Keywords versus context
A search campaign bids on phrases like 'water heater replacement cost'. A ChatGPT conversation sounds different: 'my water heater is 12 years old and the water smells like eggs, should I replace it?'. OpenAI matches ads to the topic of the conversation using context hints, so the planning job is to describe the situations where your service is the sensible next step.
The worksheet
- ServiceName one service line. One campaign per service keeps matching and reporting clean.
- SituationsList the symptoms or events that start the research: a leak, a diagnosis, a letter from the other side's lawyer.
- QuestionsWrite the questions people ask next: cost, timing, repair or replace, do I need a professional.
- ComparisonsNote the options they weigh: DIY, a cheaper alternative, a competitor type.
- AreasDefine where you can serve or are licensed. For law firms this is a hard limit.
- DisqualifiersWrite down what you don't want: jobs too small, areas too far, matters you don't take.
- ProofList what makes you credible: licenses, years in business, guarantees, financing.
An example
| Field | Entries |
|---|---|
| Situations | Old furnace, high winter bills, AC failing, new addition |
| Questions | Is a heat pump worth it in a cold climate? Rebates? Cost to switch? |
| Comparisons | Furnace plus AC replacement, dual-fuel systems |
| Areas | Columbus and surrounding suburbs |
| Disqualifiers | Commercial buildings, rentals outside the area |
| Proof | Licensed, financing available, manufacturer certification |
The same worksheet drives the ad copy and the landing page. If the context plan says people ask about cold-climate performance, the page should answer that first.
Frequently asked questions
Can I still use keywords?
Your keyword research is a good starting point for situations and questions, but the plan should read like conversations, not search phrases.
Sources
- Ad policies · OpenAI, Sep 10, 2026