Quick answer: context hints are descriptions you add to an ad group to explain to ChatGPT what you offer, who it is for and in which situations it is useful. They help the system assess relevance, but they are not keywords, targeting rules or a guarantee that your ad will be served. They work best when they are specific, accurate and up to date.
People coming from keyword-based campaigns tend to write context hints as a list of keywords. This is the most common mistake, and also the costliest: it deprives the system of the information it really needs. This guide explains what context hints are according to OpenAI’s documentation, how they factor into ad selection and how to write them using a repeatable method. The information has been verified as of 5 October 2026.
What Are Context Hints?
A practical definition
Context hints are additional details about your business, product or service that help OpenAI systems understand when an ad may be relevant in ChatGPT. They are set at ad group level and are used to share information that is not already clear from the ad or landing page.
In practice, they are how you explain to the system the context in which your offer makes sense: the type of customer, the need, the constraint or the situation that leads to a decision.
What they are not
The documentation is clear on four points:
- they are not exact-match keywords: people do not need to use the words in your hint for the underlying need to be considered relevant;
- they are not targeting rules or delivery instructions: they cannot impose geographical or time limits, or exclusions;
- they do not guarantee that an ad will be served in specific conversations, to specific audiences or in specific situations;
- they do not replace explicit targeting: geography, platform and audiences are configured at campaign level.
OpenAI gives a clear example: “residential plumbing services available in Chicago” describes a business, while “show this ad only to people in Chicago” is a delivery instruction, which must be handled using the product’s controls.
How They Factor into Ad Selection
One signal, not the only one
The ChatGPT Ads system selects ads based on expected relevance and predicted outcomes. It considers the conversation’s context and intent, the landing page, the ad’s headline and copy, the context hints and, when the user has enabled ad personalisation, some signals from their broader experience with ChatGPT.
Context hints therefore work alongside the other elements. An excellent hint cannot make up for a generic landing page or vague headline; a clear ad and a relevant page make the hints more effective.
There is also a consistency requirement: a hint describing a service that is not on the landing page creates an expectation that the page does not meet. Before writing a hint, check that the landing page backs up what the hint promises.
Why the Need Matters More Than the Exact Words
In conversation, the same need can be expressed in many ways. Someone looking for an accountant might talk about VAT registration, the flat-rate tax regime, deadlines or the fear of getting their tax return wrong. A hint that describes the need, such as “tax advice for freelancers in their first year of VAT registration”, covers all these ways of putting it better than a list of terms. This is why OpenAI asks advertisers to think of hints as thematic signals, not matches. The words customers actually use are still valuable: they help you understand which needs to describe, rather than giving you phrases to copy one by one.
The What, Who, When Method
OpenAI’s three questions
The documentation suggests a guiding question: what else should we know about your product, who it helps and when it is useful? This leads to three angles:
- What: relevant features, services, prices, availability or areas served;
- Who: customer needs, preferences or circumstances for which the offer is suitable;
- When: situations or decisions in which the offer may be useful.
Each hint should be a clear, natural sentence centred on a single idea. You do not need to cover all three angles in every hint.
The four qualities of a good hint
An effective hint is:
- additional: it provides information that is not already clear from the ad and landing page;
- specific: it describes a feature, a service detail, a need or a concrete situation;
- relevant: it explains a genuine connection between the offer and the use case;
- accurate: it reflects the current offer and is updated when features, prices, promotions, availability or areas served change.
A detail already on the landing page can still be useful if the hint links it to a need. OpenAI gives the example of breathable running shoes “for training in hot weather”: the feature becomes relevant because it explains when it is useful.
From Customer Knowledge to Hints: A Six-Step Process
The process below is a suggested working method for turning what a business knows about its customers into usable hints.
- Gather real questions. Start with emails, phone calls, quote requests and conversations with sales staff: note how customers describe the problem before they know about you.
- Identify your best-fit customers. OpenAI suggests using what you know about the customers who are the best fit and the situations that lead to the best results. Start there, rather than with the broadest possible audience.
- Group by need. Each need that requires a different message or landing page becomes a separate ad group.
- Write one sentence per idea. For each group, write five to fifteen hints, each covering one aspect: a detail about the offer, a type of customer or a situation.
- Review them against the four qualities. Remove anything that repeats the ad, is vague or is no longer true.
- Separate out anything that belongs in targeting. Anything that limits locations, times or people should be moved to the campaign settings.
The technical limit is generous, up to 2.000 hints per group, but a short, carefully curated list is easier to keep accurate.
Hypothetical Examples for Sectors Permitted in Italy
The following examples are hypothetical and are intended to show the writing style. Adapt them to what the business actually offers.
Local services
A removal company in Turin:
- weak: “removals Turin”;
- useful: “moving a one-bedroom flat in the city centre, including dismantling and reassembling furniture”;
- useful: “families who need to vacate their home by a specific date because their rental contract is ending”;
- move to targeting: “customers in Piedmont only”.
Tourism and experiences
A farm stay in Val d’Orcia:
- weak: “holidays in Tuscany”;
- useful: “a three-night farm stay with a swimming pool, suitable for families with young children”;
- useful: “couples looking for a peaceful weekend among wineries and villages without using a car every day”, if the property really offers transfers or activities that are accessible without a car.
Software and digital products
Booking software for beauty salons:
- weak: “management software”;
- useful: “online appointment book for beauty salons that want to reduce missed appointments with SMS reminders”;
- useful: “owners who manage two locations and want a single shared appointment book”.
Training
An online language school:
- weak: “English course”;
- useful: “English course for professionals who need to take meetings and calls with overseas clients within the next few months”;
- useful: “individual evening lessons for people who work full-time and cannot attend at fixed times”.
In every case, the hint works because it links a detail of the offer to a recognisable situation. Remember that eligible categories depend on OpenAI’s current advertising policies: check them before planning.
Common Mistakes and How to Correct Them
- Lists of keywords. “Shoes, trainers, running, jogging” adds no context. Fix: write one sentence per need.
- Delivery instructions. “Show only to users in Milan” or “exclude people who have already bought” do not work as hints. Fix: use geographical targeting and audience exclusions.
- Aspirational audiences. OpenAI points out that wanting to reach a group of customers does not, by itself, make an offer relevant. Fix: describe only genuine use cases.
- Out-of-date information. Prices, promotions and areas served change. Fix: include hint reviews in your promotions calendar.
- Groups that are too broad. Hints for very different products or needs in the same group muddy the message. Fix: create separate groups when the message or landing page changes.
- Copying the ad. Repeating the headline and copy in the hints adds nothing. Fix: use hints for what the ad cannot say.
Organisation, Maintenance and Measurement
One hint per idea, one group per need
OpenAI recommends organising groups around a shared product, service, theme or need, and creating separate groups when use cases require different messages or landing pages. The official example is a plumber with two distinct groups: emergency repairs and preventive maintenance. Same business, different needs.
When to update them and how to assess their impact
Hints should be reviewed whenever the offer changes and, in any case, at regular intervals. Anyone managing campaigns via API should know that updating the list replaces the existing one: to add a hint, you must send the complete list.
There is no report that attributes results to an individual hint. To understand whether a revision has worked, it is worth changing the hints in one group at a time, noting the date and details of the change, and comparing that group’s impressions, CTR and conversions over the following weeks. This is an indication, not proof: seasonality and competition also affect results.
A record for every group
To keep multiple groups organised, it is useful to have a shared record with five fields, completed before writing the hints:
- group need, in one sentence;
- landing page destination;
- list of hints, one line per idea;
- related targeting settings , kept separate from the hints;
- date of last review and reason for the change.
The record also helps anyone taking over management: it makes clear why a group exists and what it describes, without having to reconstruct decisions from the platform’s history.
To set up groups, targeting and ads in sequence, follow the guide to creating a ChatGPT Ads campaign.
Frequently Asked Questions
How many context hints do you need per group?
OpenAI does not recommend a specific number. The technical limit is 2.000 per group, but a short, specific and up-to-date list is more useful than a long, repetitive one.
Can I include prices and promotions in hints?
Yes, if they are accurate. OpenAI lists prices, availability and areas served as useful information, and asks advertisers to update hints when they change.
Do hints improve a brand’s position in ChatGPT responses?
No. Hints relate only to ad selection; ads are separate from responses and do not affect them. Organic visibility in AI responses is a separate area of work, such as the service AI Search Visibility & ChatGPT Ads.
Can I use the same hints in multiple groups?
This is not recommended. OpenAI asks advertisers to keep hints focused on the group’s theme and the ads it contains. If two groups share most of their hints, they probably describe the same need and could be combined; if they describe different needs, the hints should reflect that difference.
Can hints exclude a type of customer?
No. They can clarify which products a group promotes, but they cannot exclude people or conversations. Exclusions require the targeting controls available in the account.
If you want to turn your knowledge of your customers into groups and context hints ready to test, talk to Digital Creation about your strategy.
Sources and Updates
Updated: 5 October 2026
Did you enjoy this article?
