Context hints describe situations where your offer could be useful. They help ChatGPT Ads match an ad with relevant conversations, but they do not guarantee particular placements or function as exact-match keywords. OpenAI’s explanation of ad matching
The practical skill is describing a customer need clearly enough that you can write a useful ad and choose the right landing page. A large collection of plausible phrases is less useful than a few distinct ideas you can actually test.
Start with the decision, then add the circumstances
Use this writing framework:
Customer task + meaningful constraint + next decision.
For example: “choosing a desk for a small spare room and comparing compact models.” The task is choosing a desk, the constraint is space, and the next decision is which model fits.
This is an editorial framework, not a formula required by OpenAI. You do not need to squeeze all three parts into every hint. Include the details that change which offer or page would be helpful.
| Too broad to guide your draft | More useful planning example | What the page should answer |
|---|---|---|
| Office furniture | Choosing a compact desk for a shared room | Dimensions, storage and available models |
| Cleaning | Comparing weekly cleaning for a small office | Scope, scheduling and quote process |
| Project management | Replacing spreadsheet handovers in a remote team | Workflow, integrations and setup effort |
| Gifts | Finding a practical housewarming gift for a small flat | Suitable products, prices and delivery |
These examples are hypotheses. They are not measured winners or evidence of what a particular person has said to ChatGPT.
For the next drafting step, compare three worked ad examples that connect hints to a title, body and destination.
Give each group one coherent job
For the desk retailer, consider separate groups for compact spaces and shared workspaces. Each can have a different ad and destination. Someone comparing desk depths should see measurements; someone furnishing a room used by two people may need layout options.
Start with a short set you can review together. Five to eight distinct hints can be a manageable drafting exercise, but that is our working method, not a proven optimum or a delivery requirement.
Read every hint against the same ad. If one would need a completely different promise, move it to another group or leave it out of this test. Remove near-duplicates that add no new decision or constraint.
The campaign structure decision table helps distinguish a new customer need from a change that needs its own campaign, such as a separately controlled market or promotion.
The API exposes context hints at ad-group level. It does not turn your editorial group names into guaranteed audience segments. OpenAI’s ad-group reference
Use real targeting settings for operational boundaries
Writing a place name in a hint is not a substitute for selecting the campaign’s geographic settings. OpenAI documents country, region and market targeting; omitting location targeting can allow delivery across available locations. OpenAI’s campaign targeting documentation
Check those settings against where you can deliver or provide service. For a local business, the landing page should explain its service area too. For ecommerce, confirm that the advertised products can be shipped to the selected destinations.
Likewise, do not assume a sentence such as “not for students” acts as a negative keyword. Define unsuitable use cases in your planning notes and inspect the controls actually available in your account. A phrase inside a hint is not a verified exclusion mechanism.
Review the promise behind each hint
Before launch, use four questions:
- Can we help this person? The offer must solve the stated problem.
- Can the page prove it? Remove claims your site does not support.
- Is the next step appropriate? A comparison visitor may need specifications before a sales meeting.
- Can we evaluate the outcome? Decide what would count as a useful enquiry or purchase.
Keep a dated copy of the original hints and a short reason for later changes. If you change hints, ad copy and the landing page together, you may improve performance, but you will struggle to identify which change mattered.
Use your campaign’s own delivery and business outcomes for evaluation. Low volume does not prove that a phrase is wrong, and a high click rate does not establish profitable customer intent. The first-week playbook shows how to separate those questions.
Draft hints from your own landing page
The free AthillyAds hint generator reads a landing page and returns themed hints, an explanation for each group and ad drafts. Review the suggestions against your actual offer before using them; generated text can miss a limitation that your website does not explain.
For the wider workflow, explore campaign planning and preflight checks. If you are starting from scratch, the small-business guide connects the offer, page and budget decisions.