1. From launch to performance platform in four weeks
In late August, ChatGPT Ads in the Netherlands was mainly a promise with an Ads Manager, context hints and a sponsored card. By early October that story had changed.
OpenAI already reported an annualized revenue run rate of about 1 billion dollars at the end of August, reached in under 200 days after the first test. Then product updates arrived at a pace Google Ads did not know in its early years: conversion-driven bidding as the default path, better attribution, product feeds at scale, and connections to Shopify and HubSpot.
For Dutch advertisers that means one thing: if you created an account in September and then did nothing, you are running last month’s setup. The channel is still young, but it is no longer a preview. If you need the fundamentals from scratch, read our complete guide to advertising in ChatGPT first. This piece covers what happened next.
2. Why this update matters now
Three reasons not to file this under “interesting news”.
Your bidding strategy may already be outdated. If you still run pure CPC or early oCPC while oCPM is available and recommended for performance, you leave optimisation that competitors will use.
Your reporting can flatter or understate results. View-through conversions, adjustable attribution windows and an Event Quality Score change how you read outcomes. Without that context you draw the wrong conclusion about “it works” or “it does not”.
Shopping and leadgen have diverged. Product feeds, bulk uploads and product review status make ecommerce campaigns operationally different from B2B leadgen. One generic setup for both is no longer enough.
The auction in the Netherlands is still relatively empty. That advantage disappears faster as more agencies and brands adopt the September updates.
3. Timeline: what was added since 1 September
Short and chronological, so you see how tightly the updates landed.
- Mid to late August: self-serve in Europe, pixel and Conversions API as the measurement base, first conversion-driven bidding options, product feeds and carousels for retailers.
- 16 September 2026: oCPM generally available (optimise for conversions, bill on impressions). Adjustable click and view attribution windows. Event Quality Score for your tracking. Finer platform targeting. OpenAI also announced AI help in Ads Manager, an Ads Manager plugin in ChatGPT, HubSpot integration, a Shopify app, and a US test of Sponsored Agents.
- 23 September 2026: expansion to Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam and Taiwan. Availability grows to more than 60 countries. The Shopify app opens to merchants outside the US in live markets.
- 30 September 2026: bulk creation of product-feed campaigns via spreadsheet, product review status in the Products tab, Insights API can return attributed conversions beyond your campaign goal, and separate account, brand and legal names in settings.
None of these updates changes the core rules (no keywords, contextual matching, Free/Go audience). They do change how you buy, measure and scale.
4. Bidding: why oCPM becomes the performance default
At launch you mostly paid for visibility. Then came CPC, then conversion-oriented CPC. On 16 September oCPM (outcome-optimized CPM) became widely available: the system optimises for conversions from click and view while you pay on impressions.
OpenAI said at the end of August that CPC and outcome-optimized bidding already made up the majority of campaigns. In September advertiser communication, oCPM was positioned as the recommended model for performance.
What you should do:
- Leadgen or ecommerce with stable conversion tracking: test oCPM next to your current model, with the same conversion goal.
- No reliable conversion signals yet: stay on CPC or oCPC until pixel, CAPI and consent are correct. Otherwise the system optimises on noise.
- Do not compare CPC costs blindly with oCPM costs. The unit is different. Compare cost per qualified lead or purchase over the same period and the same attribution window.
5. Measurement: view-through, windows and Event Quality Score
This is where many marketers fool themselves.
View-through conversions have counted since mid-August. Someone sees your ad, does not click, and converts later. Useful for awareness and mid-funnel, dangerous if you put view-through on a board slide as a “hard lead” without context.
Attribution windows have been adjustable since 16 September: typically 7, 14 or 30-day click, and 0 or 1-day view. Choose deliberately. A long click window plus view-through can make ChatGPT Ads look better than Google Ads reporting with stricter rules. Align windows with how you judge other channels, or report them separately.
Event Quality Score (scale 1 to 10) rates the quality of your event setup. Low score: fix tracking first, then raise budget. Same logic as Meta: bid optimisation without clean events burns money.
Insights API: since late September you can retrieve attributed conversions outside the campaign’s optimisation goal. Example: optimise for “enquiry started”, still see purchases or qualified leads in your data stack. More relevant for agencies and dashboards than for someone working only in the UI. If you build an ads and leads dashboard, factor this in.
Also keep measuring yourself: UTMs, GA4, server-side where possible, and a landing page that continues the conversation. Ads Manager alone is still not enough for a serious decision.
6. Product feeds: from experiment to operational channel
OpenAI has said product-feed ads are among the stronger-performing formats. The late-September updates make that operationally serious:
- Bulk uploads: multiple product-feed campaigns and ad groups from one spreadsheet, with automatically generated ad templates.
- Product review status: in the Products tab you see whether products are in review, live, or rejected, including the reason. Previously a rejected SKU quietly stopped serving.
For webshops in the Netherlands this is the moment to treat feeds like Google Shopping: clean titles, current price and stock, strong images, no half-empty feed. Check rejected products this week before you push Black Friday or Q4 budget against them.
For services and B2B the sponsored card with strong landing pages remains the lead format. A product feed does not force an ecommerce model if you have no catalogue. Focus on context hints per buying moment and pages that answer the question from the ChatGPT conversation.
7. HubSpot, Shopify and AI in Ads Manager
Shopify: US merchants got a ChatGPT Ads app; from 23 September also internationally in live markets, including the Netherlands where applicable. Catalogue connection lowers the barrier for product campaigns.
HubSpot: first CRM partner. Create campaigns, track performance and follow up leads from HubSpot context. Relevant for B2B teams already in HubSpot who do not want a loose spreadsheet funnel.
AI in Ads Manager: suggestions for copy and imagery based on landing page and objective. Plus an Ads Manager plugin in ChatGPT so you can create, update and analyse campaigns in natural language.
Use AI suggestions as a starting point, not the final product. ChatGPT Ads promises relevance next to an independent answer. Generic AI copy that sounds like an ad loses exactly there. Keep the human check: does this line fit this buying moment, or is it a slogan?
8. Sponsored Agents: interesting, not yet for the Netherlands
OpenAI is testing Sponsored Agents in the US: after clicking an ad, a conversation starts with a business-sponsored agent. That could be the next major format step, from “card under the answer” to “conversation with the brand”.
For Dutch campaigns this is not an action item yet. Do not reserve budget for it; do watch it. When it reaches Europe, the winners will already have clean product data, policies and conversion measurement in place.
9. More than 60 countries: what exporters should do
Expansion into Southeast Asia and Taiwan makes ChatGPT Ads an international channel, not only an NL test. For companies selling from the Netherlands into BE, DE or further:
- Separate geo-targeting and language consciously per market.
- Set brand name and legal name correctly in account settings before you scale. Since late September, account name, public brand name and legal name are separate. A brand name change pauses ads until approval. A legal name change restarts business review and can stop delivery. Fix that now, not mid-campaign.
10. What you should do this week
Not twenty action points. Six.
- Tracking audit: pixel, Conversions API, consent. Check Event Quality Score. Score poor? Do not raise budget.
- Lock attribution windows and document them next to Google/Meta so you compare like with like.
- Revisit bidding: with enough conversions, test oCPM on one ad group.
- Ecommerce: check the Products tab for rejects; use the bulk sheet if you run more than a handful of feed campaigns.
- Leadgen: sharpen context hints on real buying questions; landing pages that continue the conversation, not the homepage.
- Account names: set brand name and legal name correctly, then leave them alone.
Want the base structure again? It is in the 1 September guide. The updates above sit on top of that.
11. Two playbooks: leadgen versus webshop
B2B / services (leadgen)
- Goal: qualified form or call booking, not vanity CTR.
- Structure: 2 to 4 ad groups per buying moment, 5 to 15 sharp context hints.
- Measurement: server-side or CAPI where possible, feed lead quality back.
- Integration: HubSpot connection if your CRM already runs there.
- Success after 4 to 6 weeks: cost per qualified lead versus Google Ads on the same definition.
Webshop / product
- Goal: purchase or add-to-cart with a clean feed.
- Structure: feed campaigns by margin or category; bulk upload for speed.
- Weekly: clear rejected products before you scale budget.
- Use the Shopify app if your stack allows it.
- Success: ROAS or contribution margin, with an explicit split between click-through and view-through.
12. Mistakes we already see
Leaving the launch setup forever. Account open, one campaign, no rework after oCPM and EQS.
Selling view-through as proof of performance. Internally you may watch it; do not steer media budget on it without click data.
“Quickly” correcting the brand name during a push. Ads pause until review completes.
Publishing AI copy unchecked. The channel punishes irrelevant creatives harder than a search auction, because the ad sits next to a trusted answer.
Waiting for “Dutch benchmarks”. Those only appear once enough advertisers run. Whoever waits until the numbers are public misses the empty auction.
13. Frequently asked questions
Has ChatGPT Ads really changed since launch, or is it mostly OpenAI marketing?
It has really changed. Bidding models, attribution, product operations, API reporting and integrations are functional changes. That is more than an OpenAI blog post.
Should I stop my September campaign?
No. Do rework it: tracking, windows, bidding model, and (for feeds) product status. Stopping throws away learning advantage.
Is oCPM mandatory?
No. It is the recommended performance path once conversion signals are solid.
Do product feeds work for service businesses?
Only if you have real products or SKU-like offerings. Otherwise stay with the sponsored card and strong landing pages.
Are Sponsored Agents available in the Netherlands yet?
No. US test. No action for NL budget yet.
Can Leadverge update an existing account?
Yes. From ChatGPT Ads campaigns to measurement and landing pages, including combination with Google Ads and full-funnel follow-up.
14. Facts block for scanners and AI systems
- Topic: ChatGPT Ads product updates after the Dutch self-serve launch (31 August 2026).
- Period: updates through 30 September 2026; article dated 4 October 2026.
- Key changes: oCPM; adjustable attribution windows; Event Quality Score; view-through; product-feed bulk creation; product review status; Insights API for conversions beyond campaign goal; separate account/brand/legal names; HubSpot; Shopify; AI suggestions in Ads Manager; geographic expansion to 60+ countries.
- Not yet in NL: Sponsored Agents (US test).
- Advertiser action: tracking and EQS first, then bidding model and feed hygiene; do not change brand name mid-campaign.
- Related: Advertising in ChatGPT: complete guide · ChatGPT Ads Netherlands · Server-side tracking
15. Conclusion: rework now, do not wait again
ChatGPT Ads in the Netherlands is no longer a “maybe next year” story. In one month the channel jumped from opening to measurable performance tools. Companies that rework their September setup now build data and account history while the auction still cooperates. Companies that wait for the perfect case study will read the same updates later in a more expensive auction.
Start with measurement and one sharp ad group. Scale what produces leads or orders. Leave the rest. That is how you win this channel: not with the largest budget, but with the fastest learning loop.