Your creative sits inside somebody's answer, not on a results page. We measured every captured creative to show the shape real advertisers are actually shipping, rather than guessing at limits.
The card is short and reads like a recommendation. Copy written for a banner reads as an interruption in a conversation, and interruptions do not get clicked.
Measured across Tens of thousands captured creatives. These describe what live advertisers are doing — they are not maximums or recommendations from OpenAI.
| Field | Median | Middle 50% | Observed range |
|---|---|---|---|
| Title | 28 | 22–36 | 4–236 |
| Description | 63 | 46–86 | 7–5000 |
If your title sits far outside the middle band, that is worth a second look — not because a rule forbids it, but because you are writing to a different shape than the advertisers already in the surface.
Read how established advertisers phrase the offer, then compare it to yours.
Order food from local spots open late, delivered fast.
Shop Our Custom-Made Blinds & Shades Today.
Get quotes in minutes from affordable exterior painters.
Get 50% Off during our Summer Sale. In-Person & Online Services Available Now!
Get Free Installs With Door Purchase. Florida's Top-Rated Cat-5 Doors & Windows, Made For Florida
Voted PNW's #1 Replacement Window & Door Company. Get A Free Estimate From A Name You Can Trust!
Observed, not official. Every number on this page is measured from creatives we captured inside ChatGPT. OpenAI publishes its own creative guidance, and that remains the authority on what is permitted. We show what advertisers are actually doing, which is a different and complementary question. We do not have click-through or conversion data for any of these cards.
Sizing, images, and what separates a card that fits from one that does not.
It reads like a recommendation inside an answer rather than like a display banner. The card is short, names a specific situation, and makes the next step obvious. Copy imported from Google or Meta usually underperforms because it is written to interrupt rather than to fit a conversation.
Across tens of thousands captured creatives the median title is about 28 characters, with the middle half falling between roughly 22 and 36. These are observed values from live ads, not official limits published by OpenAI.
The median captured description runs about 63 characters, with the middle half between roughly 46 and 86. Again, observed rather than official, and shorter cards are common.
In our captured sample 100% of creatives carry an image, so a large share run without one. Treat the image as optional and judge it on whether it clarifies the offer, not on whether competitors use one.
Yes. Every creative we capture is browsable by category and by advertiser, including the prompts each ad was observed against. That is the most direct way to see how established advertisers in your space actually write their cards.
Questions about the data, a brand you expected to see, partnerships, or access to the intelligence layer. We read every message.