Advertisers are spending real money on ChatGPT Ads, but they still don’t have the benchmarks or auction visibility needed to judge performance. That gap turns routine metrics — CPC, CPM, CTR — into signals you can’t interpret without extensive testing and external measurement.

What the Ads interface gives you — and what it hides

ChatGPT Ads Manager reports the basics: impressions, clicks, spend, CTR, average CPC, CPM and conversions. You can export CSVs and use the OpenAI Pixel or Conversions API for purchase, lead and sign-up tracking; OpenAI also provides modeled conversions when available. Ecommerce accounts can see attributed sales, cost per attributed sale and ROAS where the platform can attribute them.

Missing from the product are the comparative tools advertisers rely on to understand why performance moves: there’s no Auction Insights equivalent, no impression share, and OpenAI has not published cross-advertiser or cross-industry benchmarks. OpenAI itself says it does not yet have performance benchmarks across advertisers, industries or campaign types.

That matters because ChatGPT Ads runs a relevance-weighted, second-price auction. Ad selection depends on the conversational context and intent, plus signals from landing pages, creatives and advertiser-supplied “context hints” — short descriptors an advertiser can provide to guide relevance. Those context hints are not keywords and don’t guarantee delivery against particular words, audiences or conversations. Advertisers see their max bid and the price paid, but not the relevance scores, competing bids, or inventory pressure that produced the final price.

Early spend shows wide, sometimes contradictory outcomes

Published tests from advertisers illustrate how much results can vary.

  • Hostinger: after nearly $70,000 in testing, the team reported CPMs above $65 and uneven traffic quality. Specific use cases outperformed broad messaging, and CTR was a bigger concern than CPC.
  • Common Thread Collective (ecommerce test): scaled from $7/day to more than $1,000/day, spending $9,620 with an average $4.41 CPC and 0.94% CTR across roughly 136,000 weekly impressions. Estimated attributed revenue ranged from $19,000 to $38,000 depending on the attribution model, producing an estimated ROAS between 3.3x and 6.8x when validated with an external analytics tool.
  • B2B test (Floyd Blaikie): roughly CA$7,000 spend, averaging $9.29 CPC, $64.34 CPM and 0.7% CTR. External deanonymization of visitor data identified 146 organizations behind 336 paid clicks — only five matched the advertiser’s ideal customer profile.
  • Synter: $4,428.84 in spend produced an average $9.89 CPC overall, with large market differences: about $5.10 in the U.K., $10.62 in the U.S., and higher figures in lower-volume markets such as Australia and New Zealand.

Those examples point to three practical sources of variance: audience composition (who sees the ad), market-level competition and the types of conversational contexts where the ad is matched. With no auction-level reporting, advertisers must infer which factor is driving cost or quality changes from external analysis and segmented testing.

The $3–$5 figure is a starting bid, not a benchmark

A $3–$5 per-click figure has circulated widely. That number appears in OpenAI guidance as a recommended starting maximum CPC, not a platform average or benchmark. It’s also important because ChatGPT Ads defaults new ad groups to an automated “Maximize results” bid strategy; that setting optimizes delivery but does not guarantee specific CPC, CPA or ROAS targets. If you need a hard CPC ceiling, switch to manual bidding.

Remember too that ads currently reach a subset of ChatGPT accounts: Free and Go users are eligible to see ads; Pro, Business, Enterprise and Education accounts currently remain ad-free. Ads are also excluded from accounts identified as under 18. That restriction narrows the ad-eligible audience and affects how you should interpret user statistics when planning media.

OpenAI has expanded self-service access to Ads Manager across 52 countries and offers buyable ads through sales partners in 40+ markets, but available volume and CPCs differ materially by market.

Practical next steps for marketers

  • Run controlled A/B tests and segment results by geography, creative specificity and conversational placement. Many teams report broader messaging underperforming compared with narrowly targeted use cases.
  • Use independent attribution and analytics alongside Ads Manager to validate downstream revenue and customer fit; platform conversions alone may miss the full picture.
  • Consider manual bidding if you require a firm CPC cap. Track how the default “Maximize results” strategy changes delivery and downstream metrics.
  • Measure traffic quality against your ideal customer profile, not only clicks or CPC — especially for B2B and high-AOV offers.
  • Factor account eligibility into planning: Free/Go reach, excluded subscription tiers and age restrictions change who actually sees your ads.

What to watch next

ChatGPT Ads has moved quickly from prototype to real spend, but the platform’s lack of auction transparency and cross-industry benchmarks leaves advertisers interpreting raw metrics in isolation. For now, treat early results as directional intelligence: run segmented tests, validate outcomes with external analytics and prioritize traffic quality, not headline CPCs.

Watch for two practical product shifts that would change how you plan: any introduction of impression-share or auction-insights reporting, and the publication of cross-advertiser benchmarks by OpenAI. Either would make CPC and CPM figures far easier to evaluate across markets and advertisers.