How to Write Context Hints for ChatGPT Ads (Templates, Examples & Rewrites)
Tarun Kapoor
Updated October 13, 2026
18 min read
A good context hint describes one buyer in one situation, in the buyer's own words. This workbook shows you how to write context hints for ChatGPT Ads step by step. It covers a seven-step process, eight fill-in templates, twelve weak hints rewritten, three complete ad groups with matching titles, descriptions and landing pages, an AI critique prompt and a scorecard to run before anything goes into OpenAI Ads Manager.
The short answer
To write a context hint for ChatGPT Ads, describe one buyer in one situation in one or two plain sentences: who they are, what has just happened, and the detail that makes your offer the right fit. Use the buyer's own words, not keywords. Then make the ad title, description and landing page speak to the same moment.
This page is the hands-on workbook. It gives you the steps, templates, rewrites and worked ad groups. For the concepts behind them, such as what context hints are, how the auction scores them and what they can't target, read the definitive guide to context hints for ChatGPT Ads first.
How to write a context hint in 7 steps
- Pick one buyer moment per ad group. A moment is a person plus a trigger: “an operations manager whose spreadsheet broke at quarter close.” If you can't picture the conversation that person is having with ChatGPT, the moment is too vague. Every hint in the ad group should describe the same moment. A second moment means a second ad group.
- Collect 10–20 real buyer sentences. Take them from sales-call openings, support tickets and reviews. You're after the sentence a buyer uses to describe the problem before they know your product exists. The full list of sources is in where to find hint language.
-
Draft with the formula. Write
[who] + [situation or trigger] + [what they need]. That covers OpenAI's own what, who, when guidance in a single sentence. - Add one fit detail. This is a number, constraint, competitor or deadline that makes your offer the answer and the generic one the wrong answer: “under 20 seats”, “HIPAA-eligible”, “moving off Mailchimp”, “before Black Friday”.
- Cut to one idea in one or two sentences. Delete adjectives, your product name and anything that describes your company rather than the buyer. If the hint needs “and” to join two people or two problems, split it.
- Write four to six variants that each change one thing, such as the verb, the trigger, the constraint or the vocabulary. The variant matrix below shows how. Most ad groups end up with five to fifteen hints.
- Run the pre-submit checklist. Score every hint against the scorecard before it goes into Ads Manager. Anything scoring under 12 gets rewritten.
Fill-in-the-blank context hint templates
Copy a template, fill the brackets and then cut it down. Each one matches a stage of the buying conversation. The stage structure follows how OpenAI Ads reads conversational intent. Each stage normally gets its own ad group, because each needs a different ad and landing page.
Problem-aware
Use when: the buyer knows the symptom but not the category yet.
[Who] who are dealing with [symptom, in their words] and don't yet know what kind of tool or service fixes it.Comparison
Use when: the buyer is weighing named options or a shortlist.
[Who] comparing [option A] and [option B] (or [category] options) for [use case], deciding on [the criterion you win on].Switching
Use when: the buyer is leaving an incumbent for a specific reason.
[Who] moving off [incumbent] because [specific failure], looking for [category] that [must-have].Ready to buy
Use when: the buyer has decided on the category and is choosing a vendor.
[Who] who have decided to buy [category] and need [feature, integration, price or deadline] to sign.Trigger event
Use when: something has just changed: a hire, funding, a regulation or the season.
[Who] who just [trigger event] and now need [category] to [job].Outcome or goal
Use when: the buyer describes a result, not a tool.
[Who] trying to [measurable outcome] by [deadline or constraint] without [what they want to avoid].Local service
Use when: the job is urgent or physical. Leave place names out, because location targeting handles geography.
[Who] with [urgent problem] [timeframe] looking for [service] that [fit detail: flat fee, same day, licensed].E-commerce use case
Use when: you sell products, including through a product feed.
[Who] shopping for [product type] for [occasion or use] who care most about [deciding attribute].Want a first draft without the blank page? The free context hint generator turns a product description into situation-style hints you can then edit with the steps above.
Before-and-after rewrite clinic: 12 weak hints fixed
These are the failure modes we see most often when auditing ad groups. They're grouped by what went wrong, not by industry. For single-line examples by vertical, see the 50 examples in the definitive guide.
| Failure mode | Weak hint | Rewrite | What changed |
|---|---|---|---|
| Keyword dump | crm, sales crm, best crm, crm software, crm tool | Founders of 5–15 person service businesses tracking leads in spreadsheets who want a CRM they can set up in a day | A person and a situation replaced five synonyms |
| Category only | Project management software | Consultants juggling proposals for several clients who keep missing deadlines and want one view of every deliverable | Added who and when |
| Audience with no need | Marketing managers at SaaS companies | Marketing managers at SaaS companies who can't tell which channel drives pipeline since their attribution tool stopped tracking | Added the trigger |
| Demographic | Women aged 25–34 who like fitness | New runners building up to their first 10K who keep getting knee pain on longer runs | Replaced who they are with what they're doing |
| Location instruction | Show this to people in Austin, Texas | Homeowners whose AC failed in a heatwave and need a same-day repair with upfront pricing | Geography moved to location targeting |
| “Not for” exclusion | Accounting software, not for students or freelancers | Controllers at multi-entity companies consolidating month-end across three or more subsidiaries | Exclusions don't block anything. Describe the right buyer instead |
| Overloaded | Founders and CMOs and developers evaluating analytics, attribution and data warehouses for growth and reporting | Heads of growth at Shopify brands evaluating attribution tools because Meta and GA4 report different revenue | One person, one problem. The others become separate ad groups |
| Spec-sheet jargon | AI-powered revenue intelligence platform with multi-touch attribution | Sales leaders who can't tell which deals are real until the last week of the quarter | The buyer's words, not the product's |
| Superlatives | The best, most trusted, award-winning payroll solution | Owners of 1–20 employee businesses running payroll by hand who are worried about late tax filings | Claims about you removed. Hints describe the buyer |
| Literal user query | what is the best email marketing tool | Shopify founders whose email list just passed 10,000 subscribers and whose current tool's price doubled | A question someone might type became the situation behind it |
| Too narrow | Left-handed dentists in their first year of private practice in a two-chair clinic using Dentrix version 11 | Dentists in their first years of private practice replacing legacy desktop practice software | Three constraints cut to one that matters |
| Stale fact | Teams looking for a tool with a free plan for up to 10 users | Teams of under 10 who want to start free and upgrade only when they add seats | The fact now matches the current pricing page. Re-check whenever the offer changes |
Sentence, question or keyword list? Choosing the hint format
Advice on format conflicts. Some tutorials say to separate items with commas, others to write full sentences, and a few to phrase hints as the user's question. Here is the same hint in all three formats:
| Format | Example | Our view |
|---|---|---|
| Keyword list | switch from mailchimp, mailchimp alternative, cheaper email platform | Fast to write, but it carries no who and no why |
| User question | what's a cheaper alternative to mailchimp for an agency? | Captures one phrasing of the moment and misses the rest |
| Situation sentence | Agencies moving off Mailchimp because pricing jumped with list size, looking for flat pricing for client sub-accounts | Our default. It names the buyer, the trigger and the fit |
In one published test, Search Engine Land's ad group built from a short keyword list delivered more than one built from full questions. A single test proves little, and that's the honest position on format: write situation sentences by default, and if you want to know for your account, run format as a theme. Put each format in its own ad group with the same bid, ads and landing page. If you use bulk upload, check how the template separates hints before putting commas inside one.
The specificity dial: too broad, right-sized, too narrow
| Setting | Hint | What happens |
|---|---|---|
| Too broad | People looking for accounting software | Matches many conversations, wins few. Every competitor fits it as well as you do |
| Right-sized | Owners of independent retail shops reconciling POS sales with their books by hand at month end | A real, recurring conversation that your offer fits better than a generic one |
| Too narrow | Owners of two-location vintage furniture shops in their second year using Square and Xero who reconcile on Sundays | Describes a conversation that almost never happens. Expect little or no delivery |
Watch out for literal-word traps too. A hint built around a word with several meanings, such as “advertising,” “pipeline” or “conversion,” can pull in conversations about the other meaning. Pin the word down with the situation around it. If an ad group won't deliver at all, work through the delivery checklist before blaming the hints.
Writing variants that add coverage, not noise
Variants exist because buyers describe the same moment in different words. A good variant changes one element and keeps the moment fixed. Here are five variants for one switching ad group:
| Change | Variant |
|---|---|
| Base hint | Agencies moving off Mailchimp because pricing jumped with their list size |
| Verb | Agencies leaving Mailchimp after their bill grew faster than their client list |
| Trigger | Agencies whose Mailchimp renewal just came in far higher than budgeted |
| Constraint | Agencies that need flat-priced sub-accounts for each client after outgrowing Mailchimp |
| Vocabulary | Small marketing shops priced out of Mailchimp who send newsletters for several clients |
If a variant changes the buyer or the moment, it isn't a variant. It belongs in a new ad group. For the rules on how many hints to use, see how many context hints.
Three complete ad-group builds
Hints don't work alone. OpenAI scores the hints, the ad title, the ad copy and the landing page together. These builds show all four lined up on one moment. The title and description are kept short, because the card cuts long headlines off.
B2B SaaS: the switching group
| Moment | Marketing agencies leaving Mailchimp over price |
|---|---|
| Hints (5) |
|
| Title | Flat pricing for agencies |
| Description | Unlimited client sub-accounts. Import from Mailchimp in a day. |
| Landing page | H1 “The Mailchimp alternative built for agencies”, a price comparison at 3 list sizes, a migration checklist |
Local service: the emergency group
| Moment | Homeowners with no heat or hot water tonight |
|---|---|
| Hints (5) |
|
| Title | Out tonight, flat fee |
| Description | Licensed engineers. Price agreed before we start. |
| Landing page | Click-to-call above the fold, the flat call-out fee stated, same-evening hours, licence number |
E-commerce: the comparison group
| Moment | New runners choosing their first proper running shoe |
|---|---|
| Hints (5) |
|
| Title | Cushioned shoes for new runners |
| Description | Wide fits, 60-day run test, free returns. |
| Landing page | A collection filtered to cushioned beginner shoes, a fit guide and the return policy near the top |
Running a catalog? Product feed ads still benefit from hints written around shopping moments. See the ChatGPT product feed guide and bulk product feed campaigns. Small local businesses should also read ChatGPT ads for small business.
Message-match worksheet: hint → title → description → landing page
Use this for every ad group. The title and description fields accept up to 50 and 100 characters, but the card truncates the headline at around 35, and OpenAI's own guidance points to roughly 16 for the headline and 32 for the description. Lead with the hint's key phrase. The specs are in anatomy of a ChatGPT ad and creative best practices, and you can check truncation in the free ChatGPT ad preview tool.
| Field | Question to answer | Your version |
|---|---|---|
| Moment | Who, and what just happened? | |
| Key phrase | Which two to four words from the hints carry the moment? | |
| Title (~16–35 chars) | Does it promise the outcome this buyer wants? | |
| Description (~32–100 chars) | Does it state the fit detail and remove one objection? | |
| Landing-page H1 | Would this buyer recognise their situation within three seconds? | |
| Proof | Is every claim in the hint visible on the page (price, feature, speed)? |
Special cases: competitors, numbers, languages and regulated categories
- Competitor and tool names. “Moving off HubSpot” describes a real conversation, and switching hints are among the most specific you can write. Keep the name descriptive, never disparaging, and check OpenAI's ad policies before you scale.
- Prices, team sizes and deadlines. Numbers make good fit details (“under 20 seats”, “before Black Friday”). Use a range rather than an exact figure, and make sure it's true on the landing page today.
- Non-English markets. Write each market's hints from local buyer phrasing in the local language. Don't translate your English set. This matters more as ads open in markets such as Japan, Brazil and Southeast Asia.
- Regulated and sensitive categories. OpenAI doesn't show ads in conversations about health, mental health or politics. Hints for health-adjacent businesses should describe the operator's situation, not the patient's. The full list of limits is in what context hints can't do.
Draft with AI, edit like an operator
ChatGPT is a good first drafter and a better critic. Draft with the prompt in the definitive guide or with the hint generator. Then paste your drafts into this critique prompt:
You are reviewing context hints for a ChatGPT Ads ad group.
The buyer moment for this ad group is: [moment].
For each hint below, score 0-2 on: one buyer, one situation, a fit detail,
buyer's own words, no demographics or locations, no claims about us,
1-2 sentences, provable on our landing page [URL].
Flag any hint that describes a different buyer or moment than the others.
Then rewrite the two lowest-scoring hints. Hints:
[paste hints]Treat the output as a suggestion. The model doesn't know your buyers. Your sales calls do.
Pre-submit checklist and hint scorecard
Score each hint 0 (no), 1 (partly) or 2 (yes). Anything under 12 out of 16 gets rewritten before it goes into Ads Manager.
| # | Check | 0–2 |
|---|---|---|
| 1 | Describes one buyer in one situation | |
| 2 | Names who and what has just happened (the trigger) | |
| 3 | Contains a fit detail your offer wins on | |
| 4 | Uses words a buyer would say, not your product taxonomy | |
| 5 | No demographics, place names, “not for” clauses or superlatives | |
| 6 | One or two sentences, with no “and” joining two different buyers | |
| 7 | Every claim is visible on the landing page, and the facts are current | |
| 8 | The ad title and description speak to the same moment |
Before launching, also check the setup side: the conversion pixel and Conversions API (conversion tracking) and an ad-group parameter on the landing URL (URL parameters).
When to rewrite a live context hint
Ads Manager reports results per ad group, not per hint. So you rewrite sets, and you judge them on qualified conversions, not click-through rate. Rewrite when:
- The ad group barely delivers. Loosen the most specific constraint, or add variants in different vocabulary.
- It delivers but doesn't convert. The hints are finding the wrong moment. Tighten the trigger, or split the group into two moments.
- The offer changed. Update any hint that states a price, plan or feature on the same day you change the landing page.
- A theme wins. Write more variants of the winning moment, and retire the losing ad group rather than tweaking it forever.
Batch your edits and log the date, because each change resets the comparison. The test design itself is covered in how to test context hints.
Done for you: Context Hint Strategy
Writing hints is easy; building targeting you can actually read when the platform reports nothing per hint is not. Our context hint strategy engagement harvests your buyers’ real language and designs the ad-group architecture around it — from $3,999.
Frequently asked questions
How do I write a context hint for ChatGPT Ads?
Describe one buyer in one situation in one or two plain sentences: who they are, what has just happened and the detail that makes your offer fit. Use their words rather than keywords, write four to six variants and score them against the checklist. The full process is in 7 steps.
Is there a template for ChatGPT Ads context hints?
Yes. The core formula is [who] + [situation or trigger] + [what they need], plus one fit detail. There are eight fill-in templates by buying stage in fill-in-the-blank templates, and the free context hint generator drafts hints for you.
Should a context hint be a sentence, a question or a list of keywords?
We recommend a situation sentence because it carries who, when and why. Keyword lists and user questions each capture only one phrasing. Results vary by account, so if you want to know for yours, test format in separate ad groups. See choosing the hint format.
How long should a single context hint be?
One idea in one or two sentences. If it needs "and" to join two buyers or two problems, split it, and usually into a second ad group.
How different should my hint variations be?
Each variant should change one element (the verb, trigger, constraint or vocabulary) and keep the same buyer and moment. If the buyer changes, it belongs in a new ad group. See the variant matrix.
Can I put exclusions like "not for students" in a context hint?
It won't block anything. Hints are relevance signals, not filters. Describe the right buyer precisely instead, and use campaign-level custom audience exclusions where you need a hard exclusion.
Should I include my city or service area in a context hint?
No. Use campaign location targeting for geography. Put the urgency and the job in the hint ("no hot water tonight") and leave place names out.
Can I mention competitor brands in a context hint?
Yes, descriptively. "Moving off Mailchimp" describes a real switching conversation. Keep it factual, never disparaging, and check OpenAI's ad policies before scaling.
Should context hints repeat what my landing page says?
They should agree with it. Every claim in a hint, such as a price, feature or speed, should be visible on the landing page, because OpenAI scores the hints and the page together. Use the message-match worksheet.
How do I make my ad title and description match my context hint?
Lead the title with the hint's key phrase and use the description for the fit detail and one objection. The fields accept up to 50 and 100 characters, but the card truncates headlines at around 35, so keep them short. See creative best practices.
Can ChatGPT write my context hints for me?
It can draft and critique them. Use the critique prompt to score drafts, then check every hint against real buyer language from sales calls and support tickets.
How often should I update my context hints?
Whenever the offer changes, and when an ad group under-delivers or delivers without converting. Batch edits, log the date and judge the new set on qualified conversions. See when to rewrite.
What are examples of bad context hints?
Keyword dumps, bare categories, demographics, location instructions, "not for" clauses, superlatives and hints overloaded with several audiences. Twelve are rewritten in the rewrite clinic.
Want Tarun to rewrite your hints with you?
30 minutes, free, on a screen share. We'll run your ad groups through the scorecard and rewrite the weakest hints with you on the call.
Book your free hint reviewTarun Kapoor
Twelve years of media buying across GroupM, WPP, Ogilvy & Mather, and Neil Patel Digital — where he was directly responsible for $3.5M+ added to the bottom line of a $6.5M ARR agency.