Platform information: ChatGPT Ads are developing and platform requirements can change. This article was last reviewed on 11 September 2026. Shays Media rechecks material platform information before relying on it for client decisions.
The misconception worth correcting first
A context hint is not a traditional keyword list or exact-match trigger. That distinction causes more confusion in ChatGPT Ads setup than almost anything else, because it's tempting — especially for anyone coming from search advertising — to treat the context hint field the way you'd treat a keyword match type: list the terms you want to trigger against, add some variants, done.
OpenAI's own description allows a hint to reference the conversations, topics or even keywords a product relates to — but it's explicit that hints aren't exact-match controls and don't guarantee delivery. Shays Media therefore avoids treating the field as a pure term list; doing so loses the situational detail OpenAI's guidance is asking advertisers to provide.
What a context hint is
Context hints sit at the ad group level — not the campaign, not the individual ad. Their purpose, as OpenAI describes it, is to help the system understand the conversations, needs and topics for which that ad group may be relevant.
A good hint describes the product or service, who it helps, or when it's useful — written as a natural phrase, the way a person might actually think or talk about the situation, rather than a fragment optimised for matching. It should describe genuine use cases: things the product or service actually addresses, not a wish list of scenarios where it might theoretically apply.
What a context hint is not
Three things worth being explicit about, because each one gets assumed incorrectly on a regular basis:
- It's not an exact-match control. Writing a hint doesn't mean your ad only appears when that exact phrase shows up in a conversation, and it doesn't mean it will appear whenever something close to it does.
- It's not an audience-targeting rule. A hint describes a situation or need, not a demographic or segment. It isn't a substitute for targeting settings, and it doesn't restrict delivery to people who match some profile.
- It doesn't guarantee delivery. Writing a hint that precisely describes your ideal customer's situation doesn't mean your ad will show for that situation, every time, or at all. It's one signal among several that OpenAI's system considers alongside conversation context, landing page, title and copy.
That last point matters most for setting expectations. Nobody can promise you that a context hint will "work" in the sense of guaranteed delivery, because that's not how the mechanism is described to function.
Writing hints that describe a real situation
Two invented examples, purely illustrative — not OpenAI-provided examples, and not a Shays Media client case study:
A UK-based business offering mobile dog grooming might write hints like "looking for a dog groomer who comes to your house" or "trying to find pet grooming without taking the dog to a salon" — phrases describing an actual need and how someone might be thinking about it, rather than "dog groomer," "mobile grooming," "pet care" as a disconnected list.
A business selling an online course — say, one teaching small business owners how to build their own spreadsheets and reporting — might use something like "wants to stop paying someone else to build their reports" or "trying to learn spreadsheet skills for their own business" — again, a situation, not a term.
Notice what both examples avoid: single words, generic category labels, and anything that reads like it was written for a matching algorithm rather than a person. The test worth applying to a draft hint is whether it sounds like something a customer might actually say or think, not whether it covers as many angles as possible.
When to split into separate ad groups
If a business genuinely serves more than one distinct need — different products, different customer situations, different messaging — that's usually a signal to use separate ad groups rather than one ad group trying to cover everything. OpenAI's own guidance points the same way: meaningfully different products or use cases that require different messaging or landing pages should sit in separate ad groups, each with its own hints.
The mobile dog grooming example above and, say, a boarding/daycare service from the same business would likely warrant separate ad groups — different customer situations, probably different landing pages, and hints that would otherwise blur together if forced into one group.
How hints, ad copy, offer and landing page fit together
A context hint doesn't do useful work on its own. It's one part of a chain that needs to hang together: the hint describes the situation, the ad copy should speak to that same situation, the offer behind the ad should actually match what's promised, and the landing page needs to deliver on all of it without a gap.
If the hint describes someone wanting to learn a practical skill without paying for it to be done for them, but the ad copy is generic branding and the landing page is a homepage with no course information, the disconnect isn't really a context-hint problem — it's a congruence problem across the whole chain. Getting the hint right matters, but it isn't sufficient by itself, and it isn't something Shays Media can claim an independent, proven causal performance effect for — it's one input among several. We cover the landing-page side of this in more depth in What Makes a Good Landing Page for ChatGPT Ads?.
Why Shays Media starts with the situation, not a word list
Our approach to setting up an ad group starts with a genuine question: what is this customer actually trying to work out or achieve, and at what point would this offer sensibly enter that thinking? That's a different starting point from generating a list of terms someone might search, and it tends to produce hints that read like real situations rather than keyword exports with the spaces changed to phrases.
It's slower than brainstorming a term list, and it should be — a context hint that doesn't describe anything real isn't doing its job, however plausible it looks in a spreadsheet.
Check Your Fit — or, if you're already through the Fit Check and setting up context hints for a live account, our Setup and Management service builds these as standard rather than as an afterthought.
A few questions people ask at this stage
Do ChatGPT Ads use keywords? Not in the search-advertising sense. Context hints can reference topics or terms related to a product, but they're written as natural-language descriptions of situations and needs, they aren't a traditional keyword list, and they don't work on exact match.
Can you guarantee my ad appears for a particular conversation? No. Context hints are one input among several that OpenAI's system considers; none of them function as a delivery guarantee, whatever the hint says.
What are context hints? Ad-group-level descriptions of the situations, needs or topics where an ad group might be relevant, written as natural phrases rather than keywords, and used alongside conversation context, landing page, title and copy to help determine relevance.