A $3–5 bid per click sounds modest until you multiply it by the conversion rate. One formula decides which offers can run, and it eliminates most popular affiliate offers:
Break-even CPC = payout per conversion × conversion rate. If that figure is below $3, the offer loses money at the lowest bid OpenAI recommends.
The calculation, with real offers
- SaaS paying 30% recurring, first-month payout $40, CR 1%: break-even CPC $0.40. Real price $3 → losing 7.5×.
- Hosting paying $100 per referral, CR 1%: break-even CPC $1.00. Real price $3 → still losing 3×.
- High-ticket SaaS or a course bundle, payout $150, CR 2%: break-even CPC exactly $3.00 → it starts to work.
- US insurance PPL (case-by-case group), payout $100, CR 3%: break-even CPC $3.00 → works, if approved.
- Amazon-style offer, 4% of a $60 basket = $2.40, CR 3%: break-even CPC $0.07 → losing 41.7×. No CPC change rescues this offer.
This yields a compact rule: at 1% CR you need a $300 payout to break even; at 2% CR, $150; at 3% CR, $100. In other words, at a $3–5 CPC, only offers paying $100 or more — or with a very high conversion rate — are worth trying.
What a conversion actually costs
One figure that is easy to misjudge by feel: the cost of acquiring a conversion. At 1% CR, a $3–5 CPC means $300–500 per conversion, not a few tens of dollars. That only comes down to something comfortable when the conversion rate is much higher — and a high conversion rate is built with a landing page and an audience, not by picking a different offer.
This is why the right order is: settle the vertical first, then build the landing page to lift CR, then work out the payout you need. Picking the offer first and hunting for a high CR afterwards is backwards.
Minimum budget and the credit
- The $25/day floor at a $3–5 CPC buys 5–8 clicks a day. That is a very small amount to test with, but also a very small amount to learn from — 5 clicks a day will not establish a real CR within a week.
- 30 days at the $25 floor is $750. After the $500 credit on $500 of spend, the first test costs roughly $250–500, depending on when the credit applies.
- To get statistically meaningful data you need enough conversions — at a 2–3% CR that means a few hundred clicks, i.e. $1,000–1,500 of ad spend. That is the real threshold for "knowing whether the model works".
Three traps that break the calculation
- You cannot bid on the advertiser's brand. That is the cheapest converting channel in affiliate marketing, and many affiliate programmes forbid brand bidding. Losing it means every click is cold traffic.
- ChatGPT Ads bills per valid click, not per conversion. There is no stable CPA bidding — so the conversion risk sits entirely with you, not shared with the platform.
- The landing page must be clean end to end. An ad pointing at a policy-violating page kills the ad, and the account can be flagged.
The practical conclusion: do not start with a $20–40 payout offer — the arithmetic above shows a certain loss. Either pick a digital/education offer paying $100 or more, or build a high CR first on a cheaper channel before putting money into a $3–5 CPC.
Other posts in the series: what ChatGPT Ads is, which verticals are blocked, account setup and the first 30 days. Payout and tax mechanics for affiliates in Vietnam are covered in the Amazon affiliate series.
Source of the figures: the break-even script at /mnt/blockstore/hkd/blog_numbers_check.py — every number above is reproducible from the script, not estimated. Policy sources: Ads in ChatGPT: The Basics, Ad policies.