AI Mode Bukan Chatbot Biasa: Implikasinya untuk Arsitektur Konten

Kalau kita menganggap AI Mode hanya sebagai chatbot yang ditempel di Google, content strategy mudah salah arah.

Team mulai membuat FAQ panjang.

Semua page diubah jadi format Q&A.

Setiap paragraph dipendekkan.

Semua artikel diberi summary.

Targetnya:

“Biar gampang dijawab AI.”

Problem-nya, AI Mode bukan sekadar chatbot yang membaca satu page lalu menjawab.

Google menjelaskan AI Mode menggunakan Search infrastructure, core ranking and quality systems, dan dapat melakukan query fan-out untuk menjalankan multiple related searches.

Ia mencari web.

Membandingkan information.

Menggunakan multiple sources.

Menghubungkan subtopics.

Ini punya implication besar untuk arsitektur konten.

Website tidak perlu menjadi satu giant chatbot transcript.

Website perlu menjadi knowledge environment yang Search bisa temukan, pahami, dan hubungkan.

Page bukan prompt response

Content architecture klasik sering dibangun dari keyword.

Keyword A -> Page A.

Keyword B -> Page B.

Keyword C -> Page C.

AI Mode mendorong model yang lebih semantic.

User question bisa punya beberapa subneeds.

Satu page menjawab definition.

Page lain methodology.

Page lain evidence.

Page lain pricing.

Page lain implementation.

Google dapat fan-out.

Jadi website perlu relationship antarpages yang masuk akal.

Bukan setiap possible question punya URL.

Ini perbedaan penting.

Jika Anda membuat 300 pages berdasarkan 300 prompt variations, kemungkinan besar menghasilkan redundancy.

Google pada guidance 2026 bahkan memperingatkan membuat banyak pages untuk query variations dengan tujuan manipulasi dapat jatuh ke scaled content abuse jika tidak memberikan value.

Architecture harus mengikuti information domain.

Bukan prompt inventory.

Mulai dari entity dan decision objects

Sebelum membuat content tree, tentukan core objects.

Organization.

Service.

Product.

Industry.

Use case.

Method.

Evidence.

Customer.

Location.

Policy.

Research.

Glossary.

Event.

Person.

Setiap object punya relationship.

Service used in industry.

Case study proves service.

Method supports service.

Research informs category.

Location serves market.

Person authors research.

Ini bukan berarti semua harus jadi knowledge graph formal.

Mulai dengan mental model.

Website bukan kumpulan artikel.

Ia adalah representasi bisnis.

AI Mode memperbesar value dari representation yang coherent.

Hub pages kembali penting

Dalam era content farm, hub sering hanya category archive.

Sekarang hub bisa menjadi orientation layer.

Contoh:

/ai-optimization/

Menjelaskan:

apa itu.

Untuk siapa.

Core components.

Measurement.

Technical readiness.

Content.

Entity.

Governance.

Links ke deep pages.

Hub memberi context.

Deep pages memberi detail.

User bisa navigate.

Search crawler bisa discover.

AI retrieval punya multiple surfaces.

Hub tidak harus panjang 5.000 words.

Harus useful.

Jangan buat hub hanya list links.

Berikan information architecture.

Canonical page harus menjawab persistent question

Ada questions yang terus kembali.

“Apa itu service ini?”

“Bagaimana process?”

“Berapa biaya?”

“Siapa yang cocok?”

“Apa evidence?”

“Bagaimana data handling?”

Ini layak canonical pages.

Bukan answered repeatedly di 40 articles.

Kalau pricing fact tersebar di 50 posts, maintenance nightmare.

Make canonical pricing page.

Articles link.

If price changes, one source.

AI Search benefits from current information.

Content ops juga.

Architecture is governance.

Blog should add perspective, not duplicate core facts

Editorial articles should explore:

trends.

cases.

failure.

opinion.

news.

analysis.

scenarios.

Not copy service description every time.

If article about AI Mode reporting needs definition Search Console, link canonical guide.

Don't rewrite full definition from scratch.

This reduces drift.

POST HUMANIS can remain narrative.

Canonical pages can be structured.

Different page types have different voice.

AI Mode doesn't require all pages same format.

In fact, editorial diversity helps human experience.

Machine can understand varied content if clear.

FAQ is support layer, not architecture

FAQ useful when people ask repeated questions.

But many sites turn FAQ into dumping ground.

80 questions.

Some duplicate.

Some outdated.

Some SEO-generated.

No ownership.

Architecture should place facts where they belong.

Pricing question -> pricing page.

Security -> security.

Implementation -> implementation.

FAQ covers edge cases and orientation.

If every important fact only lives in FAQ, page discovery and maintenance get messy.

AI Mode does not require FAQ schema or FAQ format.

Google explicitly says no special markup or chunking needed for generative Search.

Design for humans.

Search can retrieve.

Comparison deserves dedicated structure

AI Mode is strong for complex comparison.

Website should support comparison honestly.

Not only “we vs competitor.”

Also:

method comparison.

approach comparison.

plan comparison.

build vs buy.

agency vs in-house.

tool vs service.

Comparison page should have criteria.

Trade-offs.

Fit.

Limitations.

Evidence.

This becomes valuable for fan-out subquestions.

If user asks:

“Is AI Optimization different from SEO?”

Comparison source.

If asks:

“Agency vs internal team?”

Dedicated decision page.

Architecture follows decision.

Not product propaganda.

Evidence layer cannot be buried in blog

B2B website often says:

“We have experience.”

Where proof?

Case studies hidden.

PDF inaccessible.

No methodology.

No dates.

AI Mode may search evidence.

Create evidence architecture.

Case study hub.

Research.

Methodology.

Public observations.

Testimonials.

Certifications.

Clients where permitted.

Not all need separate pages.

But easy to discover.

Evidence page should link to service.

Service should link evidence.

Relationship reciprocal.

This helps buyer.

Search also.

Architecture should support query fan-out naturally

Take question:

“Is GEO useful for local businesses in Indonesia?”

Potential subtopics:

definition GEO.

local Search.

Business Profile.

AI Mode.

measurement.

reviews.

location data.

Now map website.

Do we have:

GEO guide.

local AI Search guide.

Business Profile optimization.

measurement.

case evidence.

If yes, fan-out has meaningful pages.

If no, content gap.

This is better than generating “50 fan-out keywords.”

Architecture audit can use customer questions as probes.

Ask question.

Decompose information needs.

Find canonical source for each.

If no source, gap.

If five sources conflict, governance problem.

If source exists but orphaned, linking problem.

Simple framework.

Article depth should depend topic

AI Mode does not demand long content.

A simple answer may need 500 words.

Complex guide 3,000.

POST HUMANIS in this project has its own production length target, but website architecture more broadly should not force uniform length.

Google says no preferred word count for Search success.

Write enough.

No padding.

For architecture, page function matters more than word count.

Pricing page can be concise.

Research report long.

FAQ short.

Case study detailed.

Homepage brief.

Do not use editorial length as technical requirement.

Content should be self-contained enough

AI retrieval may land user directly on deep page.

Don't assume they saw homepage.

Page should explain enough context.

Who is this for?

What topic?

What entity?

What date if relevant?

What next?

But self-contained doesn't mean repeat company boilerplate 500 times.

Use concise context.

Then links.

This balance reduces duplication.

Content architecture needs currentness paths

Every core page should have owner.

Service -> Marketing/Product.

Security -> Security/Legal.

Pricing -> Commercial.

Locations -> Operations.

Research -> Editorial/Data.

When facts change, update source.

Articles can reference.

Architecture becomes living system.

AI Search increases cost of outdated pages because old URL can still be retrieved.

Governance must be built into architecture.

Not added after 1,000 pages.

URLs should reflect stable concepts, not trends

Avoid URL architecture tied to temporary labels.

Example:

/chatgpt-search-2026-best-tips-final-new/

Bad.

If topic persistent:

/chatgpt-search/

Update content over time.

Historical article can have date if news.

Separate evergreen canonical from news.

This helps long-term maintenance.

AI platform changes fast.

Stable URL is asset.

Trend article links canonical.

Canonical links latest official docs.

This allows history + current truth.

Internal linking should mirror relationship, not quota

“Each article must have 7 internal links.”

Wrong target.

Links should help discovery and next question.

From AI Mode article:

link query fan-out.

Search Console GenAI.

content architecture.

technical SEO.

From content architecture:

link evidence.

entity governance.

measurement.

Meaningful graph emerges.

Don't link homepage everywhere.

Deep relationship more useful.

Anchor should describe destination naturally.

No exact-match obsession.

Search needs crawlable links.

User needs orientation.

Architecture can include non-HTML assets, but critical facts should be text

Research PDF.

Video.

Webinar.

Interactive tool.

All useful.

But key information shouldn't exist only in inaccessible format.

Google recommends important content be available in textual form.

If report is PDF, create HTML summary.

If video, transcript or key points.

If infographic, text explanation.

This expands retrieval surfaces.

Accessibility improves.

Users benefit.

Don't hide knowledge in gated PDFs only

B2B often gates everything.

Email required.

Then AI/Search can't access.

Maybe that's business choice.

But if all evidence is gated, public information layer weak.

Use teaser + public summary.

Gate full report if needed.

Explain methodology public.

Provide enough to verify.

AI Optimization forces trade-off:

lead capture vs public discoverability.

No universal answer.

But choose intentionally.

Architecture should separate current, evergreen, and historical

Current:

pricing.

leadership.

features.

locations.

Evergreen:

definitions.

methods.

principles.

Historical:

launches.

old research.

past events.

Each has different update rules.

Don't rewrite historical page to look current.

Don't leave current page stale.

Architecture must encode temporal role.

Dates.

labels.

archive.

current canonical.

This helps AI and humans avoid timeline confusion.

AI Mode also affects CTA architecture

User can arrive deep in decision journey.

Article about procurement should link to:

security.

methodology.

demo.

Not generic homepage.

Comparison page should link evidence.

Evidence should link service.

Local page to booking.

Architecture should support action.

AI Search may compress discovery.

Website must pick up where AI leaves.

This is why content architecture and conversion architecture merge.

What should a GEO.or.id architecture look like conceptually?

Core hubs:

AI Optimization.

Google AI Search.

ChatGPT Search.

Perplexity.

Agentic Web.

Measurement.

Entity & Knowledge Governance.

Industry.

Evidence.

Each hub:

definition.

core concepts.

subtopics.

current updates.

research.

guides.

Then POST HUMANIS adds perspective.

News update.

analysis.

scenario.

conversation.

case.

Not everything needs canonical page.

But repeated foundational facts do.

AI Mode is not chatbot content strategy.

It is Search with generative reasoning and retrieval.

The wrong response is to make every page look like chatbot answer.

The right response is to make website easier to navigate as a knowledge system.

Clear entities.

Clear page roles.

Strong hubs.

Deep pages.

Evidence.

Current facts.

Meaningful internal links.

Stable canonicals.

Historical context.

Action paths.

If AI Mode decomposes one question into ten searches, architecture should not panic and create ten thin pages.

It should already have the right knowledge in the right places.

That's the difference between a content library and a content system.