OpenAI Ads Manager: ChatGPT Ads for Users Who Are Already Hot

Direct answer: do not open ChatGPT Ads only because the channel is new. Pay only for conversations where the user is already hot: the category is chosen, the constraints are named, and the user asks for the next step. Write context hints for that situation, write ads that state the price and the conditions, and measure with the OpenAI Pixel and the Conversions API.

Main condition: this method applies when your legal entity is in an available country and your business category is allowed. Indonesia is on the self-service list since 23 September 2026. Limit: a context hint is not a targeting rule, OpenAI publishes no benchmarks yet, and this article promises no results.

We wrote this article on 24 September 2026 from 3 OpenAI announcements and 17 articles in the ChatGPT Ads collection of the OpenAI Help Center. We have not run a ChatGPT Ads campaign yet, so there are no Ads Manager screenshots. The crawler test uses our own landing page, and every scenario uses dummy data.

Why FOMO is expensive on a new ad channel

On 23 September 2026, OpenAI started to open ChatGPT Ads in Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam, and Taiwan. Ads show only to users on the Free and Go plans. Source: OpenAI announcement ChatGPT Ads expands to Southeast Asia and Taiwan.

A new channel invites a fast decision: launch now, think later. That pattern is expensive in ChatGPT Ads because of 4 facts on OpenAI pages, read on 24 September 2026:

OpenAI writes that ChatGPT Ads has no performance benchmarks across advertisers and industries yet. Source: ChatGPT Ads FAQ.
Delivery and spend can move up and down during the beta because the delivery systems keep changing.
The account country, currency, and time zone cannot change after the account exists. Source: Ads Manager Beta Account Setup.
A daily budget is an average. Spend on 1 day may reach 2 times the daily budget. Source: Create Campaigns for ChatGPT Ads.

An ad launched from FOMO is usually broad: 1 ad group, a hint made of keywords, and the homepage as the landing page. The result is clicks from people who are still learning, and the Clicks objective bills every one of them. Many clicks with few sales is a setup problem, not proof that the channel fails.

Also watch for fake emails in the name of OpenAI Ads. A new launch always attracts phishing. Sign up only through ads.openai.com, typed by you.

How ChatGPT Ads picks the moment to show

Ads show below the answer, not inside it

Ads appear below a ChatGPT answer with a sponsored label. Each unit holds the advertiser name, favicon, title, copy, landing page, and image. Source: Ads in ChatGPT: The Basics.

OpenAI writes that ads do not influence ChatGPT answers, and advertisers cannot read user conversations. Ads do not show on Plus, Pro, Business, Enterprise, and Edu plans, or to accounts under 18. Source: Ads in ChatGPT FAQ for users.

So the user reads the answer first, then sees the ad. An ad that only repeats the answer gives no reason to click.

The system reads the conversation, not keywords

The ads system uses several signals: the context and intent of the current conversation, the landing page, the title, the copy, advertiser context hints, and other ChatGPT signals when ads personalisation is on. OpenAI picks ads through a relevance-weighted second-price auction. Source: Ads in ChatGPT: The Basics.

You write context hints at the ad group level. OpenAI writes that a hint is not an exact-match keyword, not a targeting rule, and does not guarantee delivery in a given conversation. Source: Write Context Hints for ChatGPT Ads.

The intent ladder: learn, compare, decide

OpenAI writes that many ChatGPT conversations are active and decision-oriented: people learn about a category, compare options, or decide what to do next. Source: New ways to buy ChatGPT ads, 5 May 2026. We use those 3 stages as an intent ladder.

Diagram with 3 columns: the Learn stage marked cold, the Compare stage marked warm, and the Decide stage marked hot, each with a sample message, signals, and an ad decision — Skip the Learn stage (left). Use the Compare stage (middle) for a small test only. The Decide stage (right) is the click worth paying for, because the user named the place, the time, and asked for the price.

A hot user shows 4 signals in the conversation:

The category is chosen. The user no longer asks "what is", but "which one".
Constraints are named: budget, place, size, number of branches, or specification.
There is a deadline: today, this week, before a given date.
The user asks for the price, who can do it, or where to buy.

Why hot users click an ad

A hot user does not need persuasion. A hot user needs 3 things: an offer that fits the constraints, a clear price or condition, and a short next step. An ad that gives those 3 things answers the next question in the conversation.

OpenAI asks for clear, specific, benefit-focused copy, not a catchy line or a generic call to action. Source: Create Ads for ChatGPT Ads. Rama Digital recommendation: write ads that also filter, so cold users pass them and do not use your click budget.

Objectives and billing in Ads Manager

Ads Manager offers 3 objectives: Views, Clicks, and Conversions. The objective sets pricing and optimisation, and it cannot change after the campaign exists. Sources: Create Campaigns for ChatGPT Ads and Conversion-optimized Campaigns, read on 24 September 2026.

Objective	You pay	The system optimises for	Use it for hot users when
Views (CPM)	Per 1,000 impressions	Impressions	Rarely. An impression does not tell a cold user from a hot user.
Clicks (CPC)	Per valid click	Clicks	The Pixel has no conversion data yet, and the ads already filter.
Conversions, Clicks billing (oCPC)	Per valid click	Conversions after a click	1 standard event already has enough volume.
Conversions, Impressions billing (oCPM)	Per 1,000 impressions	Conversions from clicks and views	Still in beta. Use it only when measurement is healthy.

The Conversions objective does not bill per conversion. Each Conversions campaign uses 1 standard event, and that event cannot change after the campaign exists.

What to prepare

A legal entity in a country marked Available on the Ads Manager Availability page.
A business category that the Ads Policy Eligibility table allows for the target country.
An OpenAI account with a work email, a logo, and a business name that match the domain.
A credit card for the billing profile, with a complete billing address.
1 landing page for each hot situation, with a price, conditions, and 1 action button.
Access to robots.txt, the WAF or CDN, and the website tracking code.
1 conversion event with value: a qualified lead, a booking, a trial, or an order.

Open the Ads Manager Availability page in the OpenAI Help Center, then find the country of your legal entity. OpenAI writes that the entity that advertises and is billed must be in an available country. On 24 September 2026, Indonesia is marked Available. Source: Ads Manager Availability.

Then check the Ads Policy Eligibility table. Outside the US, categories such as minimally invasive cosmetic procedures, dietary supplements, personal loans, buy now pay later, and cryptocurrency are Not allowed. Financial software and health software are Allowed. Source: Troubleshooting Common Onboarding and Policy issues.

Fill in the country, currency, and time zone with care, because all 3 are locked. A new account can be limited to its home country until verification is complete and the required spend is reached. Business verification may take up to 7 business days.

Verify: your country shows Available, your category shows Allowed for the target country, and the platform access email has arrived.

Step 2: Open the landing page to OAI-AdsBot

OpenAI visits the landing page of every ad with the OAI-AdsBot crawler, for policy review and to understand relevance. OpenAI requires access for OAI-AdsBot and recommends access for OAI-SearchBot. Source: Advertiser Guidance for Allowing OpenAI Web Crawlers.

Check 3 layers: robots.txt, the WAF or bot protection, and any CAPTCHA or JavaScript challenge. OpenAI writes that Cloudflare verifies OAI-AdsBot. The full user agent is on the Overview of OpenAI Crawlers page.

Terminal running curl with the OAI-AdsBot user agent against ramadigital.id: robots.txt allows every bot except GPTBot, the landing page returns 200, and the www redirect keeps the oppref parameter — In our test, robots.txt does not name OAI-AdsBot, so the User-agent: * rule applies (1). The landing page returns 200 for the OAI-AdsBot user agent (2). The oppref and UTM parameters stay intact (3), also after the www to apex redirect (4).

In our test on 24 September 2026, ramadigital.id returned 200 for the OAI-AdsBot user agent. This test only checks user agent rules from our machine, not the OpenAI crawler IPs. For an IP-based allowlist, use the list at openai.com/adsbot.json.

Verify: curl with the OAI-AdsBot user agent returns 200, and the final URL still carries every query parameter.

Step 3: Set up the OpenAI Pixel and the Conversions API before the first ad

Create a data source in Ads Manager, then send events through the OpenAI Pixel, the Conversions API, or both. OpenAI appends the oppref parameter to the landing page URL. The Pixel stores oppref in a first-party cookie so a conversion can be linked to the click. Source: Conversion Measurement.

Diagram of the 6-stage oppref path: ad click, landing page with oppref, OpenAI Pixel, lead or order event, Conversions API, and Ads Manager, with boxes for break points and fixes — Stages 2 and 5 are the weak points. A redirect that drops the query breaks oppref, and 2 different event_id values for 1 sale double the report. The lower right box holds 3 fixes.

Keep oppref through every redirect and form step. Store oppref with the lead in your CRM, then send it back through the Conversions API. When the same sale goes through both the Pixel and the Conversions API, use the same event ID so OpenAI removes the duplicate.

Send conversion data only after you give visitors clear information and get the consent that the law requires. When your leads arrive through WhatsApp, use the same principle as in tracking Click to WhatsApp ads: 1 code per visitor from the click to the payment.

Verify: the URL after a test click holds oppref, the Pixel cookie is set, and the CRM lead row stores oppref.

Step 4: Build 1 ad group for 1 hot situation, then write the context hint

Open the campaign, then create an ad group for each situation that needs a different message or landing page. OpenAI gives the example of a plumbing business with separate ad groups for emergency repairs and preventive maintenance. Source: Create Ad Groups for ChatGPT Ads.

Diagram of an AC repair campaign in Bekasi with 3 ad groups: leak today, routine clean, and new install, each with a context hint and a landing page, and a wrong hint made of keywords — Ad group 1 (amber) catches the hottest users: a leaking AC and a technician needed today. The red box shows a wrong hint: a keyword list without a situation.

Write each hint with the OpenAI pattern: What (features, price, service area), Who (customer need or condition), and When (situation or decision). Write 1 idea per hint as a natural sentence. OpenAI gives the example "After-hours residential plumbing in [city], with an $89 service-call fee for urgent repairs". Source: Write Context Hints for ChatGPT Ads.

For hot users, fill the When part with the decision moment, not the topic. "Same-day AC repair visits in Bekasi for home ACs that leak" is stronger than "AC repair". Do not write instructions such as "show only in Bekasi", because a hint cannot limit place, schedule, or exclusions.

Verify: each ad group has 1 situation, 1 landing page, and a hint that reads as a full sentence.

Step 5: Write ads that filter clicks

Write several title and copy variations for each ad group. OpenAI asks that each variation brings a different angle, a title that states the value clearly, and copy that adds new information. Source: Create Ads for ChatGPT Ads.

Dummy data simulation of 2 ads: generic ad A with a 4 percent CTR, 40 clicks, and 4 bookings, against ad B that names the place, time, and price with a 2 percent CTR, 20 clicks, and 5 bookings — Ad B (right) gets half the clicks, but the cost per booking falls from Rp50,000 to Rp20,000. The price and conditions in the title keep cold users from clicking. These numbers are dummy data.

Put 3 filters in the title or copy: a price or starting fee, a condition (place, size, customer type), and a time limit. The CTR may fall. With the Clicks objective, a CTR that falls because cold users pass the ad saves budget.

Send the click to a specific product or service page, not the homepage. Add your own UTM to the destination URL of each ad. When you turn on the text customisation that OpenAI announced on 16 September 2026, check that the price and conditions still show. Source: Reimagining advertising with AI.

Verify: each ad states at least 2 of the 3 filters, and the title and the destination page name the same offer.

Step 6: Choose the objective, bid, and a controlled budget

Start with the Clicks objective when the Pixel has no conversion data. Move to Conversions after 1 standard event has enough volume. OpenAI recommends a starting maximum bid of $3 to $5 USD per click for CPC campaigns. That page gives no rupiah figure. Source: Ads in ChatGPT: The Basics.

New ad groups use Maximize results by default. This strategy chases result volume and does not guarantee a given CPC or CPA. Choose Manual: Max bid when you have a cost limit per click. Source: Maximize Results Bid Strategy.

The minimum daily budget for an IDR account is Rp14,000. A daily budget is a 7-day average: spend on 1 day stays at or under 2 times the daily budget, and spend in 1 Sunday to Saturday week stays at or under 7 times. A campaign-total budget is the stricter total control. Sources: Create Campaigns for ChatGPT Ads and Budget Pacing.

Exclude existing customers through Tools > Audiences. An exclusion audience may have fewer than 25,000 matched users, while inclusion needs at least 25,000. Source: Set up Custom Audiences for your Campaign.

Verify: the campaign uses 1 objective, 1 country, a daily budget you can afford to lose for 14 days, and an exclusion for existing customers.

Step 7: Read the data with the right delay, then decide

Impressions, clicks, and CTR refresh about every 15 minutes. Spend arrives about 7 to 8 hours late, so a spend of 0 does not always mean no charges. Conversions take 24 to 48 hours. Source: Measure Results.

Timeline with 5 points: impressions and CTR at about 15 minutes, spend at 7 to 8 hours, conversions at 24 to 48 hours, the 7-day budget week, and the 14-day decision point — Points 1 to 3 show when each number arrives. Point 4 shows the budget cycle. Point 5 (amber) is our recommendation for the first decision, not an OpenAI rule.

Ads Manager offers click-through windows of 7, 14, or 30 days, and view-through windows of 0 or 1 day. Attribution is last-touch, not multi-touch. Changing the window changes reporting only, not optimisation or billing.

Compare the Conversions column with sales data in your CRM, not only with CTR. OpenAI writes that Ads Manager and other analytics can differ without an error. Rama Digital recommendation: make the first decision after 14 days, unless a stop criterion in the checklist is reached first.

Verify: the day-14 report holds spend, clicks, Ads Manager conversions, and CRM sales for each ad group.

Sample scenarios: when hot users click an ad

Every name, price, and conversation in this section is a simulation with dummy data. Each scenario shows the user message, why that user is hot, and the ad setup that fits.

Scenario 1: AC repair service in Bekasi

Part	Content
User message	"The 1 PK AC in my bedroom in East Bekasi has been dripping since yesterday. Can a technician come today? Roughly how much?"
Hot signals	Place, AC size, a same-day deadline, and a price question.
Context hint	Same-day AC repair visits in Bekasi for home ACs that leak water or stop cooling, with a Rp50,000 check fee.
Ad	Title: "AC Technician Today in Bekasi, Check Rp50,000". Copy: "Leaking 1-2 PK home AC. Book before 14:00 for a same-day visit."
Landing page and event	A leaking-AC service page with a booking form. Event: lead.

A user who asks "why does my AC leak water?" is still learning. Do not write a hint for that question, because that user often fixes the problem alone.

Scenario 2: Point of sale app for coffee shops

Part	Content
User message	"I run 2 coffee shop branches in Bandung. I need a POS app that deducts milk and coffee bean stock automatically, reports per branch, and costs under Rp300,000 a month. Which one fits?"
Hot signals	Number of branches, required features, and a monthly budget limit.
Context hint	POS app for coffee shops with 2 to 5 branches that need automatic ingredient stock per cup.
Ad	Title: "Coffee Shop POS for 2 Branches, Rp249,000 per Month". Copy: "Coffee bean and milk stock falls automatically per cup. 14-day trial, no credit card."
Landing page and event	A multi-branch plan page with a trial form. Event: sign up.

This ad states the price, so a shop owner with a Rp100,000 budget does not click. The click budget stays for owners who fit.

Scenario 3: Online running shoe store

Part	Content
User message	"I want to train for my first 5K. My feet are wide. Which shoes are cushioned and under Rp1,000,000?"
Hot signals	A clear goal, a foot condition, and a price limit.
Context hint	Cushioned wide-fit running shoes for beginners training for their first 5K.
Ad	Title: "Wide-Fit Running Shoes for Your First 5K, Rp849,000". Copy: "Wide sizes in stock. Free size exchange for 30 days."
Landing page and event	A wide-fit running shoe collection page. Event: order_created.

This hint follows the OpenAI example for beginner running shoes. For a large catalog, use a product feed campaign. New product feed campaigns accept country-level targeting only.

Scenario 4: A business that should wait

An aesthetic clinic in Jakarta wants to advertise fillers because a competitor may enter first. The Ads Policy Eligibility table marks minimally invasive cosmetic procedures as Not allowed outside the US. The ads will be rejected, so the money and time for creative are lost.

The same applies to personal loans, buy now pay later, and cryptocurrency in Indonesia. The right decision for these businesses is to wait for a policy change, and to strengthen other channels.

14-day simulation: AC repair in Bekasi

This table continues scenario 1 with dummy data. Daily budget Rp100,000, the Clicks objective, and 3 ad groups as in the Step 4 diagram.

Time	Input	What is recorded	Decision
Day 1, 09:00	Clicks campaign, 3 ad groups, 9 ads submitted	Ad status: review	Wait. Do not edit the ads.
Day 1, 16:00	Ads marked Serving	120 impressions, 3 clicks	Check that ads serve. Do not judge yet.
Day 2, 08:00	Day 1 spend appears	Spend Rp61,000	Match the spend against the clicks.
Day 3	Day 1 conversions arrive	2 leads in Ads Manager, 2 bookings in the CRM with oppref	Tracking works. Continue.
Day 7	End of the budget week	Spend Rp640,000. Ad group 1: 38 clicks, 9 bookings. Ad group 2: 30 clicks, 1 booking. Ad group 3: 12 clicks, 0 bookings	Add the price to the ad group 2 titles.
Day 14	Decision point	Ad group 1: Rp21,000 per booking. Ad group 3: 25 clicks, 0 bookings	Raise the ad group 1 budget by 20%. Stop ad group 3.

Ad group 3 stops because it reached the stop criterion, not because of a low CTR. This decision uses CRM data, not only Ads Manager data.

Checklist before and after the first campaign

Check the legal entity country and the policy category. Owner: business owner. Evidence: dated screenshots of the Availability and Eligibility tables.
Test the landing page with the OAI-AdsBot user agent. Owner: developer. Evidence: curl output with status 200.
Set up the OpenAI Pixel and the Conversions API with 1 event_id. Owner: developer. Evidence: 1 test conversion with oppref in the CRM.
Write 1 ad group per hot situation. Owner: marketer. Evidence: a What, Who, When hint document per ad group.
Write 3 ad variations with a price, a condition, or a deadline. Owner: marketer. Evidence: the approved list of titles and copy.
Add an exclusion for existing customers. Owner: marketer. Evidence: the audience name with status Ready.
Record spend, clicks, conversions, and CRM sales every week. Owner: marketer. Evidence: a weekly report per ad group.
Stop an ad group that reaches 25 clicks without 1 qualified lead, or a cost per sale above the margin per sale on day 14. Owner: business owner. Evidence: a decision note with numbers.

Frequently asked questions

Can Indonesian businesses use ChatGPT Ads now? Yes. OpenAI started to open ChatGPT Ads in Indonesia on 23 September 2026, and the Ads Manager Availability page marks Indonesia as Available for self-service. The account must still pass verification and review.

What is the minimum ChatGPT Ads budget for a rupiah account? The minimum daily budget for the IDR currency is Rp14,000. Spend on 1 day may reach 2 times the daily budget, and spend in 1 week stays at or under 7 times the daily budget.

Is a context hint the same as a Google Ads keyword? No. A context hint describes the product, the customer, and the situation in a natural sentence. OpenAI writes that a hint is not exact match, not a targeting rule, and does not guarantee delivery.

Can advertisers read user conversations in ChatGPT? No. OpenAI writes that advertisers receive only aggregated performance data, such as impressions and clicks. Advertisers do not see chats, chat history, memories, or personal details.

Are Sponsored Agents available to Indonesian advertisers? Not yet. Sponsored Agents are a limited alpha for selected advertisers, and OpenAI does not accept early access requests. The announcement details are in our article about ChatGPT Ads and Sponsored Agents.

When should I stop a ChatGPT Ads campaign? Set the stop criterion before the campaign goes live. For example, 25 clicks without 1 qualified lead, or a cost per sale above the margin after 14 days.

Next step

ChatGPT Ads is still in beta, so delivery, formats, and rules can change at short notice. Results still depend on the offer, the landing page, follow-up speed, and tracking quality. The background on the new features is in ChatGPT Ads and Sponsored Agents for Indonesian marketers.

When you want every ad click linked to a lead and a sale in your CRM, the Attribution Bridge service sets up the Pixel, the Conversions API, and a visitor code from the click to the payment. When you want to check your account and landing page readiness first, book a 30-minute Initial Consultation.

How to Use OpenAI Ads Manager without FOMO: ChatGPT Ads for Users Who Are Already Hot

Quick summary

Why FOMO is expensive on a new ad channel