Context Hints in ChatGPT Ads: How They Work and How to Write Them

If you have run Google or Meta ads, context hints are the part of ChatGPT ads that will feel strangest, because there are no keywords. Instead of a list of terms to bid on, you write a short description of the conversation you want to show up in. Get this wrong and your ads barely deliver. Get it right and you show up in exactly the right moments. Here is how context hints actually work, and how to write good ones.

What are context hints?

Context hints are short, plain-language descriptions you provide at the ad group level, telling OpenAI what kinds of conversations, topics, and shopper situations your product is relevant to. Per the OpenAI Help Center, the system uses them, along with your ad title, ad copy, and landing page, to decide when your ad is eligible to appear. They guide matching, but they do not guarantee delivery in any specific conversation.

The format is simple: one or two sentences describing a buyer and what they are trying to do.

Why they are not keywords (the mental shift)

This is the part that trips everyone up. A keyword is a string the system matches literally. A context hint is a description the system interprets. They are different tools entirely.

That means there are no keyword lists, no negative-keyword sheets, and no exact-match-versus-broad-match toggles. You are not telling ChatGPT "show my ad when someone types these words." You are telling it "here is the kind of person, in the kind of situation, where my product is a good answer," and letting its semantic understanding do the matching. If you write hints like keywords, you will underperform.

The mistake almost everyone makes

The single most common mistake is writing a context hint that describes your product instead of the conversation you want to appear in.

Your hint should not read like a product label. It should read like the moment a shopper is in. Compare:

Weak hint (describes the product)	Strong hint (describes the conversation)
"Our merino wool socks"	"Shoppers looking for warm, non-itchy wool socks for hiking in cold weather"
"Handmade ceramic mugs"	"People wanting a thoughtful housewarming gift under $50 for someone who loves coffee"
"Organic vitamin C serum"	"Shoppers with sensitive skin comparing gentle vitamin C serums that will not cause irritation"
"Trail running shoes"	"Beginner runners asking which cushioned shoes to buy for their first half marathon"

The weak versions describe what you sell. The strong versions describe who is asking and why, which is what the system is actually matching against.

How to write a good context hint

A reliable structure is to name four things:

1. The buyer. Who is asking? ("Parents of toddlers," "first-time home cooks.")
2. Their need. What are they trying to do or solve? ("find a durable lunchbox," "replace a knife set that has gone dull.")
3. The context. The situation or constraint that matters. ("for daycare that bans plastic," "on a small budget.")
4. The intent moment. Where they are in the decision. ("comparing options," "ready to buy a gift.")

Put together: "Parents comparing durable, plastic-free lunchboxes for daycare that need to survive daily drops."

Two things sharpen it further:

• Use high-intent language. Verbs and phrases like compare, best, gift for, which to buy, recommend, replace signal shopping intent and help the system separate buyers from browsers.
• Match the journey stage. A hint for someone exploring ("overwhelmed by too many options") is different from someone deciding ("comparing two specific brands"). Write for the stage you want.

Why context hints underperform (and how to fix it)

When hints do not deliver, it is almost always one of three things:

• The hint is too generic. Broad hints lose the relevance auction to better-scoped advertisers. Be specific about the buyer and situation.
• Your landing page does not match the hint. The system reads your landing page too. If the hint promises one thing and the page shows another, relevance drops. Point each ad group at the most relevant page.
• Your budget is too low. An underfunded daily cap throttles delivery no matter how good the hint is.

Everything has to line up

The important thing to remember is that context hints do not work alone. OpenAI matches on the hint, the ad title, the ad copy, and the landing page together. If those four tell a consistent story about the same shopper and moment, you win relevance. If they contradict each other, you lose it. Write them as one coherent message, not four separate pieces.

How to know which hints actually work

Here is the catch: you can write beautiful context hints and still have no idea which ones drive sales, unless you are tracking accurately. Because ChatGPT has no keywords, you learn by testing different hints and comparing results, and that comparison is only as good as your measurement. A browser pixel misses a large share of ChatGPT conversions, so hints that are working can look like they are failing.

To actually learn which context hints convert, you need accurate, server-side tracking. That is covered in How to track OpenAI Ads conversions on Shopify, and if your numbers already look off, Why are my ChatGPT ads showing 0 conversions?.

Write better context hints, then see which ones actually convert. Install Reach on the Shopify App Store to track exactly what your ChatGPT ads return, in about 30 seconds, no code required.