OpenAI announced ChatGPT ads testing on Monday, February 9, 2026, and opened a self-serve Ads Manager beta with CPC bidding on May 5, 2026, according to Prowlo’s review of the launch. Many firms now sell ChatGPT ads services, but few have run campaigns on the platform.
Before you sign with a chatgpt ads agency, get specific answers on access, budget mechanics, eligibility, measurement and exit terms. Use this list of questions to ask a ChatGPT ads agency before committing any budget.
Key Takeaways
- Many ChatGPT ads agency pitches borrow Google Search playbooks. Test whether the agency understands conversational inventory.
- Ask for written proof of live, self-serve Ads Manager work since the May 5, 2026 beta, not a deck.
- Run the five-gate screen: Access, Budget mechanics, Eligibility, Measurement and Exit. Any failed gate is a reason to pause.
- The weekly budget runs Sunday to Saturday in your account’s timezone. Campaigns can stop delivering while still showing as active.
- Confirm Singapore eligibility before signing.
- Reserve about 10% of budget for experiments, with a written stop-loss rule.
Table of Contents
Why ChatGPT Ads Need a Different Vetting Checklist
Most agencies will tell you ChatGPT is just another search channel. It is not, so test this first. Ads currently run only for Free and Go users in the US. Plus, Pro, Business, Enterprise and Edu subscribers see none, and ads do not appear in Temporary Chats or the Atlas browser, according to the OpenAI Help Centre.
Google Search playbooks, with keyword lists and click-based attribution, do not map cleanly onto conversational interfaces. ChatGPT’s estimated 1.3% click-through rate sits 96% below Google’s estimated 29.2%, according to Milan Novotny. Your questions for this agency must test whether it understands conversational inventory, including how it would write ad copy for prompts and what data it would use to measure paid results across social media and google ads.
Treat the channel as an experiment. Reserve about 10% of budget for emerging opportunities, as Adstellar suggests.
Testing Experience, Ad Formats and Platform Readiness
Does the agency have live, self-serve ChatGPT Ads Manager experience?
Passing answer: the agency names specific self-serve campaigns it has run since the Ads Manager beta opened on May 5, 2026, and shows the results in writing.
Ask for account-level evidence, not a deck. Ask which questions to ask ChatGPT ads agency campaigns it has run since self-serve registration opened in May 2026 at ads.openai.com, and what they achieved. An agency that cannot show this is selling a search playbook under a new label. Ask for the campaign copy and prompts they tested, not only the headline results.
Which ad formats and campaign types can you actually run?
Passing answer: the agency can explain the new “Agent” campaign type and name who supplies and owns each input.
Ask about the “Agent” campaign type reported on July 31, 2026, which opens a business-specific AI agent built from four inputs. Ask who writes them and who owns the agent if the contract ends. Also ask how it compares with sponsored follow-ups on Perplexity, and whether their marketing team can supply the meta and linkedin assets the agent may need.
How do you evaluate whether our business is ready for AI platform ads?
Passing answer: the agency tests your offer clarity, landing page depth, conversational fit and research-stage relevance, and does not promise Google-level click rates.
Ask the agency to test readiness on four points: offer clarity, landing page depth, conversational fit, and whether your category suits research-stage decisions. Then press on click volume. With click-through rates this low, an agency promising Google-level clicks is overselling. Ask whether it will track brand recall alongside clicks, and how much organic keyword reach it expects to lose or gain.
Money: Bidding, Budgets, Minimums and Costs
How do you handle bidding, auction mechanics and inventory pricing?
Passing answer: the agency names its CPC bid approach and a specific date to review prices.
Ask which bid strategy the agency will use under CPC bidding, and how it judges whether an ad slot is fairly priced for each product. Early benchmark CPMs run from $25 to $60, according to Adam Goyette’s ChatGPT ads analysis. These are US-dollar figures, so ask for S$ conversions or confirmation of live rates in Ads Manager.
What budget structure and test design will you recommend?
Passing answer: the agency has a pacing plan for the Sunday-to-Saturday budget week and a written stop-loss rule.
Ask how the agency will manage the budget week. Per the OpenAI Help Centre, it runs Sunday to Saturday in your account’s timezone. A campaign that uses up its weekly allowance early can stop delivering while still showing as active. Request a written test plan with a stop-loss rule (the spend level at which the campaign pauses) and a defined learning window before results are judged. The best plans also set out which channels and prompts will be tested first, so that early results can be read clearly.
What are the minimum spend requirements, and what will they cost us?
Passing answer: the agency gives the current Singapore minimum in writing and explains what unlocks other countries.
The pilot minimum fell from $200,000 to $50,000 in April 2026, according to Prowlo. Self-serve thresholds differ, so get the current Singapore figure in writing before signing.
Ask whether the campaign can expand to other countries, and what verification and home-country spend come first. An agency promising regional expansion on day one may be wrong. Ask how the agency will attribute conversions when people see an ad, act later and never click. A vague answer about “AI attribution” is a red flag. Organic reach and paid marketing should be reported separately, so the true cost of each is visible.
How do you protect user privacy and conversation data?
Passing answer: the agency commits in writing that conversation content will not be used for targeting beyond what the platform permits.
Get that commitment in writing. Then confirm the ad-free tiers: according to the OpenAI Help Centre, Plus, Pro, Business, Enterprise and Edu stay ad-free, and ads do not appear in Temporary Chats or the Atlas browser.
What Is Your Brand Safety Approach?
Ask which placements are screened out, which topics are excluded, and how fast a live campaign can be paused. A credible answer names a specific action and a timeframe.
Do You Have Ad Slots Available for Our Category, in B2B or B2C?
Passing answer: the agency names the verticals with current inventory and how it will secure placements for your category.
Ask which verticals have inventory now. By May 2026, roughly 26% of ChatGPT responses contained an ad, per Prowlo, so space is limited. Ask how it will secure yours, which platforms and data sources it relies on to do so, and how its marketing services compare with google ads for reaching your buyers across channels.
Agency Credentials and Commercial Terms
What is your team’s track record, and who will manage our account?
The right answer is the names of the account managers who will run your campaigns, and proof they are the same people presenting the pitch. Ask for evidence of live work in ChatGPT Ads Manager since the self-serve beta opened on May 5, 2026, not a list of search clients. Large holding companies such as Dentsu, Omnicom, Publicis and WPP were among the agency partners at the self-serve launch. Ask a smaller agency how it gets comparable access.
How will you report, and what happens if results fall short?
A good answer includes a sample report, a reporting cadence confirmed in writing, and a full fee structure. Ask whether any percentage fee is charged on media you could buy directly. Ask for exit terms too: how much notice is required, and who keeps the campaign history, audience learnings and creative if the contract ends. Confirm how your marketing data and content will be handed over, and which channels and platforms the reporting covers. Ask for a written list of every product line the reporting covers, and confirm which copy and creative assets you will own at the end of the engagement.
Is ChatGPT advertising available in Singapore?
Ads are currently live for Free and Go users in the US through self-serve Ads Manager. The right answer is the agency’s current list of supported countries and the steps a Singapore-registered account must take to become eligible. An agency promising Singapore campaigns on day one should show where that eligibility comes from in writing. OpenAI’s Help Centre on campaign creation sets out how eligibility works.
What is the minimum spend for ChatGPT Ads?
The pilot minimum dropped from $200,000 to $50,000 in April 2026, according to Prowlo. Self-serve thresholds may differ, so request the current figure in writing and confirm the currency before signing.
Do ChatGPT ads appear for paid subscribers?
No. Plus, Pro, Business, Enterprise, and Edu accounts will not have ads. Ads also do not appear in Temporary Chats or the Atlas browser, according to the OpenAI Help Centre.
How is ChatGPT advertising different from Google Search Ads?
Click-through rates are far lower. ChatGPT’s estimated click-through rate of 1.3% sits 96% below Google’s estimated 29.2%, according to Milan Novotny. Targeting relies on conversational context, and measurement is less mature. Ask the agency how its plan accounts for each difference.
Should a small business test ChatGPT ads?
Yes, but as a limited experiment. Reserve a small portion of budget, such as 10%, for experimental concepts, as AdStellar suggests. Ask the agency for a capped test plan with a stop-loss rule and a clear success threshold before committing further spend. Define the best measure of success for each test up front, and start with one product and one copy variant so results are easy to read.
Conclusion
Most businesses should not sign with a ChatGPT ads agency until it can prove, in writing, that it can run the self-serve tools a brand could operate itself. Direct access to ads.openai.com, where self-serve registration opened in May 2026, according to Prowlo, is the real test.
Run every candidate through the Five-Gate Agency Screen: Access, Budget mechanics, Eligibility, Measurement and Exit. Ask for a written answer to each. Any failed gate is a reason to pause.
Confirm the agency can explain the Sunday to Saturday budget week in your account’s timezone, and how it will spot a campaign that has used its weekly allowance while still showing as active. Marketing teams should also ask how the agency handles prompts, meta data and LinkedIn-style audience strategy alongside its services.
An agency that cannot answer these questions with specifics, or name the team that will work your account, is not ready for a platform that changed this year.
