Strategy

How to set up a ChatGPT Ads campaign: context hints and chat cards

Getting your first ChatGPT Ads campaign right starts with choosing the right context hints. This guide walks through the full setup, in the panel and via the API.

Setting up a ChatGPT Ads campaign boils down to three decisions: how much you will spend (campaign budget limit), which conversations you will appear in (context hints) and the most you will pay per click or impression (CPC/CPM max bid). For the bigger picture of what the platform is, start with our What is ChatGPT Ads guide; this article focuses on setup practice.

1) How do you set up the campaign and budget?

When creating a campaign in Ads Manager (ads.openai.com) you set a name and a spend limit: daily or lifetime. Since there is no minimum spend you can start with a test budget; in practice, a daily limit that buys at least 30-50 clicks produces meaningful signal. Budget lives at the campaign layer; bids are defined in the next step, on the ad group.

2) How do you choose context hints?

Context hints are short phrases that decide which conversation topics your ad group competes in. A solid starting structure is two groups: one with 5-8 broad hints describing your category (discovery), another with 4-6 narrow hints catching purchase intent (comparison, 'which tool should I buy' style topics). Write hints in the target market's language; on this US/UK-heavy platform that is mostly English.

Broad vs narrow context-hint setsBroadNarrow10022Impressions1,10,66Clicks1,13CTRIllustrative 30-day test
In an illustrative 30-day test the broad set bought plenty of impressions at 1.1% CTR while the narrow set delivered 3.0% CTR on far less volume. Keeping them in separate groups makes budget shifting easy.
  • Broad hint examples: 'project management', 'team productivity' (high volume, medium relevance)
  • Narrow hint examples: comparison topics like 'best project management tool for small teams' (low volume, high intent)
  • Avoid: generic topics loosely tied to your brand; they burn impressions and depress CTR

3) CPC or CPM, and what should the max bid be?

You pick the billing type on the ad group: CPC if you want clicks, CPM if you want awareness. The max bid is the ceiling you will pay for that event. Because the channel is young, the safest path is entering with a conservative cap, raising it gradually if you cannot win impressions, and locking it once CTR and cost per click settle.

4) How do you write the chat card?

A chat card has four parts: title, 1-2 sentences of copy, an image and a target URL. The critical principle is tonal fit: in the middle of a conversation the card must read like a natural answer to the question being discussed, not a slogan. Use a clear benefit in the title, one concrete value proposition in the body, a clean product image and a directly relevant landing page.

Example chat-card CTR target2,5%CTRIllustrative 30-day test
In example scenarios, cards that fit the conversation see CTRs around 2-3%; below 1% usually signals a context-card mismatch.

After setup: watch the data in one panel

Ads Sensor pulls your ChatGPT Ads campaign daily; its AI tells you which context set works, with reasons.

Join the beta →

5) How does setup work via the Advertiser API?

For programmatic setup you issue an API key in Ads Manager → Settings; each key is scoped to one ad account. The flow is three calls, with money expressed in micros (1 unit = 1/1,000,000):

  1. POST /v1/campaigns: name, status and budget (e.g. lifetime_spend_limit_micros).
  2. POST /v1/ad_groups: campaign_id, the context_hints list, billing_event_type (impression/click) and max_bid_micros.
  3. POST /v1/ads: ad_group_id plus the chat_card creative (title, body, target_url, file_id).
  4. For reporting, GET /v1/ad_account/insights returns impressions/clicks/spend daily.

Once setup is done the real work begins: tracking which hint earns clicks and which card earns CTR. Connect your account in the Ads Sensor beta and that tracking becomes automatic; on the conversion side, pixel and Conversions API logic follows the same principles as Meta's.

Frequently asked questions

What do I need to launch a ChatGPT Ads campaign?
An Ads Manager account opened with a business registered in an eligible country, plus completed brand review. After that it is three steps: a budget-capped campaign, an ad group with context hints and a bid, and a chat-card creative.
How many context hints should I use?
5-10 per ad group is a good start. Fewer constrains volume; too many broad ones buy irrelevant impressions. Keeping broad and narrow hints in separate groups makes the performance gap visible.
How do I set the max bid?
Start week one with a conservative CPC cap; raise it gradually if you cannot win impressions. Once your account average CPC settles, fixing the cap slightly above it keeps costs controlled.
Is the chat-card image required?
The creative schema includes an image (file_id) and it visibly lifts attention. Clean, single-product, text-free images perform best inside a chat interface.
What is the advantage of API setup?
Repeatability and speed for multi-campaign structures, plus versionable budget and bid changes. It also lets you pipe reporting into your own warehouse or a panel like Ads Sensor for automated daily tracking.
How does Ads Sensor help with setup?
Its campaign planner looks at your account history and produces a setup recipe for ChatGPT Ads: budget, context-hint sets, bids and chat-card variants. After launch it tracks performance daily and returns reasoned optimization recommendations.

Launch your first campaign from a recipe

Ads Sensor's campaign planner turns your history into a ChatGPT Ads setup recipe, then tracks performance in one panel.

Join the beta →