Context Hints in ChatGPT Ads Aren't Keywords

Describe when your product helps. Do not write a list of people you wish would buy it.

That is the useful distinction in OpenAI’s context-hints guidance. Hints give the ad system business context. They are not exact-match keywords, delivery guarantees, or a substitute for targeting controls.

The distinction that trips people up

Do not strip every place or time detail from a hint because it sounds like targeting. A service area or opening time describes the business. A request to restrict ad delivery describes what you want the advertising system to do. Use supported product controls for the latter, where available; a hint cannot create a control the product does not offer.

Statement	What it is	Where it goes
"Same-day laptop repairs in Bristol"	Business fact (area + service)	Context hint
"Repair desk open until 9pm"	Business fact (hours)	Context hint
"Only show this ad to people in Chicago"	Delivery instruction	Ads Manager location targeting
"Pause delivery outside business hours"	Delivery instruction	Supported delivery controls, if available

Worked example: turning a brief into hints

Say a fictional client, Northside Veterinary Urgent Care, hands you this internal brief:

• Serves pets in [City] and nearby suburbs
• Walk-ins accepted daily, 8am–10pm
• Specializes in emergency stabilization, not routine checkups
• "We want to reach worried pet owners" (marketing's own framing, not a documented offering detail)

Three hints, with facts and assumptions tagged separately so a reviewer can see what's verified versus inferred:

1. [FACT] "Walk-in emergency veterinary care, open 8am–10pm daily, serving [City] and nearby suburbs." — a straight business fact; the city and hours stay in, because they describe the business rather than instruct delivery.
2. [FACT] "Specializes in urgent stabilization rather than routine wellness visits, for pet owners who need same-day care." — combines what the clinic does with who it's for.
3. [ASSUMPTION — verify before use] "May be relevant to pet owners noticing sudden symptoms like limping, vomiting, or lethargy." — the brief never mentioned specific symptoms; this is a plausible inference someone made, not a documented fact, and should be checked against real intake data before it ships.

That third line is the one worth flagging in any review process: it's the kind of sentence that reads like a fact but is actually a guess dressed up as one.

Checklist before you submit a hint

• Does it add something not already obvious from the ad or landing page?
• Does it name a specific feature, area, hour, or situation — not a vague category?
• Is every "who" or "when" phrase something the business actually does, not an audience you'd like to reach?
• Would removing it lose real information, or just remove noise?
• Is it describing the offering, or trying to command who sees the ad? Judge the meaning, not individual words: “only works on Mac” is a product constraint.

---

Method: Full text of one Help Center page read September 19, 2026. The Northside Veterinary example above is fictional, built for this piece — it is not a real advertiser, and the [ASSUMPTION] line is an illustration of a marketing inference, not a claim about pet health signals.

Next action: Pull one ad group's hints and sort each line into FACT or ASSUMPTION. Verify assumptions before using them. Keep genuine service-area and hours facts. Move delivery restrictions into supported controls, where available; do not assume the hint enforces them.