ChatGPT ads creative and context hints, in English and Arabic

Who this is for

Advertisers who already have or are building a ChatGPT ads campaign and need the creative and context hints written properly, following OpenAI's own guidance rather than copied across from a Google or Meta campaign. It suits businesses needing Gulf Arabic creative alongside English. Businesses in health or finance should check eligibility first: outside the US, OpenAI's eligibility table marks most health categories as not allowed and many financial categories as restricted to select approved advertisers. It is a creative service, not account setup or bidding, and works best alongside our campaign management service or your own team's structure. It also suits a business that already has a campaign running with only one or two ads per group, since OpenAI's own system rewards the wider coverage that more distinct variations provide.

What OpenAI's system rewards

OpenAI's creative guidance is unusually plain. Build for coverage with many distinct variations. Be useful and specific rather than vague. Represent the offer accurately, since a misleading title or image risks rejection. Titles should spell out the value on their own, and copy should add new information instead of repeating the title. Context hints work the same way, describing what you offer, who it suits and when it is useful, without trying to issue a targeting instruction OpenAI does not support.

What is included

• Context hints written per ad group, covering what you offer, who it suits and when it is useful, following OpenAI's four qualities of additional, specific, relevant and accurate.
• Eight to twelve title and copy pairs per offer, each written from a different angle, price, speed, specialism, guarantee, location or proof.
• Images that are simple and match the message, built to OpenAI's published specifications: square PNG or JPG no larger than 1200 by 1200 pixels, as its bulk upload checklist states, at least 640 by 640 pixels, the minimum in its API documentation, and publicly accessible.
• A landing page review for each offer, so the click lands on the page that actually answers that specific need rather than a generic homepage.
• Arabic versions written in Gulf Arabic by an Arabic writer, not translated line by line from the English creative.
• A policy check against OpenAI's ads policies before submission, to reduce rejection cycles on sensitive or regulated categories.
• A recorded decision on OpenAI's automatic text optimisation setting, on by default in our own account, so your ad text runs as approved rather than as silently varied.

Deliverables

How we work with your existing Google and Meta agency

Creative for ChatGPT ads is not a resized version of a Google or Meta asset, since the format is a single six part unit with its own character limits, context hints and landing page rules. We write for this format specifically and do not need access to your other creative production to do it. If your Google or Meta agency already has brand guidelines, tone of voice notes or approved claims, share them with us so the ChatGPT ads creative stays consistent with what you already run elsewhere, but the write up, review and submission process happens inside this service, not shared across teams.

What you will see in the first report

The first report from this service is a creative log, not a performance report, since ads need time in market before performance is meaningful. It lists every title and copy variation written, grouped by offer and ad group, each with its character count against the limit, its submission status, Serving or Not serving, and the reason for any rejection along with what was changed to fix it. Once ads have been live for a full week, later reports add which specific variations are carrying delivery, described in general terms in early weeks before enough volume exists to draw firm conclusions from small differences between them.

The first thirty days, week by week

1. Week one. Offers and customer needs mapped out, one ad group per need agreed, and the first round of context hints drafted against OpenAI's four qualities.
2. Week two. Eight to twelve title and copy variations written per offer, images sourced or built to specification, and everything checked against OpenAI's ads policies before submission.
3. Week three. Ads submitted and any rejections addressed quickly, since a landing page or copy issue caught here costs less than one caught after spend has started.
4. Week four. Early delivery data reviewed by variation, so the weakest angles can be replaced with new ones rather than left running on guesswork.

What we need from you

Real details about the offer, honest pricing and availability, your logo and brand assets, and one person who can confirm accuracy on claims before they go live, since OpenAI says ads may be rejected if they are misleading, inconsistent, or include disallowed content. If you sell in Saudi Arabia or elsewhere in the Gulf, tell us early, since a Gulf Arabic writer works differently from an Egyptian or Levantine one and the difference is obvious to a local reader.

Why the character limits matter more than they look

A title field allows up to 50 characters, but that is not the same as the length that reliably displays everywhere. In our own account, a title sitting well inside the limit, 34 of 50 characters, already carried the note that copy may be truncated in some placements. We treat that as a reason to write toward OpenAI's shorter recommended range, 16 to 24 characters for a title and 32 to 48 for copy, as the default, and use the fuller 50 and 100 character maximum only when a specific detail earns the extra length, a delivery time, a warranty period, a fee inclusion. A title padded out to the maximum with adjectives is the one place brevity and platform mechanics point the same direction.

What you own

Every ad, image, context hint and Arabic translation we write is submitted into your Ads Manager account and belongs to your business. We do not hold creative under a separate license or require it to be pulled if the relationship ends.

What we do not do

We do not invent claims, guarantees or numbers to make an ad stand out, and we do not write context hints that try to instruct delivery, since OpenAI states plainly that hints are not targeting rules. We do not sell SEO services.

Where to go deeper

The full method for writing context hints, with worked UAE examples, is in the context hints guide. Worked ad examples across eight UAE industries, including the full image and copy specification, are in the ad examples guide, and the ad format itself is explained in the ChatGPT ads explained guide.

Quick answers

How is the fee set?

Pricing is on request. Your exact fee comes after a short call where we look at your accounts and tracking. We work with a minimum ad budget of AED 5,000 a month (SAR 5,000 in Saudi Arabia), paid by you directly to OpenAI and separate from our fee.

Who owns the account and data?

You do. The Ads Manager account, pixel, feeds and data sit under your company. We work as invited users.

Do you use AI to write the ads?

We use every tool that helps, and a person who knows your offer writes and checks the final copy. Accuracy matters more here than on most platforms, because a misleading ad can be rejected.

How many ad variations do you produce per offer?

Typically eight to twelve title and copy pairs per ad group, each written from a different angle, since OpenAI's own guidance is that a high volume of distinct variations helps its system find more relevant moments to show your ad.

Can context hints target a specific area like Dubai Marina?

No. OpenAI states hints cannot enforce geographic limits, schedules or exclusions. We describe your service area honestly inside the hint instead, since that is what the guidance actually supports.

Do you write context hints or just ad copy?

Both, and they are written together. Context hints and ad copy inform the same relevance signal, so we treat them as one piece of work per offer rather than two separate tasks.

We already have a Google or Meta creative team. Do you replace them?

No. We write only for the ChatGPT ads format, which is a single six part unit with its own character limits and context hints, not a resized version of a Google or Meta asset. Your existing creative team keeps their channels.

Will you leave OpenAI's automatic text optimisation switched on?

We check it, and it is a decision we make with you, not a default we leave untouched. In our own account this setting was on by default and could generate or translate versions of ad text without individual review, so we confirm your preference before launch and record what was chosen.

How do you handle a rejected ad?

OpenAI names ad copy or landing pages that do not align with its Ad Policies, and landing pages that block OpenAI user agents or IP addresses, as common causes, and says an ad can be edited and resubmitted, or another review requested if it was rejected in error. We read the rejection message in Ads Manager and check both before changing anything.

Do titles and copy actually get checked against the character limits?

Yes, every title and copy line is checked against the field limits before submission, not just the shorter recommended range, since a title that is too long risks truncation in some placements even under the maximum.

Related