ChatGPT ads best practices for targeting, creative specs and ad policy

ChatGPT ads best practices start with context hints that match real conversations. Then you check creative specs, ad policy, and tracking before you spend.

AI search means answers from assistants like ChatGPT. Paid placements now sit beside those organic answers. So plan paid and organic work on the same prompts.

Platforms like AirOps show which ads, yours and competitors', appear on the prompts you track. This guide covers OpenAI's rules plus field practice. Every OpenAI Help Center fact below is as read October 14, 2026.

What are the best practices for advertising on ChatGPT?

Match your ad to the conversation, then clear specs, policy, and tracking before launch. These eight ChatGPT ads best practices follow the order you build a campaign.

1. Write context hints as natural phrases about one situation.
2. Set location, platforms, and audiences in campaign settings.
3. Keep titles, copy, and images inside OpenAI's limits.
4. Start each new ad group with several ad variants.
5. Send clicks to a matching page that OAI-AdsBot can crawl.
6. Check your category against the live Ad Policies.
7. Track conversions with a shared event ID and a holdout.
8. Check who advertises on prompts that name your brand.

The last practice sits outside your ad account. An AirOps measurement study covered 2.8 million ChatGPT answers across 131 fintech brands in 2026. On prompts that named a brand, that brand received 3.1% of observed ad placements.

That finding covers fintech only. It still makes your own brand prompts worth a check before launch. Each section below covers one of these ChatGPT ads best practices in detail.

Are ChatGPT ads targeted?

Yes. ChatGPT ads targeting is contextual and starts with the conversation on screen. OpenAI's ads FAQ lists the signals it uses to pick an ad:

• The context and intent of the current conversation
• The ad's landing page, title, and copy
• Advertiser-provided context hints
• Your targeting selections

Personalization goes further only when the person turns it on. OpenAI's FAQ says memories and recent chats inform ads only with ads personalization and memory both on. OpenAI also says advertisers never receive chats, chat history, memories, names, or email addresses. You get aggregated reporting from OpenAI instead.

Eligibility narrows the audience further. OpenAI's FAQ sets these limits:

• Ads appear on the Free and Go plans.
• Plus, Pro, Business, Enterprise, and Edu accounts see no ads.
• Accounts OpenAI identifies as under 18 see no ads, based on age data and age prediction where available.
• Temporary Chats show no ads.
• Ads stay away from personal health, mental health, and politics.
• Ads in logged-out sessions must suit all audiences.

The plan rule shapes your reach. Buyers on paid ChatGPT plans never see your ad. You reach them only through organic answers, so keep that work funded.

Estimate the split before you set a budget. Your analytics, sales notes, or customer surveys can show which plans your buyers use. A business-to-business (B2B) audience on paid plans needs more organic investment.

How to write ChatGPT ads context hints

ChatGPT ads context hints describe the conversations where your ad fits. You add them at the ad-group level. OpenAI's context hint guide asks for "a clear, natural phrase focused on one idea" instead of a keyword list.

That guidance treats hints as relevance signals. The same guide says hints cannot control geography, schedules, or exclusions, and they guarantee no delivery. OpenAI documents no length limit for hints.

Length still needs a working rule. A useful hint packs a product and a buying moment into one phrase.

Alex Hallenbeck, Lead Product Manager, Enterprise at AirOps, builds hints by "defining, really, like, 5 to 7, words within those context [hints], where you're defining the product, the situation, and the persona around who that's for".
Treat that range as one practitioner's working method. OpenAI sets no hint length, so use it as a starting habit. Then check it against your own delivery data.

Keyword-style hints still deserve a test. In a Search Engine Land article from August 10, 2026, John Horn of StubGroup reported one agency test. He compared an ad group with keyword-list hints against an ad group with full-question hints. In his test, the keyword-list ad group showed more often than the question-list group. It also had a slightly higher click-through rate (CTR) and lower average cost per click (CPC).

That is one test with no published sample. So test both styles in separate ad groups and keep the stronger one. The examples below are illustrative, and each natural phrase covers a product, a moment, and a buyer.

Use campaign controls for what context hints cannot do

Campaign settings handle the hard limits in ChatGPT ads targeting. Hints shape relevance, and settings decide where your ads can run. OpenAI's campaign setup guide lists the controls below.

Two rules catch new advertisers. The guide says new self-serve accounts can stay limited to their home country until identity verification and a spend threshold. OpenAI's ads FAQ says personalized ads are not initially available in the European Economic Area (EEA) or Switzerland, and the campaign guide tells you to avoid custom audiences there.

The country rule trips up US launches. OpenAI's guide says excluding other countries alone does not restrict delivery. Add the United States as an explicit location instead.

What are the creative specs for ChatGPT ads?

OpenAI's bulk upload checklist recommends 16–24 character titles, 32–48 character copy, and square images. Those ChatGPT ad specs keep recommended ranges separate from hard maximums.

The checklist states these limits for bulk uploads. Apply them to every ad anyway, because one standard keeps review simple. The checklist also requires images at a publicly reachable URL.

Third-party guides add their own numbers. The same Search Engine Land article lists a 256 × 256 pixel minimum image size. OpenAI's checklist names no minimum, so follow OpenAI's ChatGPT ad specs first.

Spec errors multiply as your account grows. Hallenbeck also wants a governed system that checks character length, counts, and structure for every entity you create. Run that check before upload.

ChatGPT ads best practices for creative and landing pages

Creative is the part of the ad you fully control. Give each ad group enough options to learn from. The same AirOps product lead recommends "anywhere between, like, 3 to 5 ad variants as a starting point" for each new ad group.

Change one element per variant so each result is readable. Swap the title in one variant and the image in another. These ChatGPT ads best practices apply to every new ad group you open.

Write each title around the need in its hint. A hint about first invoices pairs with a title about first invoices. Keep the copy on that same single idea.

Lead with a product or person image rather than your logo. The image should show what the buyer gets or who uses it.

The landing page is part of the ad you ship. A page the crawler cannot load stalls your launch. OpenAI's crawler guidance is direct: "You must allow OAI-AdsBot."

Treat that bot as a launch blocker. Then work through the failure points OpenAI's crawler guidance flags:

• Allow OAI-AdsBot in robots.txt, and allow OAI-SearchBot as OpenAI recommends.
• Check that web application firewall (WAF) rules return no 403 errors to the bot.
• Check content delivery network (CDN) settings for the same blocks.
• Remove CAPTCHAs and login walls from ad landing pages.
• Watch rate limits that return 429 errors to the crawler.

OpenAI's guidance notes that Cloudflare verifies and allowlists OAI-AdsBot. Then match each ad group to one page. Send a context hint about first invoices to an invoicing page, not your homepage.

What is the ChatGPT advertising policy?

The ChatGPT advertising policy is OpenAI's Ad Policies page, version 1.6, updated September 10, 2026. The page sets which advertisers, creatives, and landing pages OpenAI accepts.

OpenAI's ads FAQ names five principles for ads. They are mission alignment, answer independence, conversation privacy, choice and control, and long-term value.

Version 1.6 splits categories into disallowed and restricted. The table shows the main categories in OpenAI's policy.

The policy's approved financial list includes credit cards, mortgages, insurance, and personal loans. The policy excludes credit repair, debt settlement, and alternative investments such as bullion. Approval can require proof of licensure.

Read the whole page before you plan a regulated campaign. The page also contradicts itself. Its introduction still says legal and financial services are "disallowed at launch."

The category sections allow approved US advertisers in both. Check the live policy before you brief creative. Outside the US, the policy generally prohibits financial and health ads and prohibits legal ads.

Review follows a fixed path. OpenAI's policy describes it this way:

• Automated systems handle the bulk of reviews, with human oversight and calibration.
• Review covers the advertiser, the creative with its landing page, and the placement.
• Approved creatives cannot link to pages with disallowed content.
• Ads cannot imitate the ChatGPT interface.

How to measure and test ChatGPT ads

Set up tracking before your first impression. OpenAI's conversion measurement guide supports the OpenAI Pixel and the Conversions API. The guide says to send the same event ID through both, so OpenAI can deduplicate.

Next, plan a control group that sees no ads. The recap of a joint AirOps webinar with Ethan Smith, CEO of Graphite, calls for tests with clear control groups. If the control moves too, the change did not win. Hold back a matched region or audience from day one.

Expect gaps between systems. The same OpenAI guide says platform totals can include modeled conversions and differ from your analytics. Pick one source of truth for each decision before launch.

Pick an objective that matches the result you report. OpenAI's campaign guide offers Views, priced by cost per thousand impressions (CPM), plus Clicks and Conversions. The guide bills Conversions campaigns per click or per 1,000 impressions, not per conversion.

OpenAI's daily budget guide sets a standard US minimum of $25 a day. The guide treats that budget as an average across seven days. Per the same guide, spend on a single day can reach twice your daily budget.

Last-click reports give a new channel little credit. A holdout shows what changed with ads running. Kevin Indig, author of Growth Memo, put it this way in a recent session: "last-click attribution, is super 2006, because it measures what, or who got credit, where incrementality actually measures what works."

A holdout gives you that incrementality read. Compare conversions in exposed and held-back groups over the same dates. The gap between them is your lift estimate.

Self-reported attribution adds a second view. Ask new buyers where they first heard about you. Log the answer in your customer relationship management (CRM) system next to the click data.

In the clip below, Kevin Indig, author of Growth Memo, discusses AI search winners, losers, and attribution.

Change one variable at a time, so each result has one cause. Keep the holdout untouched while you test hints and variants. Give each change enough spend to read before you judge it.

Read paid results next to organic ones. Track the same prompts for ad placements and organic mentions. A mention is an AI answer naming your brand, with or without a link.

Pair spend data with AI visibility metrics for those prompts, and review both together. Paid share and organic share on one prompt show where your next dollar belongs.

Your first 30 days with ChatGPT ads best practices

Use this plan as a launch checklist. It puts the ChatGPT ads best practices above into launch order. Each stage clears one group of risks before more budget moves. Keep the holdout untouched from day one through day 30.

1. In days 1–7, check your category against the live policy and estimate how many buyers use Free or Go.
2. In days 8–14, allow OAI-AdsBot and test the Pixel and Conversions API with shared event IDs.
3. In days 15–21, launch three to five variants per ad group, with natural-phrase hints and a holdout.
4. In days 22–28, pause weak variants and add a keyword-style ad group as a side test.
5. On day 30, compare paid placements and organic mentions on the same prompts across your AI search channels.

AirOps for ChatGPT ads best practices

AirOps AI Ads shows where your ads and competitors' ads appear across tracked prompts. Ad Load %, Paid Share of Voice, and Ads Library turn that view into brand-prompt defense. Observed ads also cover Google AI Mode answers. You create and update on-brand ads with Playbooks and your Brand Kit, following your hint, spec, and policy rules. Spend, impressions, clicks, CTR, CPC, and CPM connect to the same market data behind your organic AI search work.

AI Ads is live for Enterprise customers, starting with ChatGPT Ads. AirOps ad experts can help you choose your first opportunities, set up the channel, and support execution and measurement. Read how AI Ads works before you plan your first campaign.

Work with us