ChatGPT Ads Retires Max CPC on Jan 1, and Your Per-Click Ceiling Goes With It

ChatGPT Ads Retires Max CPC on Jan 1, and Your Per-Click Ceiling Goes With It
Max CPC worked like a speed limiter. Target CPC only promises the average, and the old cap retires January 1, 2027.

OpenAI told ChatGPT Ads advertisers that Max CPC and Max Bid retire on January 1, 2027. Target CPC and Target CPA replace them and become selectable in Ads Manager on November 9. Target CPC lets OpenAI bid above your number in any single auction, so the hard per-click ceiling goes away. Use the seven-week overlap to set targets from your real CPC data.

The news came by email to advertisers and was first reported by Search Engine Roundtable on November 5. Barry Schwartz also posted the details on X. On paper it reads like housekeeping: two new bid strategies, a rename, and a retirement date. I think it matters more than the email suggests, mostly because of one sentence OpenAI put in it.

What actually changed (and what quietly didn't)

The timeline, as OpenAI laid it out:

  • November 9, 2026: Target CPA and Target CPC show up in Ads Manager when you create or edit a campaign. Picking a cost target is optional for now.
  • New conversion campaigns: Target CPA replaces Max Bid. You set an average cost per conversion (a purchase, a lead) and the system adjusts bids toward it.
  • New click campaigns: Target CPC replaces Max CPC. You set an average cost per click.
  • January 1, 2027: Max Bid and Max CPC retire. Conversion campaigns still on Max Bid move to Target CPA, and click campaigns still on Max CPC move to Target CPC.
  • The rename: "Maximize results" (autobid) is now called intelligent bidding. OpenAI calls this part of its "intelligent-by-default" approach.

Existing campaigns keep running as they are until the cutoff. That part is generous. It gives you roughly 53 days to compare the two modes side by side, which is more runway than ad platforms usually give before a forced migration.

Here's the sentence that matters, quoted in the SER writeup: "Target CPC lets our system bid above or below your target in each auction."

Under Max CPC, a $4 bid meant you never paid more than $4 for a click. Under Target CPC, $4 is the number OpenAI tries to average. Any single click can cost more, as long as the blended number lands somewhere near your target.

An average is not a limit.

Why losing the ceiling stings more on ChatGPT than on Google

Target bidding isn't new. Google's own Target CPA documentation says it plainly: "Some conversions may cost more than your target and some may cost less." Search marketers have lived with that for a decade.

The difference is the guardrails around it. Google lets you set bid limits on portfolio Target CPA strategies, though it doesn't recommend them. The email OpenAI sent, at least as reported, says nothing about bid caps on the new strategies. Maybe one shows up in the Ads Manager UI on November 9. I wouldn't plan around it until I see it.

Then there's the price level. When OpenAI turned on CPC buying in April, Digiday reported a bid range of $3 to $5 per click, and Search Engine Journal noted that CPMs had already slid from $60 to as low as $25. Earlier this week we worked through the revenue math and found ChatGPT ad clicks look worth about $0.26 on a revenue-per-visit basis. That's a rough estimate, but even if it's off by 3x, a $3 suggested bid is aggressive.

So the one control that stopped a bad auction from getting expensive is the control being retired. Think of Max CPC as a speed limiter. Target CPC is cruise control that promises you'll average 60 and is allowed to do 90 on the downhill stretches. Most days that's fine. You just can't see the downhills coming.

And to be fair, this isn't purely a cash grab. Average-based bidding seems to deliver more volume on most platforms, because the algorithm can pay up for the clicks it thinks are likely to convert and go cheap on the rest. The problem is that ChatGPT Ads is young, the signal is thin, and nobody outside OpenAI really knows how the auction weighs relevance yet. We dug into that in how the ChatGPT Ads auction picks winners, and honestly the picture still isn't very clear.

The seven-week test I'd run before January

Personally, I wouldn't wait for the January 1 auto-migration to make the decision for me. The overlap window is the only time you'll have both modes running against the same audience, so use it.

  1. Pull your trailing 14-day actual CPC for each click campaign. Use the CPC you've actually paid, not your Max CPC setting, and definitely not OpenAI's suggested bid.
  2. On November 9, clone your best click campaign and set the clone to Target CPC at that actual CPC. Leave the original on Max CPC.
  3. Split the budget 50/50 and leave both alone for at least 14 days. Fiddling resets whatever learning is happening.
  4. Compare three things: average CPC against the target, total clicks per dollar, and daily spend pacing.

For a benchmark, my rough rule: if the Target CPC clone's average CPC lands within 10% of the target and it buys at least as many clicks per dollar as the capped campaign, move everything over before January and don't look back. If the average runs more than 15% over target for a full week, lower the target and treat the clone as the expensive one.

The other failure mode is quieter. OpenAI warns that "restrictive targets or weak conversion signals can reduce ad delivery and leave budget unspent." If the clone is spending less than about 70% of its daily budget, the target is probably too tight. Raise it in small steps rather than one big jump. Google's Search Ads 360 guidance on bid strategy learning recommends adjusting targets "gradually over several iterations," and I'd assume the same logic applies here until OpenAI says otherwise.

Conversion campaigns live or die on the pixel

Target CPA is only as good as the conversions it can see. OpenAI's email tells advertisers to confirm the Pixel or Conversions API is actually sending the selected event to the linked data source, and to work through any Event Quality Score (EQS) warnings in Ads Manager.

That sounds obvious, and I'd still bet a good chunk of accounts have a gap somewhere. A purchase event firing on the wrong page, a CAPI integration that dedupes badly, that kind of thing. On Max Bid, a broken pixel mostly hurt your reporting. On Target CPA, it hurts your bidding, because the system is steering toward conversions it can't count.

Do the EQS check before November 9. It takes maybe ten minutes per account and it's the cheapest insurance you'll buy this quarter.

Intelligent by default, one lever at a time

Zoom out and there's a pattern. Last month OpenAI made creative enhancements the default on October 30, which meant your ads and landing page content could get rewritten unless you opted out. Now manual bid caps are going away and autobid gets a friendlier name. Each change on its own is defensible. Together, they're a steady handoff of control from advertiser to platform, which is roughly the path Google and Meta took, only compressed into about nine months instead of nine years.

My prediction: by the end of Q2 2027, more than half of ChatGPT Ads spend will run on intelligent bidding with no cost target at all, mostly because it's the default and the targets are a hassle to manage. If that happens, the advertisers who kept a target (and kept testing it) should have a much clearer view of what a ChatGPT click is actually worth than everyone else.

Something I keep going back and forth on: is this good for small advertisers? Probably yes, if their conversion tracking is clean, since target bidding tends to reward clean signal over budget size. Probably no if they were relying on Max CPC as a hard budget brake on a channel they don't trust yet. Maybe both, depending on the week.

Keep a record of the cap you're losing

Before January 1, write down your current Max CPC settings and your actual CPCs by campaign. Screenshot them if you have to. Once the migration happens, that history is the only baseline you'll have for judging whether Target CPC is costing you more per click than the old cap did.

I don't expect this change to blow up anyone's budget overnight. It just takes away the one number you could point to and say "it never goes above this," and it's worth knowing what that number was before it's gone.