Write ChatGPT Ads context hints as specific product-use situations supported by the landing page. Keep that relevance brief separate from audience eligibility, bid preferences and geographic settings. GPTAd recommends reviewing each hint against the actual offer, saving the existing configuration before changes and treating hints as context rather than exact-match keywords or guaranteed placements.
Reviewed October 15, 2026. GPTAd is an independent publisher. The checklists and fictional examples below are editorial guidance.
OpenAI describes context hints as product and use-case context, not exact-match keywords or guaranteed triggers. Audience inclusion restricts eligibility; an audience bid multiplier changes a fixed bid without itself excluding nonmembers. Targeting guide
That distinction creates two separate jobs: explain relevance clearly, then decide who is eligible. Mixing those jobs makes it harder to review what the campaign is trying to do.
Write from the offer outward
Start with a plain-language offer statement. A fictional retailer, North Window, sells compact grow lights for indoor herbs. Its planning statement might be: “A small light designed for customers growing culinary herbs indoors where natural light is limited.” This is a hypothetical product, not a performance claim about a real advertiser.
From that statement, draft a short set of possible context hints:
- Planning an indoor herb-growing setup in a small kitchen.
- Comparing lighting options for herbs near a shaded window.
- Looking for a compact grow-light arrangement for a countertop.
These examples describe situations. They do not guarantee placement in those situations. They also do not claim the product solves every indoor gardening problem.
Compare them with a weak draft: “gardening, food, apartments, savings, best light, productivity.” The weak version combines a category, broad interests and unsupported praise. It leaves the reviewer guessing which need the product actually meets.
Use the evidence test
For each proposed hint, ask three questions:
- Does the destination page support this use case?
- Can the product owner explain the connection without adding an unstated feature?
- Would a customer recognize the same offer after clicking?
If a hint refers to outdoor growing but the fictional product is designed only for indoor use, remove it. If the landing page never explains size or placement, either improve the page or narrow the hint. Do not use targeting language to compensate for missing product information.
An internal worksheet can have four columns: customer situation, proposed hint, supporting page section and reviewer. This is an editorial planning tool, not a required platform schema. It makes vague assumptions visible before they become configuration.
Separate eligibility from bid preference
Suppose the fictional retailer wants one campaign available only to an eligible included audience. That is a different instruction from preferring that audience while leaving others eligible. Before configuring anything, write the intended behavior in a complete sentence. Avoid ambiguous requests such as “target our existing audience more.”
Inclusion and multipliers require 25,000 matched users; exclusion has no minimum. Uploaded rows are not matched users. Exclusions prevail over inclusion when membership overlaps, and the same audience ID cannot appear in both lists. Targeting guide
Treat audience readiness as a separate checkpoint. A spreadsheet's size does not establish the platform's matched count. Do not restructure a campaign around an audience assumption that has not been verified in the account.
Review geography and platform independently
An offer's service area and its destination's usability deserve their own review. Ask where the business can actually fulfill the offer, then test the page on the platforms the campaign is intended to reach.
OpenAI distinguishes granular web platforms from native apps; its documentation includes desktop_web, ios_web and android_web, with web as shorthand for web platforms. Platform targeting
Do not carry over another advertising network's device labels by habit. The operational question is whether the selected values match the intended customer journey. A page that is comfortable on a laptop may still have an awkward booking form on a phone.
Make changes with a saved comparison
Updating context_hints replaces the list; omitting it leaves the existing list unchanged. Other targeting updates can also replace criteria, so retrieve the current settings before editing. Targeting guide
Keep a before-and-after note stating what changed and why. For North Window, the reason could be removing an unsupported outdoor use case while preserving the approved indoor situations. Review the resulting configuration, rather than assuming the submitted change preserved everything else.
Finally, evaluate the whole offer. ChatGPT ad selection uses contextual relevance alongside other documented signals. A hint is one input, not a purchased keyword position. Ads basics The strongest preparation is a specific product explanation, a credible destination and deliberate eligibility settings.
