ChatGPT Product Feed Tidak Sama dengan Organic Product Discovery

Feed berhasil di-upload.

100.000 product accepted.

Dashboard green.

Merchant berkata:

“Berarti produk kita sekarang otomatis muncul organic di ChatGPT?”

Belum tentu.

Ini salah satu distinction paling penting dalam commerce AI.

Product feed adalah structured data supply.

Organic product discovery adalah outcome di pengalaman ChatGPT Search/commerce.

Ads product delivery adalah paid outcome lain.

Feed membantu OpenAI memahami product record.

Tetapi upload feed bukan guarantee bahwa setiap product akan ditampilkan, direkomendasikan, diklik, atau dibeli.

OpenAI product specification bahkan memisahkan flags:

`is_eligible_search`

`is_eligible_checkout`

`is_ads_eligible`

Tiga fungsi.

Tiga eligibility.

Satu product bisa searchable.

Tidak checkout-enabled.

Tidak ads-enabled.

Atau berbeda sesuai registered feed configuration.

Ini menunjukkan architecture OpenAI commerce sendiri tidak menganggap “product in feed” sebagai satu status universal.

Mulai dari product feed sebagai data layer

Feed berisi:

item ID.

title.

description.

brand.

URL.

image.

price.

availability.

variant.

identifiers.

category.

attributes.

OpenAI menggunakan data ini untuk membangun normalized product records sesuai schema dan registered configuration.

Tujuannya:

identify product.

understand properties.

link to seller.

know current state.

Itu foundation.

Kalau data buruk:

title spam.

price salah.

stock stale.

variant kacau.

image mismatch.

organic discovery dan ads sama-sama punya problem.

Feed quality is shared infrastructure.

Tetapi discovery still depends on system choosing product for user context.

Feed accepted does not mean selected.

Organic Search eligibility berbeda dari Ads eligibility

OpenAI schema has `is_eligible_search`.

This controls whether product can be surfaced in ChatGPT search results for non-Ads feed behavior.

Ads uses `is_ads_eligible` as per-product Ads processing control in the stable schema, while Ads feed behavior also depends registered feed configuration.

The Ads product feed guide says Ads feeds use same base schema and add Ads eligibility requirement.

Still, Ads eligibility doesn't guarantee serving.

Campaign.

ad group.

ad.

budget.

review.

availability.

auction/relevance.

All matter.

Likewise search eligibility isn't a guarantee user sees product.

Eligibility is permission.

Selection is outcome.

This distinction prevents merchant from making fake KPI:

“100% products organic-visible.”

No.

Maybe 100% feed accepted and search-eligible.

Visibility must be observed/measured separately.

Google-compatible ingestion adds another nuance

OpenAI supports a Google-compatible file profile in certain registered feed setups.

Core fields map:

id.

title.

description.

link.

image.

availability.

price.

brand.

variants.

Identifiers.

But compatibility isn't identical behavior.

OpenAI normalizes according its system.

Some input fields are accepted but not stored.

Some unrecognized columns dropped.

For accepted products in the documented compatibility flow, OpenAI can enable search and disable checkout under configured behavior.

Merchant should not assume Google Merchant settings copy one-to-one.

Read OpenAI validation report.

Registered feed configuration matters.

Feed format compatibility is not platform semantics compatibility.

Stable item ID matters for both paid and organic

If SKU changes every day, OpenAI sees unstable records.

Product history breaks.

Delta update difficult.

Variant mapping breaks.

Use stable merchant product/variant ID.

Same principle across channels.

One product record should have persistent identity.

Title can update.

price changes.

stock changes.

But item ID stable.

This is catalog governance.

If ecommerce team cannot maintain stable SKU, AI feed isn't main problem.

Core commerce data model is.

Organic discovery may use public web too

ChatGPT Search is broader than merchant feed.

OAI-SearchBot can surface public websites.

Reviews.

Publisher articles.

Product pages.

Comparison.

Third-party sources.

Feed is one structured route for product information.

Organic product discovery can therefore involve more than feed record.

A user asks:

“best trail shoes for wet weather under $150.”

ChatGPT may use product data plus web sources.

Merchant feed provides:

price.

stock.

variant.

Product page.

Review gives:

real-world grip.

Publisher test gives:

durability.

Brand site gives:

spec.

Organic discovery is information ecosystem.

Feed is catalog truth.

Don't treat feed as replacement for product page/content/reviews.

Product page still needs useful information

Feed description max allows rich text, but user click lands page.

Page should have:

specification.

size.

material.

compatibility.

warranty.

shipping.

returns.

photos.

video.

reviews.

care.

availability.

Feed can't carry every user decision context.

And organic Search may crawl page independently through OAI-SearchBot.

Good merchant strategy:

feed for structured facts.

page for detailed context.

third-party ecosystem for experience/reputation.

Ads for paid reach.

Separate layers.

Organic discovery isn't “free ads”

If product appears in ChatGPT Search organically, that is not sponsored.

If product appears through ChatGPT Ads, it should be labeled sponsored.

OpenAI states ads are separate from chat answers and do not influence responses.

Therefore reporting must distinguish:

Organic product presence.

Paid product ad impressions.

Feed acceptance.

Referral.

Conversion.

Don't combine.

A product could be organic-visible without active Ads campaign.

A product could serve as ad but not be naturally surfaced in sampled organic query.

These are normal.

Feed processing report is not marketing report

Accepted/rejected rows answer:

Can OpenAI parse and normalize data?

Errors:

missing field.

bad URL.

invalid currency.

bad price.

bad image.

identifier.

This is data quality.

Not customer demand.

Marketing performance report answers:

impressions.

clicks.

conversions.

Organic observation answers:

product appears in Search context.

Different dashboards.

Don't show “accepted 99%” as “AI visibility 99%.”

That's misleading.

Feed acceptance is readiness KPI.

Discovery is visibility KPI.

Ads delivery is media KPI.

Product feed can enable search but selection remains dynamic

This is core.

`is_eligible_search=true` means product may be surfaced under applicable product feed configuration.

It does not mean every relevant query includes it.

User query.

market.

availability.

price.

product fit.

system relevance.

source quality.

Other merchants.

All can affect.

OpenAI doesn't publish a deterministic organic product ranking formula in the feed spec.

Do not invent.

Optimize data quality and product truth.

Observe outcomes.

Don't keyword stuff title.

Product title should identify product clearly.

No “BEST CHEAP TOP CHATGPT RECOMMENDED”.

That harms usability.

Organic discovery is user relevance problem, not feed spam competition.

Ads feed also has separate serving conditions

OpenAI's Ads product feed guide says uploading feed does not automatically advertise every product.

Product must:

be Ads eligible.

remain available.

have usable data.

pass processing/review.

Campaign/ad group/ad active and funded.

Then Ads system chooses eligible product and runs through normal ad delivery/auction mechanisms.

So merchant cannot say:

“Feed uploaded = all catalog advertised.”

No.

The same mental discipline should apply organic.

Feed uploaded = data available, subject to configuration and selection.

Not guaranteed surface.

Search vs checkout vs ads flags should be owned commercially

Who decides `is_eligible_search`?

Merchandising.

Legal.

Who decides checkout?

Commerce ops.

Who decides ads?

Media/commercial.

Don't let developer default all true.

Some products:

out of region.

low stock.

regulated.

low margin.

discontinued.

not eligible.

Flags should follow business policy.

This is structured governance.

If one field says eligible but landing page says unavailable, system mismatch.

Automation without policy is dangerous.

Market configuration matters

OpenAI feed docs mention registered feed configurations for market-specific processing, currencies, target countries.

Indonesian merchant shouldn't upload IDR and assume accepted globally without confirmed configuration.

Current program availability may be limited.

For Ads specifically, Indonesia advertiser availability is not currently listed as available as of 9 August 2026.

So Indonesian merchants should distinguish:

feed technical readiness.

program/account availability.

target market.

Don't make public claim “our products are live in ChatGPT Ads Indonesia” without actual access.

Read official availability.

Feed docs can exist before every market is open.

Organic commerce programs may also have onboarding requirements.

Verify.

Product URL must work outside feed

Feed says URL.

User clicks.

OAI-AdsBot validates for ads.

OAI-SearchBot may crawl public page for Search.

If URL:

404.

403.

geo blocked.

login.

redirect loop.

feed record cannot save journey.

Product feed and web infrastructure must agree.

This is why catalog integration has three validation layers:

schema validation.

HTTP validation.

human/agent conversion.

All necessary.

Price and availability are temporal facts

Feed today:

in stock, 1.299.000.

Tomorrow:

out.

If feed not updated, organic answer may show stale info depending system update behavior.

Ads can promote unavailable product.

Bad.

Use delta updates/API according integration.

Define SLA:

inventory every X minutes.

price after promotion.

full feed daily.

No universal.

Business velocity.

For flash-sale merchant, hourly is slow.

For furniture made-to-order, daily enough.

Operational accuracy matters more than article volume.

Organic discovery also needs reputation context

A merchant feed can say:

“waterproof shoe.”

But user asks:

“is it actually waterproof in heavy rain?”

Reviews matter.

Independent test.

Return rate.

Brand reputation.

Feed claim alone may not satisfy.

Organic ChatGPT can search web.

This is why merchants should invest in:

authentic reviews.

clear product specs.

publisher outreach.

customer support.

FAQ.

No fake review.

No synthetic third-party site.

Product feed gives first-party facts.

Public ecosystem gives evidence and experience.

AI shopping likely blends.

Measurement framework

Data readiness:

accepted rows.

rejected.

freshness.

price accuracy.

stock accuracy.

variant accuracy.

Search eligibility count.

Ads eligibility count.

Organic visibility:

controlled ChatGPT product queries.

brand/product mention.

owned link.

product surfaced.

Paid:

ad product impressions.

clicks.

spend.

conversion.

Business:

revenue.

margin.

returns.

repeat purchase.

This stack avoids metric confusion.

Don't optimize product discovery if returns explode.

Fit matters.

One useful test

Take 20 hero products.

For each, verify:

feed accepted.

stable ID.

price current.

stock current.

image right.

URL 200.

variant mapping.

search eligibility intended.

ads eligibility intended.

Then run organic ChatGPT Search questions realistic for buyers.

Do products appear?

If not, don't immediately rewrite title 20 times.

Check:

program availability.

query fit.

market.

competition.

public page.

third-party sources.

feed freshness.

Then iterate.

AI commerce should be tested like product discovery, not keyword density.

What if feed says product eligible but organic never appears?

That can happen.

Eligibility is not guarantee.

Do not escalate as bug unless official expected behavior says otherwise.

Collect evidence.

Maybe search integration not enabled for merchant.

Maybe market.

Maybe relevance.

Maybe system.

Support can investigate configuration.

But don't promise client “product feed guarantees organic placement.”

It does not.

What if product appears organic without feed?

Possible through public web/Search routes depending product and system.

OAI-SearchBot can discover website.

Third-party source can mention.

Feed isn't only possible visibility route.

But structured feed gives OpenAI cleaner first-party product data for commerce integrations.

Think of it as high-quality data channel.

Not exclusive gate to all ChatGPT knowledge.

The opening merchant saw 100.000 accepted rows and thought marketing objective complete.

Actually data engineering objective was complete.

Now marketing starts.

Which products are relevant?

Which queries?

Which markets?

Which get clicks?

Which convert?

Which get returned?

Product feed is infrastructure.

Organic product discovery is behavior.

Ads is paid distribution.

Checkout is transaction.

Each stage requires different controls.

The merchant who understands this won't celebrate the wrong dashboard.

And they won't blame the feed when the real problem is product fit, market availability, stale stock, weak landing page, or nonexistent customer evidence.

Structured data makes products understandable.

It does not make them automatically desirable.