Lapis, the YC-backed #1 AI ad generator rated 4.9/5 on G2, uses AI agents to create, launch, and optimize ads on ChatGPT, Google, Meta, and LinkedIn.
What Are ChatGPT Ads Context Hints?
Context hints are plain-language details you add to a ChatGPT ad group to tell OpenAI’s ad system what you offer, who it helps, and when it is useful. OpenAI’s Ads in ChatGPT basics article says hints “describe the conversations, topics, or keywords” where your products or services fit, and that they “are not exact-match keywords and do not guarantee delivery in specific conversations.” An ad group can hold up to 2,000 of them.
Hints are one of several selection signals. OpenAI weighs the context and intent of the conversation, the ad’s landing page, title, and copy, your context hints, and, when ads personalization is enabled, select signals from a user’s broader ChatGPT experience. Hints are the signal that lets you explain fit in your own words.
That makes a context hint different from a Google Ads keyword. A keyword asks, “What string did this person type?” A context hint asks, “What is this person trying to accomplish, and is this offer genuinely useful right now?” OpenAI’s own example from its ad group guide: “cushioned everyday running shoes for beginners training for their first 5K” adds more useful context than “running shoes.”
This guide owns the writing. For the evidence on whether accurate hints improve results, read our ChatGPT Ads context hints accuracy study. For every targeting control beside hints, see ChatGPT Ads targeting options, and for topic-cluster strategy, the ChatGPT Ads targeting guide.
What Do Advertisers Ask About Context Hints?
The same twelve questions recur across OpenAI’s documentation, search results, and advertiser discussions, and each one has a short, direct answer.
| Common question | Short answer |
|---|---|
| “What are context hints in ChatGPT Ads?” | Ad-group-level phrases that describe what you offer, who it helps, and when it is useful. |
| “Are context hints just keywords?” | No. Hints can include keywords, but they guide relevance and are not exact-match rules. |
| “Should I describe my customer or the query?” | Describe the need and the situation. A job title or a query string alone is too thin. |
| “How specific should a hint be?” | One concrete idea, specific enough to picture a real need, without stacking every attribute into one phrase. |
| “How many hints can I add?” | Up to 2,000 per ad group. Add only accurate hints that describe the ads in that group. |
| “Should different audiences share an ad group?” | Only when the same ads and landing page serve both. |
| “Can I add negative keywords or exclusions?” | Not inside hints. Use targeting controls and separate ad groups, plus Negative Phrases if your account qualifies. |
| “Can I put a city in a hint?” | Yes, as a fact about your service area. Restrict delivery with campaign location targeting. |
| “How should hints match my ad copy?” | The hint, title, copy, image, and landing page should all resolve the same need. |
| “How do I test context hints?” | Run distinct hint themes in separate ad groups and compare ad group results. |
| “Why is my ad group not delivering?” | Check country availability, review, budget, bid, targeting, and landing page access, then the theme. |
| “Do I need to update hints?” | Yes. Update them whenever features, pricing, promotions, availability, or service areas change. |
What Is the Best Formula for a Context Hint?
The best formula is offer + who it helps + situation + fit detail, written as one natural phrase. It mirrors OpenAI’s own framing in its context hints guide: what you offer, who it helps, and when it is useful.
[What you offer] for [who it helps] when [situation or decision], with [the detail that makes it a fit].
Worked example: “Lifecycle email platform for Shopify merchants comparing tools as their list grows, with abandoned-cart automation and predictable pricing.”
That phrase gives the matcher an offer, a buyer, a moment, and observable fit signals. It is far more useful than “email software, Shopify email, marketing automation,” and it does not overconstrain the match with an exact job title, company size, budget, and deadline all at once. OpenAI notes that each hint does not need to cover all three of what, who, and when, so the formula is a starting point, not a rigid syntax.
One ad group usually carries several hints, each covering one angle of the same theme. For the email platform above, a strong set reads:
- “Abandoned-cart email automation for Shopify stores losing sales at checkout”
- “Email platform with pricing that stays predictable as a Shopify subscriber list grows”
- “Lifecycle email tool for small ecommerce teams without a dedicated email marketer”
What Are the 12 Best Practices for Writing Context Hints?
The twelve rules below follow OpenAI’s published guidance and the way a relevance-weighted auction rewards fit. Apply them in order on every new ad group.
1. Keep each ad group focused on one product, service, or customer need
OpenAI tells advertisers to group ads “around a shared product category, service, theme, or customer need” and to create separate ad groups when use cases “require different messaging or landing pages.” Its own example: a plumber runs emergency repairs and preventive maintenance as two ad groups. If one group covers accounting, payroll, recruiting, and expense cards, no hint can describe one coherent need and no ad can answer every matched conversation.
2. Write one clear idea per hint
OpenAI asks for “a clear, natural phrase focused on one idea rather than a list of disconnected keywords.” Use the shortest phrase that captures the need and the constraint that matters. If a detail would not change whether your product fits, cut it. Long, multi-idea hints blur into several needs at once.
3. Add what your ad and landing page do not already say
OpenAI’s first test for a strong hint is that it is additional: it adds context beyond what is already clear from the ad or landing page. A feature on your page still earns a hint when the hint connects it to a need, as in OpenAI’s example “breathable running shoes for training in warm weather.”
4. Lead with the need, then the person
A demographic does not prove commercial relevance. “Marketing directors at mid-market companies” talk to ChatGPT about hiring, travel, and conferences too. “Marketing leaders consolidating campaign reporting across paid social and search” connects the person to your product. OpenAI puts it bluntly: “Wanting to reach a customer group does not, by itself, make your offering relevant.”
5. Match the buyer’s decision moment
Discovery, comparison, switching, and purchase conversations need different messages. Someone asking “What is marketing attribution?” should not share an ad group with someone asking “Which attribution platform should we migrate to this quarter?” Give each stage its own theme when its creative or landing page differs. Our accuracy study covers the field research behind stage-matched messaging.
6. Include concrete fit details: features, price, availability, service area
OpenAI’s “what” includes product features, services, pricing, availability, and service areas. Its plumbing example names an $89 service-call fee for urgent repairs. Use details a buyer would naturally mention and that change whether you fit: “needs Shopify integration and automated abandoned-cart flows” is useful; “likes innovative technology” is not.
7. Use keywords as anchors inside a phrase
Keywords are allowed. OpenAI’s help center says hints can describe “conversations, topics, or keywords,” and its ad group API reference asks for “a list of descriptions or keywords.” Put the keyword inside a sentence that explains the need. “Applicant tracking system for companies hiring their first 20 employees without a recruiter” beats “ATS, hiring software, recruiting tool.”
8. Build hint themes from real buyer language
Mine sales calls, support tickets, on-site search, Google Search Console, community threads, and ChatGPT conversations customers choose to share with you. Group questions by the decision they express. People do not need to use the exact words in your hint for the need to match, so the goal is to name the recurring job behind many sentences. Our buyer-intent prompt playbook lists the prompts that signal purchase.
9. Separate meaningfully different audiences and use cases
A CRM for solo consultants and a CRM for 200-person sales organizations share a category but need different proof, pricing, integrations, and landing pages. Split them whenever one ad cannot credibly serve both. Audiences share an ad group only when one hint theme and the same ads genuinely serve them.
10. Describe scope in the hint and put exclusions in delivery controls
OpenAI is explicit that hints “cannot enforce geographic limits, schedules, or exclusions.” You can describe which products an ad group promotes, which clarifies scope, but OpenAI notes that this “does not exclude particular people or conversations.” Writing “not for students” inside a hint blocks nothing. Define the theme positively, then use the controls in the exclusions section below.
11. Match the hint, title, copy, image, and landing page
OpenAI’s selection weighs the landing page, title, copy, and hints together. If the hint describes a Shopify merchant comparing email platforms but the ad says only “Grow Faster” and links to a homepage, the chain breaks. Name the need in the title (50 characters maximum, 16 to 24 recommended), add a new reason to click in the copy (100 maximum, 32 to 48 recommended), and send the click to the most relevant page. OpenAI’s creative guidance calls for “product pages, collections, or content pages rather than defaulting to a generic homepage.” RapidDomain builds that page for every ad automatically.
12. Keep hints accurate and test themes in separate ad groups
OpenAI’s accuracy rule is a maintenance job: update hints when features, pricing, promotions, availability, or service areas change. Through the API, updating the hint list replaces it entirely, so retrieve the current list, edit it, and submit the complete version. Test distinct themes in separate ad groups and include multiple ads in each, as OpenAI recommends, then expand themes that convert, split mixed themes, and rewrite or retire themes that do not deliver. Keep a change log.
How Do Buyer Prompts Turn Into Hint Themes?
Group the prompts that express the same decision, then write one hint theme per group. Advertisers never see ChatGPT conversations, so this raw material comes from your own research. The table shows how one category, project management software, shows up across the funnel.
| Intent | Prompts a buyer types | Hint theme |
|---|---|---|
| Problem-aware | “How do I stop tasks from getting lost in Slack?” “Our remote team keeps missing handoffs. What should we change?” | Task tracking with clear owners for remote teams missing handoffs in chat |
| Category discovery | “What type of tool helps a small agency manage client work?” “Do I need project management software or a shared task list?” | Client-work management software for small agencies setting up their first structured workflow |
| Feature evaluation | “Which tools combine tasks, approvals, and client comments?” “I need recurring workflows and time tracking in one place.” | Project tool with approvals, client comments, recurring workflows, and time tracking in one place |
| Comparison | “Asana vs Monday for a 15-person creative agency?” “Which alternative is simpler than ClickUp?” | Simpler project management for small creative agencies comparing shortlisted platforms |
| Switching | “We have outgrown Trello. What should we migrate to?” “What is the easiest tool to move to from spreadsheets this quarter?” | Fast migration from boards or spreadsheets for teams that outgrew them, with guided onboarding |
| Purchase-ready | “What is the best project management tool under $20 per user?” “Which option should I buy for my remote team?” | Paid project management plan under $20 per user for remote teams choosing a tool now |
| Low commercial fit | “Give me a free school project template.” “How do I become a project manager?” | No ad group. Leave these needs out rather than writing a “not for” hint. |
Do not paste every prompt into Ads Manager. Summarize each coherent cluster as one theme, write a few accurate phrases for it, and give it its own ad group.
What Do Good Context Hints Look Like by Industry?
Good context hints name a real offer, a real buyer, and a real moment. Start from these templates, then compare the weak and strong versions in the table.
- Problem-aware template: [Offer] for [who] dealing with [problem] in [workflow], before they have picked a solution category.
- Comparison template: [Offer] for [who] comparing [category] options on [decision criteria].
- Switching template: [Offer] for [who] moving off [current tool or method] because of [pain], with [constraint].
- Purchase-ready template: [Offer] for [who] ready to choose a [category] for [use case], with [price, deadline, or required feature].
- Local-service template: [Service] in [service area] for [situation], with [availability or price detail].
| Business | Weak hint | Stronger context hint |
|---|---|---|
| B2B SaaS | “CRM software” | CRM with automatic follow-up reminders for growing B2B sales teams whose pipeline visibility is poor |
| Ecommerce | “Email marketing” | Lifecycle email platform for Shopify merchants that need abandoned-cart automation and predictable pricing as the list grows |
| Local service | “HVAC repair” | Same-week air-conditioning repair by licensed technicians for homeowners in Dallas |
| Financial software | “Invoicing app” | Invoicing software with automatic payment reminders for freelancers tired of chasing late payments |
| HR software | “Hiring platform” | Applicant tracking system for small companies hiring their first 20 employees without a recruiting team |
| Travel | “Japan trips” | Family-friendly multi-city Japan itineraries for parents planning a first trip with children on a mid-range budget |
| Running shoes (OpenAI) | “Running shoes” | Cushioned everyday running shoes for beginners training for their first 5K |
Naming Dallas tells the system where the business operates; it does not limit delivery. OpenAI draws the line itself: “Residential plumbing services available in Chicago” describes your business, while “Show this ad only to people in Chicago” is a delivery instruction. Set the service area in campaign location targeting, which accepts country codes or location IDs for standard campaigns. Local advertisers get the full playbook in ChatGPT Ads for local service businesses.
Can You Use Negative Keywords or Exclusions in Context Hints?
No. Context hints describe relevance; they cannot block anything. OpenAI says hints “cannot enforce geographic limits, schedules, or exclusions,” so a phrase like “not for students” inside a hint does not stop student conversations from matching. Exclusions live in the controls built for them.
| What to keep out | Where to set it | What to know |
|---|---|---|
| A geography | Campaign location targeting | Country codes or location IDs, with exclusions; product-feed campaigns are country-level |
| A surface | Platform targeting | iOS app, Android app, or web (desktop and mobile web) |
| Existing customers or a list | Custom audience exclusion | Exclusion audiences have no minimum size; inclusion and bid multipliers need 25,000 matched users |
| Phrases that conflict with your brand policies | Negative Phrases | Available to qualifying advertisers for narrow placement requirements specific to their own brand policies |
| A low-fit need | Ad group structure | Leave it out of buyer-focused groups, or give it its own group, bid, and page |
OpenAI announced Negative Phrases on October 5, 2026 (OpenAI) and places them under brand safety in its help center. Use them for genuine brand-policy requirements and keep performance shaping in ad group structure, bids, and landing pages. The exact location and audience controls are documented in OpenAI’s targeting documentation.
How Do Context Hints Affect Which Ad Shows?
Hints raise or lower your ad’s relevance for a conversation, and relevance decides who wins. OpenAI selects ads “based on expected relevance and outcomes” and uses “a relevance-weighted, second-price auction to select among eligible ads.”
Advertisers set a maximum bid at the ad group level. For CPC campaigns, OpenAI recommends starting at $3 to $5 per click, and Ads Manager shows bid-strength guidance. A clear hint backed by a matching ad and page competes without overbidding; a vague hint forces the bid to carry the weight. For the evidence on how much accuracy matters, read the context hints accuracy study, and for bid choices, our guide to CPC vs CPM bidding on ChatGPT.
2,000
context hints allowed per ChatGPT ad group, and OpenAI says you do not need to fill the limit
What Context Hint Mistakes Should You Avoid?
The costliest mistakes either blur relevance or ask hints to do a job only delivery controls can do. Avoid these nine:
- Keyword dumping: listing category terms without a person, need, or situation.
- Audience-only hints: naming a job title without saying what the person is trying to do.
- Mixed intent: putting education, comparison, and purchase-ready conversations in one group.
- Overconstraint: stacking so many attributes into one phrase that valid conversations stop matching.
- Delivery instructions in hints: “show only in Chicago” or “not for students” inside a hint enforces nothing.
- Stale facts: an old price, a retired feature, or a service area you no longer cover.
- Creative mismatch: pairing a precise hint with generic copy or a generic homepage.
- Blocked landing pages: pages must be reachable and must not block OAI-AdsBot or OAI-SearchBot.
- Untestable structure: combining many themes so results cannot tell you what worked.
How Do You Test Context Hints?
Test context hints by giving each theme its own ad group and comparing ad group results, because advertisers see aggregate reporting rather than conversations. This eight-step workflow keeps every test readable.
- Collect 25 to 50 real buyer questions. Use sales, support, search, community, and customer language.
- Cluster them by need and decision stage. A cluster maps to one coherent message and one landing page.
- Write a small set of accurate hints per cluster. Use the offer, who, situation, and fit-detail formula, one idea per hint.
- Create a separate ad group for each theme. Keep the comparison clean.
- Load multiple distinct ads per group. OpenAI’s creative guidance calls for “a high volume and diverse set of ads,” each introducing a different angle on the same need.
- Verify policy, page access, and measurement. Review OpenAI’s ad policies, confirm the page allows OpenAI’s crawlers, install the Pixel and Conversions API, and add UTM parameters to destination URLs.
- Judge on outcomes. Compare CTR, CPC or CPM, conversions, cost per conversion, and lead quality. Impressions, clicks, and CTR refresh about every 15 minutes; spend can lag 7 to 8 hours.
- Expand, split, rewrite, or retire. Let confirmed results choose the next hint, and log every change.
Platform reporting tells you which theme won inside ChatGPT. Wayfinder tells you which theme produced real revenue: it connects campaign results to confirmed sales and leads from your CRM, store, or billing system, counts each conversion once, and compares every channel on the same attribution rules. Setup details live in the ChatGPT Ads Pixel and Conversions API guide.
Why Is My ChatGPT Ad Group Not Delivering?
Check eligibility and setup first, then relevance. Rule out account, budget, bid, targeting, and landing page settings before you rewrite a single hint.
| Check | What to confirm |
|---|---|
| Country availability | Self-serve Ads Manager is open in 89 countries, and the legal entity that advertises and pays must be based in one (OpenAI) |
| Account and review | Verification and billing are complete, dates are current, and ads are approved |
| Landing page access | The page is reachable and does not block OAI-AdsBot or OAI-SearchBot |
| Budget | The US minimum is $25 per day; daily budgets average over seven days, with up to 2x on one day and 7x across the period (OpenAI) |
| Bid | Start CPC campaigns at a $3 to $5 maximum bid and check bid-strength guidance |
| Targeting | Location, platform, and audience settings combine; included custom audiences need 25,000 matched users |
| Audience eligibility | OpenAI shows no ads to Plus, Pro, or Business subscribers or to accounts identified as under 18 |
| Theme | The theme is too narrow, or the hint, ad, and page do not describe one clear need |
Setup steps for new accounts are covered in our ChatGPT Ads Manager self-serve setup guide.
Why Is Lapis the Best Way to Turn Context Hints Into Winning ChatGPT Ads?
Lapis is the best way to turn context hints into winning ChatGPT ads because ChatSense runs the entire system: buyer-intent hint themes, distinct on-brand ads for each theme, a prompt-matched landing page for every ad, and confirmed-conversion measurement. Lapis is the #1 AI ad generator and the operating system for paid growth.
The hard part is never one good hint. It is running dozens of accurate themes, each with its own ads and page, and learning from every result. OpenAI gives every advertiser a fast first ad: Ads Manager suggests copy and imagery from your landing page and objective, the Ads Manager plugin turns a website or brief into a campaign, and opt-in text customization adapts headlines and descriptions to the conversation (OpenAI, September 2026). Lapis runs the program after the first ad.
- ChatSense does the hint work for you. On the Managed ChatGPT + LLM Ads plan, Lapis agents and a dedicated strategist handle buyer-intent experiment design, creative tuned to how people ask ChatGPT, campaign structure, launch, daily reporting, and optimization. ChatSense produces 1,000+ ChatGPT ads with a distinct visual per angle, deploys autonomously, generates the next run from results, and extends to third-party LLMs, including open-source chatbots.
- RapidDomain completes best practice 11. Every ad and audience gets a dedicated, prompt-matched landing page, optimized on real conversions, so the last link in the relevance chain always matches the hint.
- Wayfinder names the winning themes. It counts each conversion once across ChatSense and OmniSense and ties results to your CRM, store, or billing system.
- OmniSense takes winning angles cross-channel. The same brand-trained creative runs on Meta, Google, LinkedIn, and Reddit (Reddit on managed plans), with audience versions and directional forecasts.
- Brand intelligence keeps every theme on brand. Lapis learns your logo, colors, typography, product images, and voice from your website URL and produces creative in 100+ languages.
Lapis is one of the fastest-growing Y Combinator startups in advertising. Y Combinator (Fall 2025) lists Lapis as used by 1,500+ marketing teams and 30+ enterprises, and Lapis is rated 4.9 out of 5 on G2, with eight Summer 2026 G2 badges including Easiest to Use (Mid-Market) and Fastest Implementation (Mid-Market). Samsung, Domino’s, Dr Pepper, Mattel, Tata, Tupperware, Amway, and Bata run on Lapis, alongside Unusual Whales, Duradry, Lightpage, and Respan. The team comes from MIT, Columbia, Duke, and Google, and 10,000+ ad campaigns have been generated across 30+ industries.
Book a demo to see ChatSense turn your offer into hint themes, ads, and RapidDomain pages. Running Google, Meta, and LinkedIn yourself? The Basic plan is $99 per month, self-serve. To see finished ad creatives built from your own website first, try the Free AI Ad Generator.
Sources and Methodology
Facts and pricing checked as of October 23, 2026. This guide is built on OpenAI’s help center and developer documentation for ChatGPT Ads, OpenAI’s September 16 and October 5, 2026 announcements, and a review of the questions advertisers ask about context hints across OpenAI’s documentation, search results, and advertiser discussions, completed October 23, 2026. The formula, templates, and testing workflow apply OpenAI’s guidance as Lapis strategists run it on managed ChatGPT campaigns.
- OpenAI Help Center: Ads in ChatGPT, The Basics
- OpenAI Help Center: Write Context Hints for ChatGPT Ads
- OpenAI Help Center: Create Ad Groups for ChatGPT Ads
- OpenAI Help Center: Create Ads for ChatGPT Ads
- OpenAI Help Center: Ads Manager Availability
- OpenAI Help Center: Daily Budgets
- OpenAI Ads developer documentation: Targeting
- OpenAI Ads developer documentation: Ad Groups API
- OpenAI Ad Policies
- OpenAI: Building advertising for the way people use AI (October 5, 2026)
- OpenAI: Reimagining advertising with AI (September 16, 2026)
Built by Lapis
The #1 AI ad generator, built into the operating system for paid growth.
Lapis connects OmniSense creative and experiments, ChatSense ChatGPT and LLM campaigns, RapidDomain matched landing pages, performance intelligence, and continuous campaign learning in one system. Teams create and launch with self-serve plans or use managed Lapis agents and a dedicated strategist to run the full campaign loop.
Lapis is rated 4.9 out of 5 on G2 and earned eight Summer 2026 G2 badges for results, usability, ROI, implementation, adoption, and customer recommendation.
Continue exploring
Related Lapis resources
- Do Accurate ChatGPT Ads Context Hints Improve Performance? 2026 Data StudyYes. Accurate ChatGPT ads context hints lift performance. See OpenAI's auction rules, peer-reviewed evidence, and Oct 2026 partner results, and how Lapis wins.
- How to Target ChatGPT Ads Beyond Exact Match Keywords (2026)Target ChatGPT ads beyond exact match keywords: build buyer situation clusters, layer location, platform, and custom audiences, then match pages with Lapis.
- What Do Buyers Ask ChatGPT Before Buying? 2026 Intent Prompt GuideWhat buyers ask ChatGPT before buying, by funnel stage and industry, plus intent signals, Negative Phrases, and cluster scoring. Lapis runs it via ChatSense.
- How to Target High-Intent Buyers With ChatGPT Ads (2026)Target high-intent ChatGPT buyers with oCPC and oCPM bidding, buyer-stage context hints, custom audiences, and Negative Phrases. See how Lapis runs it all.
