CheatsheetAd SpecsReferenceCampaign Setup

ChatGPT Ads Creative Specs Cheatsheet

Chat card specs, context-hint rules, image sizes and QA bands for OpenAI Ads Manager — confirm live fields on the day you build; this sheet is the working reference.

Updated 2026-08-2411 min read

ChatGPT Ads is a labeled card under an answer, not a 9:16 hook. Specs, geos and bidding have moved through 2026; treat every UI label as directional and re-read ads.openai.com help on the day you launch. This sheet is the creative unit, the hint rules and the QA bands — not a second Meta playbook.

The unit is a chat card, not a UGC file

Each ad is a sponsored card beneath a ChatGPT response: advertiser name, favicon, title, body copy, landing page, square image. It is matched with context hints, not keywords and not a mute-scroll first frame. If your first draft could be a TikTok hook, it is the wrong object. Lead with the job the user is already in — a true constraint, a noun, who it is for — and let the model’s answer stay above you. Inventory has been aimed at logged-in adults on Free and Go tiers in rolling geos; Plus-class seats are not the default audience. If that is not your buyer, stop before you design. Keep Meta and TikTok UGC in their own lane: steal language, not files.

Context hints are conversations you will not see in a query report

Hints describe the chats where your product is a sane next step: who is talking, what they are trying to do, and what you are not for. 'I'm comparing travel backpacks that fit under a seat' is a hint. 'backpack, rucksack, carry on' is leftover Search copy. Include disqualifiers. Cover the job, the comparison moment, and one objection, then stop — a novel is more ways to match badly. Reporting has often been limited to delivery and clicks, so you will not get a transcript. Write like you will never see the query log. Two ad groups only if you have two genuinely different conversation types. Duplicate hint soup across ten groups is theatre.

Character caps move — front-load the noun and confirm live

Plan for a short title and a short body, then check the live caps in Ads Manager before you lock a grid. Some accounts have seen title fields in a ~16–24 or ~50 character band and body in a ~32–48 or ~100 band as the product matured; truncation is common either way. Front-load the noun and the constraint. Put the offer on the landing page if it will not fit. The image is a square product or in-use crop that still reads as a thumbnail, PNG or JPG, not a 9:16 talking head with burned-in captions. The landing page must continue the same SKU name. Conversation-to-PDP mismatch is more punishing here than in social because the user is mid-research.

How to use the tables below

Confirm live specs on the day you build, then QA the card against the bands: unit fields, copy length, image, hints, campaign structure, and launch gates. The last table is a pass/fail before spend. Klip Kanvas still belongs on Meta and TikTok — URL to talking-head UGC — not as the native ChatGPT file. Translate proof nouns from a winning script into the title; do not paste the voiceover. Unlock once; keep this sheet next to Ads Manager, not in a brand deck.

6 lookup tables: chat-card field specs, copy-length and truncation bands, square-image rules, context-hint writing bands, campaign/ad-group structure, and a 10-point pre-launch QA — each row is a number or pass/fail, not prose. Confirm live caps in Ads Manager on the day you build.

Locked

Unlock the full cheatsheet

Enter your email and we'll instantly unlock the complete cheatsheet — plus send you fresh ad playbooks every month.

No spam. Unsubscribe anytime.

Ready to put this into practice?

Create your first AI UGC video ad in minutes — no filming, no actors, no editing.

Try Klip Kanvas free

More in this section

Ready to make ads like these?

Paste a product link and Klip Kanvas writes the script, casts the creator and renders the ad — no filming, no actors, no editing.