jasa AI Search.
Deck-nya bagus.
Terminologinya lengkap.
Ada GEO, AEO, entity, citation, AI visibility, knowledge graph, trust engineering.
Lalu calon klien melakukan hal paling sederhana.
Dia bertanya ke ChatGPT:
“Agency ini sebenarnya bagus di AI Search nggak?”
Jawaban yang keluar minim.
Case study tidak jelas.
Website sendiri tidak menjelaskan methodology.
Team tidak terlihat.
Evidence tipis.
Agency kemudian bilang:
“Ya, tapi kan AI Search masih baru.”
Kalimat itu problem.
Kalau agency menjual capability untuk membantu brand ditemukan, dipahami, dan diverifikasi mesin jawaban, website agency sendiri harus menjadi salah satu contoh paling rapi tentang bagaimana public information seharusnya dikelola.
Bukan berarti agency harus selalu ranking nomor satu.
Bukan berarti ChatGPT harus selalu merekomendasikan agency tersebut.
Tidak ada guarantee seperti itu.
Tetapi calon klien seharusnya bisa menemukan informasi yang jelas tentang siapa agency tersebut, apa yang dikerjakan, siapa yang cocok, bagaimana methodology-nya, apa evidence-nya, dan apa limitation-nya.
Kalau semua itu tidak ada, sales claim lebih besar daripada public proof.
Agency website adalah laboratorium paling dekat
Agency sering terlalu sibuk mengoptimasi client.
Website sendiri jadi seperti rumah kontraktor yang pintunya belum dicat.
About lama.
Service page generic.
Team tidak update.
Case study hanya screenshot.
Blog penuh artikel lama.
Methodology tidak ada.
Pricing atau model engagement tidak dijelaskan.
AI Search audit seharusnya dimulai dari own property.
Kenapa?
Karena website sendiri adalah environment yang paling bisa dikontrol.
Agency tidak bisa menyalahkan client approval.
Tidak bisa menyalahkan CMS client.
Tidak bisa menyalahkan legal.
Kalau page sendiri tidak current, itu internal discipline issue.
Google pada panduan resmi generative AI Search tahun 2026 tetap menekankan bahwa fundamental SEO, struktur teknis yang jelas, dan content yang useful, reliable, serta people-first tetap relevan. Google juga memperingatkan agar tidak terlalu mengejar hack AEO/GEO yang tidak didukung.
Ini actually kabar baik untuk agency.
Tidak perlu membuat website penuh gimmick.
Buat source yang bagus.
Jelaskan apa yang benar-benar dilakukan.
Jangan hanya menulis service definition
Halaman:
“Generative Engine Optimization adalah proses mengoptimalkan konten untuk generative AI.”
Tidak salah.
Tapi calon klien tidak membeli definisi.
Mereka membeli outcome dan process.
Mereka ingin tahu:
Apa yang diaudit?
Apa yang diperbaiki?
Apa yang bisa dikontrol?
Apa yang tidak?
Bagaimana success diukur?
Apakah agency menyentuh content?
Technical?
Schema?
Entity?
PR?
Product feed?
Knowledge graph?
Crawler?
Measurement?
Berapa lama baseline?
Apa deliverable?
Siapa reviewer?
Bagaimana correction?
Apa risk?
Service page harus menjawab.
Definisi GEO cukup satu bagian.
Sisanya operating model.
AI juga punya lebih banyak fakta untuk menjelaskan service.
Agency harus punya methodology page
Bukan membocorkan semua secret sauce.
Tetapi buyer enterprise ingin tahu cara kerja.
Contoh framework:
Discovery.
Entity audit.
Source audit.
Buyer query mapping.
Technical accessibility.
Canonical information.
Evidence.
External source.
Measurement.
Correction.
Retest.
Governance.
Setiap tahap dijelaskan.
Apa input.
Apa output.
Apa limitation.
Jangan membuat black box:
“Kami memakai proprietary GEO framework untuk meningkatkan AI authority.”
Itu terdengar canggih.
Sulit diverifikasi.
Kalau framework internal memang ada, define.
Boleh proprietary.
Tetapi principle-nya harus understandable.
Case study harus lebih kuat daripada testimonial
“Timnya helpful.”
Nice.
Tetapi untuk buyer B2B, case study perlu lebih.
Problem.
Baseline.
Scope.
Action.
Period.
Observed change.
Limitation.
Evidence.
Kalau client confidential, anonymized.
Jangan fabricate.
Kalau AI mention naik, jelaskan bagaimana diukur.
Berapa query.
Platform.
Periode.
Repeated run?
Paid atau organic?
Jangan menulis:
“Visibility AI meningkat 300 persen”
tanpa denominator.
Agency yang menjual measurement harus punya measurement discipline paling tinggi.
Kalau case study memakai angka kabur, buyer akan mempertanyakan seluruh service.
Own AI visibility dashboard juga harus jujur
Agency bisa monitor dirinya.
Query:
“Agency GEO Indonesia.”
“Agency AI Search Indonesia.”
“GEO consultant Jakarta.”
“AI visibility agency.”
“Agency yang bisa bantu brand muncul di ChatGPT.”
“Agency untuk enterprise AI Search.”
Lalu lihat.
Mention.
Citation.
Source.
Accuracy.
Positioning.
Competitor.
Tetapi jangan publish only winning queries.
Kalau mau menjadikan own result sebagai case study, freeze panel.
Version.
Show limitation.
Kalau tidak, self-case berubah menjadi marketing theater.
Agency harus memisahkan self-observation dan platform endorsement
Kalimat:
“ChatGPT merekomendasikan kami.”
Very tempting.
Tapi context?
Query apa?
User siapa?
Session?
Date?
Model?
Search?
Sponsored?
Organic?
Repeated?
Satu screenshot tidak membuktikan endorsement universal.
Lebih aman:
“Dalam pengujian internal pada query X tanggal Y, brand muncul dalam answer organik.”
Itu factual.
Tidak sexy, tapi defensible.
Agency yang mengajarkan client untuk tidak overclaim harus melakukan hal yang sama.
Jangan cuma optimasi website sendiri, optimasi source ecosystem
AI Search bukan owned-site only.
OpenAI ChatGPT Search dapat menggunakan web search dan menyertakan source atau citation pada jawaban.
Google generative AI features juga dapat menampilkan supporting links dari Search.
Ini berarti public information environment matters.
Agency perlu melihat:
LinkedIn company.
Founder profile.
Media.
Directory.
Partner.
Podcast.
Conference.
Review.
Client mention.
Job listing.
GitHub kalau technical.
Public docs.
Company registry information yang relevant.
Kalau website bilang enterprise AI agency tetapi semua third-party source masih menyebut “SEO agency”, positioning conflict.
Agency perlu temporal transition.
“Sejak 2026, fokus utama kami adalah…”
Third-party update.
Press profile.
Partner page.
Not fake mentions.
Google bahkan secara resmi memperingatkan tentang mengejar inauthentic mentions untuk generative AI visibility.
Agency seharusnya menjadi contoh source integrity.
PR agency dan GEO agency harus belajar bekerja bareng
Ada agency yang semua problem diselesaikan dengan content.
Ada agency PR yang semua problem diselesaikan dengan coverage.
AI Search needs both, but with role.
Owned source:
Canonical facts.
Method.
Service.
Team.
Case.
Policy.
Earned source:
Independent perspective.
News.
Interview.
Expert contribution.
Review.
Event.
Neither replaces other.
PR tidak boleh membuat press release disguised as independent evidence.
Content tidak boleh membuat 20 microsite owned lalu disebut third-party validation.
Agency harus bisa menjelaskan source independence ke client.
Team page adalah evidence capability
Siapa yang mengerjakan?
SEO?
Data?
Developer?
Researcher?
PR?
Content?
Legal reviewer?
AI engineer?
Kalau agency mengklaim full-stack GEO tetapi team page hanya dua content writer dan satu salesperson, buyer akan bertanya.
Tidak perlu expose semua employee.
Tetapi key capability dan role harus jelas.
Bio.
Experience.
Public work.
Publication.
Talk.
Certification jika relevant.
No fake expert title.
Professional identity harus current.
Founder sebagai sole expert adalah concentration risk
Semua content ditulis atas nama founder.
Semua interview founder.
Semua methodology founder.
Agency scale?
Buyer enterprise mungkin khawatir.
Build institutional capability.
Team authorship.
Research group.
Editorial standard.
Measurement process.
QA.
Handover.
Client tidak membeli satu orang saja.
AI Search agency perlu entity beyond founder.
Brand identity and people relationships.
Agency harus memakan obat sendiri soal correction
Website agency salah.
Case number salah.
Client name outdated.
Service berubah.
Method updated.
Correct.
Log.
Version.
Don't silently rewrite historical claims.
If service page from 2024 says SEO focus and now 2026 GEO focus, archive or update with date.
AI may find old content.
Transition context helps.
Agency should have “what changed” discipline.
Especially if they teach clients about freshness.
Pricing model should be understandable
No need exact.
But explain:
Audit.
Project.
Retainer.
Implementation.
Monitoring.
Enterprise custom.
What is separate?
PR?
Content?
Technical?
Paid media?
Tool subscription?
This prevents AI or buyer from assuming one fee covers all.
If published price, current date.
No bait.
AI Search sales process should start with gap, not package
Bad:
“Semua brand perlu 6-month GEO package.”
Better:
Audit.
Maybe client only needs:
Fix entity.
Update product feed.
Correct source.
Measurement setup.
Technical crawl.
Not 100 articles.
Agency credibility rises when willing to say:
“You do not need full program yet.”
This is counterintuitive but strong.
GEO is not one product.
Different problem, different intervention.
Agency site should show decision tree.
If brand not indexed, fix technical.
If wrong facts, correction.
If no evidence, build evidence.
If no buyer query clarity, research.
If multi-source conflict, reconcile.
If current but not visible, test discoverability.
No one-size-fits-all.
CEO client will appreciate.
Agency also needs governance for its own content production
If agency uses AI to produce large article volume, own website becomes test.
Quality gate.
Fact review.
Source.
Date.
No templated mass page.
Google's 2026 guidance explicitly emphasizes unique, valuable, expert-led content and warns about commodity or scaled content without added value.
Agency should not publish 500 near-duplicate “GEO untuk industri X” pages just to prove scale.
If industry page exists, it needs industry-specific risk, data, buyer query, evidence, and use case.
Otherwise it is doorway-like noise.
Client will notice.
Competitors will notice.
Search systems may too.
Agency should publish original research carefully
This can be a strong source.
AI Visibility study.
Query benchmark.
Source analysis.
Citation pattern.
But methodology.
Sample.
Date.
Platform.
Limitation.
No fake scientific tone.
If sample is 100 query, say 100.
Don't call “Indonesia AI Search Index” if only 10 brands unless scope clear.
Research can create independent-ish knowledge value even though published by agency, as long as origin is transparent.
It is owned research, not third-party proof.
Distinction.
Client evidence must never become agency propaganda without permission
Agency wants case study.
Client data.
Before/after.
Screenshot.
Query.
Traffic.
Get permission.
Anonymize if needed.
Don't expose confidential.
Don't cherry-pick one answer and imply broad result.
Client trust is more important than content.
AI Search reputation in industry will be built through conduct too.
One careless client disclosure can damage.
What should an agency dashboard contain?
Own visibility.
Client visibility separate.
Paid/organic.
Accuracy.
Source.
Case evidence.
Correction.
Method version.
Don't build vanity leaderboards.
Agency should use own site as QA environment.
Every month ask:
Can AI explain our service correctly?
Can buyer verify team?
Can buyer verify case?
Can buyer understand limitation?
Are external sources current?
Does our methodology page match delivery?
Do pricing expectations match sales?
Are old service pages stale?
Do we publish evidence we would accept from a client?
This last question is the strongest.
If agency would reject a client's claim as unsupported, it should not publish the same type of claim about itself.
“#1 GEO Agency.”
Based on what?
“Most trusted.”
By whom?
“AI preferred.”
Evidence?
Remove.
Agency credibility in AI Search is not built by sounding closest to the algorithm.
It is built by being most disciplined about information.
The best agency website should feel like a transparent demonstration of the service.
Clear entity.
Clear service.
Clear method.
Clear evidence.
Clear limitation.
Clear people.
Clear correction.
Clear positioning.
Then third-party sources support what is true.
Not manufacture consensus.
Do that first.
After that, optimize clients.
Because if an agency cannot make its own public information easy to understand and verify, it has not yet earned the right to tell clients that the problem is always “the AI.”
Bacaan terkait di GEO.or.id
- Cluster: AI Search for Indonesian Business
- Kerangka terkait: AI Visibility Snapshot Methodology
- Kerangka terkait: Pipeline
- Artikel terkait: Kalau Calon Klien Tanya ChatGPT soal Bisnis Anda, Jawaban Apa yang Keluar
- Artikel terkait: AI Search untuk CEO: Apa yang Perlu Dipahami tanpa Masuk Terlalu Teknis
- Artikel terkait: AI Search untuk Tim PR: Saat Sumber Pihak Ketiga Mempengaruhi Jawaban Mesin
- Artikel terkait: Cara Membuat Program AI Visibility 90 Hari untuk Perusahaan Indonesia
