A ChatGPT ad appears below a ChatGPT answer and shows the advertiser’s name, favicon, a title of 3–50 characters, body copy of up to 100 characters, an image and a landing page. The strongest examples match one specific conversation situation, make a promise the landing page can prove and point to an obvious next step. Below are 11 annotated example ads across the categories OpenAI allows, plus rewrites of weak ads and examples that would be rejected.
Every business, domain and ad on this page is fictional and illustrative. The domains end in .example on purpose. These are teaching mock-ups, not real campaigns, tested winners or evidence of results. Check every claim against your own offer before you reuse a pattern.
What a ChatGPT ad looks like, and the limits that apply
According to OpenAI, ads are shown to logged-in users on the Free and Go plans. Plus, Pro and Business plans don’t show ads, and neither do users identified or predicted to be under 18. Ads appear below the response, are labelled, and are kept separate from the answer itself. OpenAI: Ads in ChatGPT, the basics
| Element | Official requirement (OpenAI) | Practical note |
|---|---|---|
| Title | 3–50 characters | Name the product, service or situation. Spaces and punctuation count. |
| Body | Up to 100 characters | One concrete reason to click. |
| Image | JPEG, PNG or WebP, at least 640×640 px | Show the real product or service. |
| Brand icon (favicon) | At least 128×128 px | Should be recognisable at small sizes. |
| Landing page URL | Up to 2,048 characters, tracking parameters included | Send people to the page that answers the ad. |
Limits are from OpenAI’s Ads developer documentation. Product-feed ads look different: they show an image, title, star rating, price and sale price taken from the feed. See the product-feed guide.
Why the examples below are shorter than 50/100. AthillyAds drafts ad cards with stricter default limits of 24 characters for the title and 48 for the body, so cards fit the app’s workbook-compatible export and read in full on a phone. Those are our app defaults, not OpenAI rules. Every example below fits both. If you need more room, you can use OpenAI’s full limits. Here is the compact desk example at full length:
| Version | Title | Body |
|---|---|---|
| Compact (24/48) | Desks for small spaces (22) | Compare compact desks by width and depth. (41) |
| Full length (50/100) | Compact desks for small rooms and home offices (46) | Filter by width and depth, see real room photos and check delivery before you order. (84) |
How ads are matched. OpenAI chooses ads based on conversation context and intent, your landing page, title, copy and context hints. Where ads personalization is enabled, select signals from the user’s wider ChatGPT activity are also used. Personalization isn’t available in the EEA or Switzerland, so ads there are matched on context. That is why each example starts with the conversation situation. For the full mechanics, read what context hints are and how to write them.
Ecommerce and consumer goods examples
1. Compact desks: Nordvik Desks (nordvik-desks.example)
| Field | Example |
|---|---|
| Conversation situation | Someone asks how to fit a workspace into a small spare room and what desk depth to look for. |
| Context hints | choosing a desk for a small spare room; comparing desk depths for a home office; compact desk that leaves room for a chair |
| Exclusion hints | office furniture for large companies; standing desk motor repair |
| Title (22) | Desks for small spaces |
| Body (41) | Compare compact desks by width and depth. |
| Image idea | A stocked desk photographed in a real small room, in true proportion |
| Landing page | nordvik-desks.example/compact-desks, with width and depth on every product card, plus price, stock and delivery |
| Why it works | It names the constraint (space) and promises a comparison the page actually delivers. It doesn’t claim the desk fits every room. |
2. Wide-fit trail shoes: Fjällsteg (fjallsteg.example)
| Field | Example |
|---|---|
| Conversation situation | A runner with wide feet asks which trail shoes to consider before a first autumn trail race. |
| Context hints | trail running shoes for wide feet; choosing shoes for a first trail race; grip on wet rocky trails |
| Exclusion hints | marathon training plan for road running; children’s school shoes |
| Title (20) | Wide-fit trail shoes |
| Body (43) | Filter trail shoes by width, grip and drop. |
| Image idea | One shoe on a wet trail, with the sole visible |
| Landing page | fjallsteg.example/trail/wide-fit, pre-filtered to wide sizes, with a size guide and the returns policy |
| Why it works | The body repeats the three things a runner compares. The pre-filtered page means the ad’s promise is true from the first scroll. |
3. Office catering: Gröna Lådan (gronaladan.example)
| Field | Example |
|---|---|
| Conversation situation | An office coordinator asks for lunch ideas for a 20-person workshop with vegetarian and allergy needs. |
| Context hints | ordering lunch for a team workshop; vegetarian catering for 20 people; office lunch delivery with allergy labelling |
| Exclusion hints | vegetarian recipes to cook at home; wedding catering |
| Title (24) | Lunch for team workshops |
| Body (48) | Vegetarian boxes for 10–60, labelled by allergen |
| Image idea | Real lunch boxes on a meeting table, with labels legible |
| Landing page | gronaladan.example/workshop-lunch, showing minimum order, delivery area, order deadline and allergen list |
| Why it works | It answers group size and dietary needs, which are the coordinator’s two worries. Exclusion hints keep out home-cooking conversations. |
Local service examples
4. Office cleaning: Klarrum (klarrum.example)
| Field | Example |
|---|---|
| Conversation situation | An office manager compares recurring cleaning services and asks what an office-cleaning agreement should include. |
| Context hints | comparing recurring cleaning for a small office; deciding what to include in an office cleaning agreement; scheduling cleaning around office hours |
| Exclusion hints | move-out cleaning for apartments; cleaning products to buy |
| Title (24) | Office cleaning, Uppsala |
| Body (45) | Weekly or biweekly plans. Get a scoped quote. |
| Image idea | An authentic photo of the team cleaning a real office |
| Landing page | klarrum.example/office-cleaning, with service area, included tasks, how scheduling works and a short quote form |
| Why it works | It names the service and city and makes a modest promise. Geography still has to be set in the campaign’s location targeting; a city name in a hint or title doesn’t do that. |
5. Local moving: Vasa Movers (vasamovers.example)
| Field | Example |
|---|---|
| Conversation situation | Someone moving out of a two-room flat asks whether to hire movers and how to get a reliable estimate. |
| Context hints | hiring movers for a small apartment move; planning a move within the same city; moving quote with packing help |
| Exclusion hints | international relocation; buying moving boxes |
| Title (20) | Moving a small flat? |
| Body (47) | Fixed quote after a 10-minute video walkthrough |
| Image idea | The crew carrying furniture into a van, with the company name visible |
| Landing page | vasamovers.example/quote, with move date, from/to, size, what’s included and how the video call works |
| Why it works | It removes the main uncertainty (price) with a concrete process. Use it only if the quote really is fixed. |
Travel and entertainment examples
6. Family ski cabins: Åreby Cabins (areby-cabins.example)
| Field | Example |
|---|---|
| Conversation situation | A parent plans a February ski weekend for five people and asks where to stay close to the lifts. |
| Context hints | family ski weekend accommodation near lifts; cabin for five people at a ski resort; self-catering stay for a winter holiday |
| Exclusion hints | ski equipment reviews; summer hiking routes |
| Title (23) | Ski cabins for families |
| Body (45) | Check lift distance and dates for 4–6 guests. |
| Image idea | A real cabin exterior in snow, with the lift visible if it truly is |
| Landing page | areby-cabins.example/winter, showing availability by date and guest count and walking distance to the lifts |
| Why it works | It gives the planner the two facts they are checking: distance and capacity. Live availability prevents a dead-end click. |
7. Team escape rooms: Lockroom (lockroom.example)
| Field | Example |
|---|---|
| Conversation situation | Someone organising an activity for 12 colleagues asks for indoor ideas that take under two hours. |
| Context hints | team activity for 10–15 colleagues; indoor team event in the city; after-work activity under two hours |
| Exclusion hints | escape room puzzles to print at home; children’s birthday party |
| Title (22) | Escape rooms for teams |
| Body (45) | Two rooms side by side for groups of 8 to 16. |
| Image idea | A real group inside one of the rooms, faces used only with consent |
| Landing page | lockroom.example/teams, with capacity, price per person, available times and a group booking form |
| Why it works | It solves the group-size problem directly, which ordinary escape-room ads don’t mention. |
Digital products, education and B2B software examples
8. Online Spanish classes: Lingora (lingora.example)
| Field | Example |
|---|---|
| Conversation situation | Someone moving to Spain in six months asks how to learn conversational Spanish while working full time. |
| Context hints | learning Spanish before moving abroad; online conversation classes with a teacher; evening language course for working adults |
| Exclusion hints | Spanish homework help; free translation |
| Title (23) | Spanish before you move |
| Body (44) | Live small-group classes, evenings, A1 to B1 |
| Image idea | A screenshot of a real class session, with participants’ consent |
| Landing page | lingora.example/relocation, with levels, schedule, class size, price and a trial lesson |
| Why it works | It ties the course to the trigger (moving) and states the format. It avoids promising fluency in a set number of weeks. |
9. Meal-planning app: Matplan (matplan.example)
| Field | Example |
|---|---|
| Conversation situation | A parent asks for help planning a week of family dinners and a shopping list. |
| Context hints | planning weekly family dinners; making a shopping list from a meal plan; reducing food waste at home |
| Exclusion hints | restaurant recommendations; weight-loss diet plans |
| Title (21) | Weekly dinner planner |
| Body (46) | Plan 7 dinners and get a sorted shopping list. |
| Image idea | The actual app screen showing a week plan and list |
| Landing page | matplan.example, with screenshots, trial terms and price after the trial |
| Why it works | It shows what the user gets after one use. The weight-loss exclusion keeps the app away from health-related conversations it shouldn’t target. |
10. Agency handover software: Handoverly (handoverly.example)
| Field | Example |
|---|---|
| Conversation situation | An agency lead says client context gets lost when projects move from sales to delivery. |
| Context hints | replacing spreadsheet handovers for client projects; moving project context from sales to delivery; shared checklist before an agency starts client work |
| Exclusion hints | free project management templates; personal to-do apps |
| Title (24) | Clearer client handovers |
| Body (46) | See project handovers in one shared workspace. |
| Image idea | An accurate product screenshot with invented sample data |
| Landing page | handoverly.example/agencies, showing the workflow, supported integrations and a demo request |
| Why it works | It invites the reader to inspect a real workflow instead of claiming a “50% faster” result nobody can verify. |
11. Salon booking software: Salongo (salongo.example)
| Field | Example |
|---|---|
| Conversation situation | A hair-salon owner asks how to cut no-shows and let clients book online instead of by phone. |
| Context hints | online booking for a small hair salon; sending appointment reminders to clients; switching from phone bookings to online booking |
| Exclusion hints | booking a haircut near me; salon furniture |
| Title (22) | Salon booking software |
| Body (44) | Online booking and SMS reminders for salons. |
| Image idea | The booking calendar on a phone and on the salon’s screen |
| Landing page | salongo.example/salons, with features, monthly price, how to move existing bookings, and a free trial |
| Why it works | The exclusion hint filters out consumers looking for a haircut, the most likely mismatch for B2B salon software. |
Want examples like these for your own business? The free hint generator reads your landing page and returns grouped context hints with draft ads you can edit.
Before and after: rewriting weak ads
Most weak ChatGPT ads fail for the same reasons: they talk about the brand instead of the reader’s situation, they claim things nobody can check, or they’re vague.
| Weak draft | What’s wrong | Rewrite (title / body) |
|---|---|---|
| Best cleaning in town / We are the best, call now! | Unprovable superlative, no service named | Office cleaning, Uppsala / Weekly or biweekly plans. Get a scoped quote. |
| Welcome to Nordvik! / Quality furniture since 1998 | Greeting with no reason to click | Desks for small spaces / Compare compact desks by width and depth. |
| Revolutionize your workflow / The future of work is here | Hype, no concrete function | Clearer client handovers / See project handovers in one shared workspace. |
| Fluent Spanish in 30 days / Guaranteed results | Promise you can’t back up | Spanish before you move / Live small-group classes, evenings, A1 to B1 |
| AMAZING DEALS!!! / Click here | Shouting, generic, no product | Wide-fit trail shoes / Filter trail shoes by width, grip and drop. |
Pre-submission checklist
- The title is 3–50 characters (24 or fewer if you use AthillyAds’ export defaults).
- The body is 100 characters or fewer (48 or fewer for AthillyAds’ defaults).
- Every claim (price, speed, “fixed”, “free”) is true and visible on the landing page.
- The landing page opens on the promised product or service, not the homepage.
- The image shows the real product or service, is at least 640×640 px and is JPEG, PNG or WebP.
- The favicon is at least 128×128 px.
- Context hints describe situations, and exclusion hints cover the obvious mismatches.
- Location targeting is set in the campaign, not only written into a hint.
- UTM parameters are added and the URL stays under 2,048 characters. See UTM tracking for ChatGPT ads.
- The category is allowed in every country you target.
For the full list, use the pre-flight checklist.
Examples of ads that would be rejected
OpenAI’s ad policies allow mainly household and consumer goods, local services, travel and entertainment, and digital products and education. Outside the US, health, financial and legal services are prohibited. Some categories are never allowed anywhere, and ads aren’t shown next to sensitive conversation topics.
| Draft ad (fictional) | Why it fails |
|---|---|
| Dental check-ups, Malmö / Book online today | Health services are prohibited outside the US |
| Loans approved in 1 hour / Apply now | Financial services are prohibited outside the US, and the claim is unsupported |
| Divorce lawyer near you / Free first call | Legal services are prohibited outside the US |
| 100 free spins tonight / New players only | Gambling is always disallowed |
| Craft beer delivered / 24 IPAs to your door | Alcohol is always disallowed |
| Find your match today / Join free | Dating is always disallowed |
| 2-bed flat for rent / Södermalm, from May | Individual housing listings are disallowed |
| Barista wanted / Start Monday | Individual job listings are disallowed |
Political, adult, tobacco and recreational-drug ads are also disallowed. Check the current policy version before planning.
Allowed category, risky claim. An ad can be in an allowed category and still mislead. “Sweden’s cheapest movers”, “Learn Spanish in 2 weeks” and “Cuts admin by 50%” all need evidence the reader can see. If you can’t link to it, rewrite the ad as in the table above.
Frequently asked questions
What do ChatGPT ads look like?
They appear below ChatGPT’s answer, labelled as ads and separate from the response. Each ad shows the advertiser name, favicon, a title, short body copy and an image, and links to a landing page. Product-feed ads show an image, title, star rating, price and sale price instead.
How long can a ChatGPT ad title and body be?
OpenAI allows a title of 3–50 characters and body copy of up to 100 characters. Images must be at least 640×640 px and the brand icon at least 128×128 px. AthillyAds uses stricter defaults of 24 and 48 characters for its workbook-compatible export.
Are these real ChatGPT ads?
No. All the businesses on this page are fictional and use .example domains. They show how to connect a conversation situation, context hints, copy and a landing page within the official limits. They are not results from real campaigns.
Can I run ChatGPT ads for a clinic, accountant or law firm in Sweden?
Not at the moment. OpenAI’s ad policies prohibit health, financial and legal services outside the US. Advertising on ChatGPT in Sweden covers the other local rules, including the 175 SEK minimum daily budget.
How do I get example ads for my own website?
Paste your landing page into the free AthillyAds hint generator. It suggests context hints grouped by intent, explains each group and drafts ad copy. Then check every suggestion against your real offer before you use it.
How AthillyAds helps
AthillyAds turns your website into a campaign draft: a brand book, context hints grouped by intent, ad cards within the length limits, pre-flight checks and an Ads Manager export or API publishing. Start with the hint generator to see examples for your own site. The Free plan covers one full campaign, and pricing shows what Pro includes. To decide how many ad groups you need, see campaign structure.
Sources and scope
- Ads in ChatGPT: The Basics (OpenAI Help Center)Read
- OpenAI Ads developer documentation (full export)Read
- OpenAI ad policiesRead
- Create campaigns from product feeds (OpenAI Help Center)Read
Working methods and examples are editorial suggestions. Check current platform requirements and available features before implementation.