How to write context hints for ChatGPT Ads
Write context hints around a real customer need and a verified product detail. See practical rewrites, an ad-group method and the limits of hints.
Write a context hint as a plain description of a customer need your offer can actually meet. Add a detail that explains why the product fits that situation. For example, a desk's shallow depth can matter to someone furnishing a narrow room; “best desk, home office, buy now” gives the system much less useful information.
Context hints are additional business information supplied at the ad group level. OpenAI's current writing guidance says they inform relevance but do not guarantee an audience or conversation, require exact wording, or enforce geography, schedules or exclusions.
Separate product context from instructions to the platform
“Two-hour upholstery workshops in Portland for adults learning to repair a chair seat” describes an offering. “Only show this to Portland residents on Saturdays” asks for delivery restrictions. Put restrictions in supported campaign controls where available; a sentence in a hint cannot enforce them.
A hint also cannot make ChatGPT recommend your business inside its answer. OpenAI states that ads and organic responses operate independently. Asking the model to “always recommend our brand” has no place in this field.
Hints are one input to matching. OpenAI also considers the landing page and ad copy, among other signals. A careful hint cannot rescue an ad that promises one thing while its destination sells another. See the official selection and delivery explanation.
Turn a product fact into a useful buying situation
Start with the offer in this ad group. On a working sheet, put a verified fact next to the buyer problem it solves. Then write the connection in ordinary English. This is our suggested drafting method, not a required OpenAI template.
- Choose the product or service the visitor will see after the click. Use that page as your starting point.
- Find a fact that could change a buying decision: dimensions, supported workflow, available appointment type, delivery terms or a genuine price condition.
- Describe the situation in which that fact helps. Ask what the customer is trying to accomplish, not which flattering audience label you want to reach.
- Write one focused phrase and check every promise against the offer. Remove details you cannot substantiate.
A useful working pattern is “Offering with [verified detail] for [specific need].” You can also lead with the situation. There is no need to force every hint into the same grammar or fit a complete customer biography into one sentence.
A worked example: a desk for a narrow room
Assume the retailer sells a desk that is 80 cm wide and 45 cm deep, with room for a laptop and delivery within the contiguous United States. These are invented product facts for the exercise.
Its first draft is “Home office furniture for remote workers.” That covers a large category and says nothing about why this particular desk fits a buyer's problem.
A more useful hint is “An 80 cm-wide, 45 cm-deep laptop desk for fitting a work surface into a narrow bedroom.” It connects dimensions to a constraint the product can solve. It does not claim the desk fits every small room or promise that a particular prompt will trigger an ad.
The ad could discuss the compact footprint, and the destination should show the same desk with measurements and delivery terms. If those dimensions change, revise the hint. The unit choice here follows this hypothetical product specification; use the measurements your customers actually see.
Four rewrites, with the reason for each change
All businesses and offers below are hypothetical. Treat the improved versions as examples of reasoning, and replace their facts with your own verified details.
Appointment software: describe the workflow it supports
Weak: “Booking, calendar, reminders, scheduling software.”
More useful: “Appointment software with text reminders for independent salons that lose bookings when clients forget their time.”
The rewrite connects one supported feature to a business problem. Before using it, verify that text reminders are included in the advertised plan and work in the intended market. If reminders cost extra, the ad and destination should make that condition clear.
A separate hint about coordinating staff across several locations would need its own product evidence. Do not add it simply because larger salons would be attractive customers.
Meal delivery: name an occasion the offer can serve
Weak: “Busy people who want healthy meals.”
More useful: “Prepared dinners delivered chilled for people who get home late and want a meal they can heat in five minutes.”
The revised hint gives the reader's situation a concrete product answer. Check the preparation time and delivery coverage. Avoid adding dietary or health claims just to widen the audience; “supports weight loss” needs different evidence from “heats in five minutes.”
Local workshops: describe the service area without pretending to restrict delivery
Weak: “Show only to beginners in Portland on weekends.”
More useful: “Beginner upholstery workshops in Portland with tools supplied for repairing a dining-chair seat.”
This describes a local offer and the task taught. It removes instructions the hint cannot carry out. The advertiser must separately confirm that supported location controls fit the service area. Keep “tools supplied” only if the workshop price includes them.
Bookkeeping support: replace a desirable audience with a relevant problem
Weak: “Successful founders ready to grow.”
More useful: “Monthly bookkeeping for small online stores that need payment-processor fees and refunds reconciled before reviewing cash flow.”
The customer need is specific enough to check against the service. An agency offering only annual tax preparation should not use this hint. An owner with plenty of revenue but no reconciliation problem is not automatically a match.
Decide what belongs in one ad group
OpenAI recommends grouping related ads around a shared product or need, and separating use cases when they require different messaging or destinations. You can test copy variations within a group. Read the ad-group guidance.
For the hypothetical compact desk, “a narrow bedroom” and “a small apartment's laptop corner” could share an ad group if the same product page answers both needs. A standing desk for frequent sitting-to-standing changes would deserve separate consideration if it uses a different offer and page.
Do not create a new group for every synonym. “Tiny room” and “small room” do not establish two different buying problems. Equally, avoid combining workshop enrollment and professional furniture restoration merely because both involve upholstery.
Review hints before launch, then evaluate the whole ad group
Before launch, remove the brand name and read the hint on its own. Ask a colleague which customer problem it addresses and what makes the offer suitable. If they answer only “people looking for quality,” revise it.
- Check availability, prices and feature conditions against the current offer.
- Read the hint beside the ad and destination. Look for a promise that appears in only one place.
- Move show/hide instructions into supported delivery controls; remove them from the hint.
- After saving, review the stored wording in the account. Keep a dated copy so later edits have a clear baseline.
Once the campaign runs, compare business outcomes for the ad group with its spend. If you change the hint, copy, page and bid together, you will have little basis for attributing a result to the hint. Where practical, make one meaningful change at a time and let enough outcomes accumulate to assess it. A handful of clicks cannot prove a rewrite improved relevance.
Do not interpret an aggregate performance change as proof that the ad appeared in the exact conversations you imagined. Delivery and conversion performance also depend on the offer, other matching signals and the visitor's next step.
Check the destination and measurement before launch; our campaign preparation process explains those steps. If you are still deciding whether to fund a test, the business-fit guide includes a customer-acquisition calculation.
Sources and review date
Official documentation checked on 2026-10-12. Availability and features can change; confirm them before launch.