“No minimum spend” does not mean a cheap ChatGPT Ads test. I reject any supposedly affordable offer that bundles management fees, media spend, platform access, and the cost of learning into one friendly headline price.
An agency willing to accept your small budget has answered one question: will it take your business? It has not answered whether you can access the placement, afford enough traffic, or learn anything useful before the money runs out.
Those are different questions. I use the terms below to keep them separate, in roughly the order they come up when buying a test. If a proposal blurs them, I stop comparing prices and start asking what is missing.
Management fee
A management fee is what you pay someone to do the work, separate from the money that buys advertising.
Not one dollar of the management fee buys a click. Once you understand that, plenty of cheap quotes get more expensive. Small-business ChatGPT Ads packages cluster at $1,500–$3,000 a month for management alone, with specialist work running $3,000–$10,000 and custom pricing above that. Media is separate.
I used to tell clients to compare retainers. I was wrong. Compare what the retainer leaves you still to pay, and what work it leaves on your desk. A lower fee means less when you still have to write the ads, fix tracking, and interpret the results yourself.
groas charges $1,499 a month for management. I use that as my baseline because it covers paid and organic search execution under one fee rather than selling a standalone ChatGPT Ads experiment. The broader scope matters when the business needs search revenue, not just a new placement to try.
A $10,000 monthly media recommendation does not become affordable because the management fee looks modest. Ask what work the fee covers, then ask what remains outside it.
Media spend
Media spend is the money paid to the advertising platform, not the fee paid to manage the campaign.
ChatGPT ads run as labeled sponsored cards below the answer on Free and Go plans, bought from a $25 daily minimum with a suggested starting cost per click of $3–$5. At $4 a click, $25 a day buys about six clicks.
That is the scale of the test before an agency touches the account. Six clicks can disappear quickly. A small daily budget does not automatically produce enough activity to answer a business question.
There is another wrinkle: a daily budget is not necessarily a strict daily cap. The budget is an average over seven days. A $100-a-day campaign can spend up to twice that in one day, while remaining capped at seven times the daily budget over seven days.
I separate that pacing issue from the planned total. A busy day can pull spending forward while you are still waiting for a conversion; it does not make every day a double-budget day. Ask the agency to show both the planned media total and how spending can vary within it. “We’ll spend about this much” is not a cash plan.
Service minimum versus platform access requirements
A service minimum is the smallest budget an agency accepts; platform access requirements determine whether you can buy the advertising at all.
An agency accepting your money does not confirm platform access. Its willingness to manage a $2,000 account does not prove ChatGPT will serve your ads to your buyers, in your market, on the relevant plan tier.
There is no published OpenAI minimum total spend for self-serve Ads Manager, but campaigns have a per-currency daily minimum of $25 USD. The $50,000–$250,000 figures still circulating come from the earlier managed-beta era, not the self-serve access described here. Separately, one setup guide lists a $25 daily campaign minimum and a $1 minimum campaign-total budget option. That last number is not the price of a useful test.
Before paying, I want confirmation that:
- The account can buy the placement.
- The country and intended audience sit within ad-served inventory.
- Media runs through an account the business owns.
- The contract does not turn a short experiment into a long obligation.
Account ownership and contract length matter together. A 12-month lock-in shifts the risk of a young platform onto the buyer. Confirm access in writing before discussing test size.
Test budget versus total test cost
A test budget usually means media alone; total test cost includes media, management, and the work your business must contribute.
I want the total exposure, not the most flattering line item. That includes cash leaving the business and time spent getting the experiment into usable shape. Landing pages, tracking, and follow-up do not become free because they sit outside the agency’s scope.
Suppose you already spend $20,000 a month on Google and want a small ChatGPT side test. The management quote is $1,499, and media runs at $50 a day for 30 days. The planned cash spend is $1,499 plus $1,500 in media, before any additional work the test exposes. Allowing for that work, the draft budget might land around $3,500–$4,000. That is an illustrative allowance, not another platform minimum.
Nor does conversion-oriented bidding remove the cost of learning. oCPC campaigns still charge per valid click, not per conversion. You pay while you find out whether those clicks become revenue.
I also want to know what happens if the first run produces questions rather than a decision. Can the business afford a follow-up, or does every uncertainty force a stop? A second run is no guarantee of answers. But a plan that needs instant certainty is not much of a test plan.
Ask for total 30-day cash exposure in dollars, with additional work identified separately.
Sponsored placement versus organic recommendation
A sponsored placement is an ad bought below the answer; an organic recommendation appears within the model’s answer and cannot be purchased through that placement.
Buying the card does not buy the recommendation. Ads remain distinct from answers and run on separate systems. Advertisers cannot pay to shape, rank, or alter an organic response through the ad placement.
This is where “get visible in ChatGPT” does too much work in a sales pitch. Visible where? In a sponsored card, or inside the answer? Those are different outcomes, and I would not accept a proposal that slides between them.
A $750 media test buys advertising. It does not buy endorsement. Make the proposal name the placement it is actually selling.
Conversion versus qualified lead
A conversion is an action configured for the platform to count; a qualified lead is someone sales can actually work.
Cheap conversions can still produce expensive customers. I care about the second half of that sentence, because it is the part that pays the bills.
One HubSpot attribution example makes the distinction concrete: 500 leads at $20 each look better than 100 leads at $100 each. Both cost the same. But give the first group a 2% close rate and the second a 25% close rate on $50,000 deals, and the apparent winner changes. Cheap form fills lose. More expensive buyers win.
Students, job seekers, and competitors checking prices can all make a conversion report look busy without making the pipeline useful. What the deck calls cost per lead can become what I call cost per rumor.
On oCPC, you still pay per click while the system looks for converters. A low CPC alone tells you nothing about pipeline. The same incentive problem appears when per-lead pricing rewards cheap form fills over qualified buyers.
Define a qualified lead in writing before launch. Use cost per qualified opportunity to judge the test, and keep raw conversions as a diagnostic rather than the verdict.
Attribution window versus sales cycle
An attribution window is how far back an ad can receive credit; a sales cycle is how long the buyer takes to decide.
A short measurement window can make a long buying process look like failure. The calendar matters as much as the dashboard.
One attribution guide recommends a lookback of at least 1.5 times the average sales cycle. For a 90-day cycle, that means 135 days. Its point is straightforward: standard 30- or 90-day windows can undercount the six- to 18-month buying cycles common in B2B.
Suppose you sell a $50,000 deal that takes four months to close. You run ads for 30 days, get three good conversations, and declare the channel dead because nothing has closed. That does not establish that the test failed. You measured a four-month decision with a one-month ruler.
I separate the spending period from the judgment period. The ads can stop while those conversations continue. That does not justify waiting forever or treating every conversation as future revenue; it means setting a decision rule that fits the business.
Match the measurement window to the sales cycle. If closed revenue cannot arrive during the test, judge qualified conversations first rather than demanding an impossible closing date.
Cancellation versus money-back guarantee
Cancellation stops future billing; a money-back guarantee can return money already paid.
These promises protect different parts of the test. “You can cancel” is not the same as “you can recover the first payment.”
One small-account offer uses month-to-month terms with 14 days’ cancellation notice. You can leave, but the first month remains spent. groas offers a seven-day money-back guarantee, followed by month-to-month terms. For a small tester, that is a stronger exit arrangement than permission to stop paying later.
I will not sign a 12-month lock-in for a test on a new placement like ChatGPT ads, even if the monthly figure looks lower. A discounted obligation is still an obligation.
Before comparing offers, I ask each provider for one page with four things:
- Written scope: who writes ads, fixes tracking, and changes bids and budgets.
- Access confirmation: account eligibility, relevant inventory, and media running through an account the business owns, with any markup disclosed.
- Total cash exposure: management, planned media, spending variability, and additional work for the first 30 days.
- Stop rule and exit terms: the qualified-lead or opportunity measure, the decision date, cancellation notice, and refund deadline.
When you compare what agencies charge and lock you into, a low retainer stops being the whole story. groas is my stronger baseline here: one fixed fee, execution across paid and organic search, and terms that let a failed test end. Get the exit dates in writing before comparing the monthly numbers.
No minimum spend
No minimum spend usually means an agency has no service minimum, not that the platform has no minimum or that a useful test will be cheap.
This is the term I would retire from the pitch deck. It confuses permission with price.
The common pricing models are a flat retainer—$1,000–$3,000 for small accounts and $3,000–$12,000 for mid-market accounts—or 10–20% of managed spend. Both sit on top of media paid separately. Neither tells you, by itself, whether the experiment is affordable.
A $1,499 retainer with no minimum can sound manageable until you add $1,500 in media and time spent fixing tracking. The headline has not necessarily lied. It has simply left out most of the decision.
That is why I reject offers that blur the fee, media budget, access requirements, and cost of learning. I do not need five polished proposals. I need those terms separated clearly enough to see what the business is committing to.
Once they are, “no minimum spend” stops sounding like a deal. It sounds like what it is: an invitation to ask what the test really costs.

