A product name is often a poor starting point for a ChatGPT advertising context hint. It may identify what your team sells without explaining when that offer is useful. In a fictional example, “acoustic panel” names a category. “Sound-absorbing panels for reducing echo in a meeting room” describes a use case, provided the product actually supports that application. The useful work is the translation between those two descriptions.
OpenAI’s targeting documentation says hints can add information about products, use cases and needs. It also makes clear that hints are not exact-match keywords and do not guarantee an ad will appear in a particular conversation. The method here is an editorial preparation process, not a way to buy specific questions. Official targeting documentation
Extract the task behind the category
Start with the product’s current documentation and actual customer questions. Write down the action the buyer is trying to complete. A person comparing a scanner might need to digitise bound books, archive receipts or capture large drawings. The category alone does not tell you which task your offer serves.
For each possible task, identify the feature that makes the offer relevant. A flatbed size, supported material or installation method may matter. Do not add a use case simply because buyers sometimes associate it with the category. If a scanner cannot handle bound books safely, that task does not belong in the hint set for that offer.
Use a small worksheet with four fields: product, task, constraint and supporting fact. Keep the supporting fact even if it will not appear in the final hint. It lets a reviewer understand why the wording was chosen.
Describe one useful relationship
A useful draft connects an offer to a situation. You do not need to tell the customer’s entire story. The following examples are fictional editorial examples and assume the stated product facts have been checked.
| Category-only draft | Use-case draft | Fact to verify |
|---|---|---|
| Document scanner | Scanning large technical drawings without folding the original | Supported document size and handling |
| Folding table | Temporary work surfaces for rooms that need to be cleared after sessions | Folding mechanism, stability and intended use |
| Project software | Assigning review responsibilities before a design is sent to a client | Actual review assignment workflow |
| Outdoor lighting | Lighting a covered entrance where mains power is unavailable | Power source and approved installation conditions |
The revised versions are not inherently better for performance. They are easier to review because the relationship between the task and the offer is explicit. If the product check fails, revise the wording rather than retaining the attractive scenario.
Separate a feature from the work it supports
A common shortcut is to put a customer label before the category: “scanner for architects”. That identifies the person the writer has in mind without explaining the task. An architect could need to copy loose drawings, preserve sketches in a bound book, or capture material samples. The occupation does not establish that one scanner suits all three jobs.
Instead, write a sentence with a concrete action and object: “capture loose drawing sheets”, “digitise pages in a bound sketchbook”, or “reproduce material samples”. Then add the context needed to understand the task. These are candidates to investigate, not three finished hints to save automatically.
For the fictional large-format scanner, loose drawing sheets might be an intended use, while bound books are ruled out by the feeding mechanism. Material samples require a separate assessment of the reproduction the buyer needs. One category name has become three distinct questions, and perhaps only one can proceed. The number of possible tasks is not a measure of quality.
When a feature cannot be connected to a customer task, it does not have to become a hint. An internal model name may matter when ordering but add little context about use. Keep that identifier in the worksheet’s product field so the correct offer is reviewed without turning the hint list into a copy of the catalogue.
When several practical conditions matter, the customer constraints guide helps identify which one determines suitability. First establish the task the product will support.
One category, three different tasks
- Loose drawing sheets
Possible use case if sheet size and feeding method fit.
Checkpoint - Bound sketchbook
Remove if the mechanism requires loose sheets.
Reconsider - Material samples
Needs separate evidence for the reproduction the buyer requires.
Checkpoint
Check the surrounding ads before saving
Hints apply at the ad-group level in OpenAI’s documented campaign structure. Review the actual ads that will share the list. A use case can be accurate for one product while misleading for another ad in the same group. Official campaign management documentation
Imagine a group with a lightweight portable scanner and a large-format office scanner. A large-drawing hint should not be carried across merely because both products scan documents. Decide whether the ads share a coherent use case or whether the editorial brief needs separating before setup.
Check the advertised offer and the destination’s description of it as part of that review. You are checking consistency, not redesigning a page. The hint must not introduce a capability that the advertised product lacks or that only another product in your range provides.
Select a use case you can explain from end to end
Take one candidate and follow the relationship from product to task. In the fictional worksheet, the row could read: “large-format scanner, capture loose technical drawings, preserve the original without folding it, documented sheet size and intended feeding method”. A reader should understand why this product helps. “Efficient document management” does not expose the same relationship for review.
Test the reverse direction as well. If the buyer wanted to scan a bound book, would you still propose this product? If the answer is no, you have found a meaningful boundary between tasks. Retain the rejected task in the worksheet notes. Do not rescue it by describing the whole category more broadly.
A possible final candidate is “Scanning loose technical drawings that need to remain unfolded”, provided the product check supports it. That describes a task, rather than a particular person or a question that must appear in ChatGPT. The wording does not establish demand or promise better advertising results.
Hand over the product, the selected task and the reason they belong together. Identify the ad and destination that present the offer. A reviewer should be able to distinguish a wording change from an entirely new use case that requires fresh evidence.
Continue with the hint evidence check when candidates need factual review. Use the hint specificity guide when the task is selected but the sentence needs shortening or clarification. The translation is complete when another person can explain both what the product is used for and why the selected product fits that task.
The selected customer task becomes part of the campaign’s wider relevance planning. The targeting overview connects use cases with the other decisions to make when preparing the advertising.
Sources and scope
Translate a product name into a concrete customer task and select a use case the advertised product actually supports.
Working methods and examples are editorial suggestions. Check current platform requirements and available features before implementation.