ChatGPT Ads Context Hints: How the Native Targeting Primitive Actually Works
On Google Ads you tell the system which words to match: “running shoes,” “marathon trainers,” maybe a phrase or two in quotes. On ChatGPT Ads there is no keyword box. Instead you describe your buyer in plain English and let the platform figure out which conversations fit. That plain-English description is called a context hint, and it is the main lever you have for deciding who sees your ad.
This page explains what a context hint is, why it works by meaning instead of by exact words, and how to write good ones. The single most useful mental model: a context hint is a one-line briefing to a smart new intern. You are not handing them a list of keywords. You are telling them, in normal language, exactly who your customer is, so they can spot that customer in a conversation they have never read before.
One thing to settle up front, because it changes who this is for. ChatGPT Ads is live only in the US, UK, Australia, New Zealand, and Canada, and it is not bookable for EU or DACH advertisers as of 2026-06-13 (per OpenAI’s help center, as-of 2026-06-11). If you are in Germany, Austria, or Switzerland, treat this as readiness reading. The mechanic will be the same when access opens.
Key Takeaways
- A context hint is a plain-language buyer description, set at the ad-group level, that ChatGPT matches to conversations by meaning, not by exact-match keywords. (per OpenAI's "Create ad groups for ChatGPT," as-of 2026.)
- Matching runs on embeddings, which is a way of turning text into a meaning fingerprint so two phrases that say the same thing in different words still line up. You write the intent, the system finds the match.
- Delivery is not guaranteed. A good hint makes your ad eligible for a fitting conversation. It does not force the ad to show. Bid, relevance, and the auction still decide. (per OpenAI's help center.)
- Write 3 to 5 hints per ad group, each a specific intent in your customer's own words, not a broad category. Specific in, relevant out.
What a context hint is (in plain words)
A context hint is a short description of the person you want to reach, written the way you would explain your customer to a colleague. You attach it to an ad group, which is the box that holds a set of related ads. (per OpenAI’s “Create ad groups for ChatGPT,” as-of 2026.)
Here is the difference from search advertising, stated as simply as possible.
- Google keyword: “best CRM for sales teams.” The system looks for that string, or close variants, in what someone typed.
- ChatGPT context hint: “A sales manager at a 30-person company who is frustrated that deals slip through the cracks and wants a simpler pipeline tool.” The system reads the meaning, then watches for conversations that carry that same meaning, however the person phrased it.
Notice the hint reads like a sentence, not a term. That is deliberate. You are describing a situation and an intent, not a phrase to find.
Why it matches by meaning, not exact words (embeddings, explained)
The reason a context hint can be a full sentence is the technology underneath it: embeddings.
An embedding is a way of turning a piece of text into a set of numbers that captures its meaning. Think of it as a meaning fingerprint. Two phrases that mean nearly the same thing get nearly the same fingerprint, even if they share no words. “My subscription business is losing customers” and “how do I stop churn” look completely different to a keyword matcher. To an embedding system they sit right next to each other, because they mean the same thing.
So when you write a context hint, ChatGPT turns your description into a meaning fingerprint, turns the live conversation into one too, and checks how close they are. Close enough, and your ad becomes eligible to appear. (Targeting is described as embedding-matched plain-language buyer descriptions, not exact-match keywords, per OpenAI’s “Create ad groups for ChatGPT,” as-of 2026.)
This is why the intern analogy holds. A keyword list is a brittle instruction: match these exact words. A good briefing is a flexible one: here is the kind of person and problem to watch for. The intern who understands the brief can spot the customer phrased a hundred different ways. That is what embedding matching does for you.
Delivery is not guaranteed (what a hint does and does not do)
This is the part advertisers most often get wrong, so be clear with yourself about it.
A context hint controls eligibility, not delivery. It tells the platform, “this conversation is a fit for my ad.” It does not say, “show my ad here no matter what.” Whether your eligible ad actually appears depends on the auction. (Delivery is described as not guaranteed, per OpenAI’s “Create ad groups for ChatGPT,” as-of 2026.)
ChatGPT Ads runs a relevance-weighted second-price auction, where relevance is weighted more heavily than the raw bid (per OpenAI’s help center). In plain terms: being eligible gets you into the room. Being the most relevant, fitting offer at a sensible bid is what gets you on the screen. A perfect hint paired with an off-topic ad or a tiny bid will still lose.
How to write a good context hint
A strong hint has four traits. Miss them and you either reach the wrong people or reach almost no one.
- It names a specific person, not a category. “Marketers” is a category. “A B2B marketing lead under pressure to show pipeline from a small budget” is a person.
- It states the intent or the problem, because intent is what the embedding can match against a live conversation. A description with no problem in it is just a label.
- It uses your customer’s own language, the way they would describe the situation to a friend, not the way your brand deck describes it. “Spreadsheet chaos” beats “operational inefficiency.”
- It stays one clear idea per hint. If a hint contains “and also,” it is probably two hints. Split it.
Write 3 to 5 hints per ad group. That range is wide enough to cover the real shapes of your buyer and narrow enough that every hint is still genuinely about the same offer. Fewer than 3 and you may starve the ad group of eligible conversations. Many more than 5 and the group drifts away from a single, coherent buyer.
Good and weak hints, side by side
| Weak hint | Stronger hint (specific intent, buyer's words) |
|---|---|
| Small business owners | A solo founder doing their own bookkeeping in spreadsheets who dreads tax season and wants software that just files for them |
| People who want a CRM | A sales manager at a 20 to 50 person company whose reps forget follow-ups and who wants a pipeline they can see at a glance |
| Coffee lovers | Someone with a tiny kitchen who wants real espresso but has no counter space for a bulky machine |
| Companies needing security compliance | A startup founder who just got asked for SOC 2 by an enterprise prospect and has no idea where to start |
The pattern in every strong example: a recognizable person, a real problem stated in plain words, and a clear thing they want. That is exactly the briefing you would give a smart intern so they could spot your customer in the wild.
A worked example: one ad group, 3 to 5 hints
Say you sell a simple invoicing tool for freelancers. One ad group, one buyer, a handful of hints that each catch a different shape of the same person:
- A freelancer who is tired of chasing late payments and wants invoices that send their own reminders.
- A new self-employed designer setting up their finances for the first time and unsure how to bill clients properly.
- A consultant who works with international clients and needs invoices in different currencies without the headache.
- A side-hustler turning a hobby into a business who wants to look professional when they send a bill.
Four hints, one coherent buyer, four real situations. Each is a person and a problem, in their words. None is a keyword. That is a healthy ad group.
Where geo and tier targeting fit in
Context hints are the meaning-based lever, but they are not your only control. ChatGPT Ads also lets you target by location: country, region, or DMA (a DMA is a designated market area, the regional groupings used to buy media, like a metro TV zone). You set those through the targeting locations include field, supported by a CSV catalog and a Geo Lookup API (per OpenAI’s campaign targeting docs). As of the same documentation, there is no demonstrated demographic, device, or language targeting. So you can say “the US, plus Greater London,” but you cannot natively say “women aged 25 to 34 on iPhones.” The buyer detail that would normally live in demographics has to live in your context hints instead, which is one more reason to write them carefully.
There is also an audience floor you cannot move with a hint: ads only serve to Free and Go tier users, and not to anyone predicted or self-reported to be under 18 (per OpenAI’s help center). If your customer mostly sits on Plus, Pro, Business, or Enterprise plans, no context hint reaches them, because those tiers are ad-free. That is a fit question to settle before you write a single hint.
Context hints versus Google keywords, at a glance
If your team thinks in Google Ads, this is the translation table to keep handy.
| Google Ads keywords | ChatGPT Ads context hints | |
|---|---|---|
| What you supply | Exact words and phrases, with match types | A plain-language description of the buyer and their intent |
| How matching works | String and close-variant matching of typed queries | Meaning matching via embeddings, across the conversation |
| Set at which level | Ad group, with bids per keyword | Ad group, inherited by every ad in the group (per OpenAI docs) |
| Negative controls | Negative keywords to exclude terms | No documented negative-keyword equivalent for hints as of 2026 |
| Guarantees | A match makes you eligible; rank and bid decide delivery | A match makes you eligible; relevance-weighted auction decides delivery (not guaranteed) |
| Best unit of thought | The query | The customer |
The deeper platform contrasts (auction, pricing, reporting) live in our ChatGPT Ads vs Google Ads comparison. If you run both channels, knowing how each one targets stops you from forcing keyword habits onto a system that does not use keywords. Our Google Ads team and ChatGPT Ads team build the two in a way that fits each platform instead of copy-pasting between them.
Common mistakes with context hints
- Writing keywords disguised as hints. “Best invoicing software” is a search term, not a buyer description. Rewrite it as the person who would type that.
- Going broad to “reach more people.” A vague hint like “professionals” makes you eligible for noise, then loses the auction on relevance anyway. Broad does not mean bigger results here.
- Cramming one ad group with ten hints for three different buyers. That is three ad groups. Keep one buyer per group so the ads inside stay relevant to the people the hints invite.
- Assuming a hint forces delivery. It never does. When something is wrong, check creative fit, bid, and tier eligibility before blaming the hints.
- Ignoring measurement. If you cannot see which conversations and hints actually drove leads or sales, you are tuning blind. The native dashboard reports impressions, clicks, spend, and click-through rate only, with no conversion attribution unless you install the pixel or Conversions API (per practitioner reporting on ChatGPT ad tracking). Getting that wired up is exactly what our tracking and measurement work covers.
Frequently Asked Questions
What are context hints in ChatGPT Ads?
Context hints are plain-language descriptions of the buyer you want to reach, set at the ad-group level. Instead of choosing keywords, you describe the person and their problem in normal English, and ChatGPT matches that description to fitting conversations by meaning rather than by exact words (per OpenAI's "Create ad groups for ChatGPT," as-of 2026). Think of a hint as a one-line briefing to a smart intern about exactly who your customer is.
How are context hints different from keywords?
Keywords match the specific words someone typed, with match types and close variants. Context hints match meaning. Because they run on embeddings (a way of turning text into a meaning fingerprint), two phrases that mean the same thing line up even if they share no words. So "my subscription is losing customers" and "how do I reduce churn" can both fit one well-written hint, where a keyword would have to be listed in every phrasing.
Do context hints guarantee my ad will show?
No. A context hint makes your ad eligible for a fitting conversation, but delivery is not guaranteed (per OpenAI's "Create ad groups for ChatGPT," as-of 2026). ChatGPT Ads runs a relevance-weighted second-price auction where relevance counts more than raw bid. Eligibility gets you into the auction; relevance, fit, and a sensible bid decide whether you actually appear.
How many context hints should I use per ad group?
Aim for 3 to 5, each a specific buyer intent written in the customer's own words. That range covers the real shapes of one buyer without letting the ad group drift across several different people. If you need to describe a genuinely different buyer, create a separate ad group rather than overloading one with mismatched hints.
Is behavioral, demographic, or language targeting available?
Not as a documented native lever. As of OpenAI's campaign targeting docs, you can target by country, region, or DMA (a regional media-buying zone), but there is no demonstrated demographic, device, or language targeting. Audience detail that would normally sit in demographics has to be expressed through your context hints instead. Ads also only serve to Free and Go tier users and not to anyone under 18 (per OpenAI's help center).
Get your targeting right before you spend
Context hints are simple to write and easy to write badly. The advertisers who win on ChatGPT are the ones who stop thinking in keywords and start describing the customer: a real person, a real problem, in plain words, three to five times per ad group. Do that, keep one buyer per group, and remember that a hint earns eligibility, not a guaranteed slot.
If you would rather have the buyer descriptions, ad-group structure, and measurement built for you, our ChatGPT Ads team writes the hints and wires up the tracking, and our Google Ads team keeps your keyword channels aligned with them. Book a strategy call to talk through whether the channel fits your offer and your market.