ChatGPT Ads Benchmarks: ROAS, CPC, and Cost From 15 Days of Real Spend

Over 15 days of real spend on a live account, the scaled ChatGPT Ads campaigns put in roughly $60K and returned about $89K in revenue. That is a 1.49x blended ROAS, at around $1.72 per click. Day to day the number was choppy, with ROAS swinging from roughly 0.2x to 2.9x, so the honest read is a rolling window and not any single day. The widely reported $3 to $5 cost per click did not hold for this account this early.

Here is what 15 complete days looked like on a live account in 2026, running the scaled ChatGPT Ads campaigns.

These are actuals from a running account, not a pricing-model estimate or a projection off the platform’s rate card. The cumulative view shows revenue pulling ahead of spend and holding the gap as the budget scaled.

Figure 1. Cumulative spend versus revenue across the 15 days. Cumulative revenue stays above cumulative spend the whole way and widens late, ending near $89K revenue on about $60K spend.

The scaled campaigns ran at a 1.49x blended ROAS over the 15 days. Spend went in around $60K and revenue came back around $89K.

One word in that sentence is doing a lot of work: scaled. This is the set of campaigns we pushed budget into once they showed signal, not an average across every test we launched. Stating the scope is what keeps the number honest. A blended ROAS that quietly folds in your throwaway tests tells you nothing you can act on.

At 1.49x this early, the read is a working signal rather than a mature-channel result. Part of that return comes from being in before the auction filled up. Inventory is cheap and competition is thin when a channel is weeks old, and early-mover economics flatter the math in ways that will not last. So treat 1.49x as proof the channel can return money today, not as a ceiling or a promise it stays here.

It is also a single number sitting on top of a wide spread. The blended figure smooths over days that ran well under 1x and days that ran close to 3x. That spread is the part most benchmarks hide, and it is where the real lesson lives, so it gets its own section below.

This account ran at about $1.72 per click across the 15 days.

That sits well under the figure circulating in the press. When the platform switched to cost-per-click pricing in April 2026, advertisers could set bids between $3 and $5 per click, and early coverage reported clicks landing in that range [1][2]. Ours came in at less than half of it.

The framing that matters: this is our account over that window, not a universal ChatGPT Ads CPC. The same forces that flatter early ROAS depress early CPC. Thin competition and cheap inventory keep the auction soft, and as more advertisers move in, the reported press range is the direction cost is more likely to travel. The CPM the platform launched at had already eroded sharply before the move to cost-per-click, which is the kind of repricing a young auction does [3].

Cost only matters next to what it produced. At roughly $1.72 a click and a 2.35% conversion rate, the unit math is what carried the campaigns to 1.49x. A cheaper click that does not convert is not a benchmark worth chasing. If you are setting up an account and want the mechanics behind these numbers, our companion guide on how to advertise on ChatGPT walks through the build.

Day to day, it was choppy. ROAS swung from roughly 0.2x on the worst days to about 2.9x on the best.

Figure 2. Daily spend and ROAS across the 15 days. Daily ROAS (the line) swings from roughly 0.2x to 2.9x while daily spend (the bars) scales up, which is the volatility a rolling-window read smooths over.

The chart makes the point better than a paragraph can. There is no smooth climb here, just a saw blade. The lesson is to judge ChatGPT Ads on rolling windows, not single days. A 0.2x morning does not mean the channel is broken, and a 2.9x afternoon does not mean you cracked the code. Both are noise until enough days stack up to show a trend, which is exactly why a one-week read will mislead you.

One honest caveat on the data. Day two was a launch-day budget dump, so we smoothed that day’s spend across the series to keep the daily picture readable. The totals are unchanged and the rest of the series is real and as choppy as it looks. We left the chop in because flattening it would hide the thing you most need to plan around.

Not for ROAS. The platform’s in-tool tracking reports cost per click and cost per acquisition, but it does not hand you return on ad spend, and even the cost numbers it does show come in under what actually happened.

That gap is the whole reason the 1.49x in this article is a number we can stand behind. The dashboard can tell you what a conversion cost, but it cannot tell you what that conversion was worth, so ROAS only shows up if you run your own first-party tracking and tie revenue back to spend yourself. We did, which is how we know the channel returned about $89K on roughly $60K. The platform’s own view would have stopped at a cost-per-acquisition figure, and it would have counted fewer conversions than actually landed.

The takeaway for anyone testing the channel is plain: stand up first-party tracking before you read the dashboard as truth. The in-tool numbers are fine for a directional read on cost, but ROAS, the number that decides whether the channel earns its budget, exists only if you measure it yourself. Our benchmark held up because we did.

Across our portfolio, ChatGPT Ads landed between search and paid social on efficiency this early. It cleared the bar we hold paid social to and ran behind mature paid search, which is a respectable place for a channel measured in weeks.

Read that as a practitioner’s view across accounts, not a published benchmark with hard numbers attached. Channel comparisons move with maturity, auction density, and the mix of accounts in front of them. The placement that holds today will shift as more advertisers enter and the early-mover discount fades. What is fair to say now is that the channel earned its spend next to the incumbents rather than sitting in a test corner, and that alone is more than most new ad surfaces can claim in their first month.

Plan for roughly $5K to $10K to get a read you can trust. That is enough ad spend across enough days to clear the daily noise and see a real trend instead of a lucky or unlucky week.

A small one-week test will mostly show you volatility, as the daily swing above makes clear. You need the days more than you need the dollars per day.

If you take one thing from these numbers, let it be this: use them to sanity-check your own account, not to predict it. What worked for one account over one 15-day window will not map cleanly onto every budget, vertical, or offer. These are the kind of accounts we run for clients, and even across them the spread is wide. Run your own window, measure it honestly, and judge the channel on the trend.

Do ChatGPT Ads actually work? For this account, yes. The scaled campaigns returned a 1.49x blended ROAS over 15 days of real spend. That is a working early signal rather than a mature-channel guarantee, and the result leans on early-mover economics that will not hold forever.

How much do ChatGPT Ads cost? This account averaged about $1.72 per click over the window, below the $3 to $5 range reported when the platform moved to cost-per-click pricing. Cost is account-specific right now and is more likely to rise than fall as competition enters the auction.

What is a good ROAS on ChatGPT Ads? This early, anything clearing 1x on a rolling window is a real result. We saw 1.49x blended, with daily swings from roughly 0.2x to 2.9x. Judge the channel on the rolling number, not the best or worst single day.

What is the average CPC on ChatGPT Ads? Reported figures cite $3 to $5 per click. Our live account ran about $1.72 over 15 days. Treat any published average as a moving target while the channel is young and the auction is still filling up.

Are ChatGPT Ads worth it for B2B or ecommerce? For accounts already spending more on search than on social, a $5K to $10K test is usually enough to get a baseline worth trusting. Smaller tests tend to surface volatility rather than signal, so budget for the days, not just the daily spend.

Why don’t these ChatGPT Ads benchmarks include a click-through rate? This report sticks to outcome metrics, ROAS and cost per click, because a single account’s click-through rate would not generalize while the channel is this young. CTR moves with placement, query intent, and creative, so an early CTR tells you less than what the spend actually returned. Judge the channel on the blended ROAS and the cost per click that produced it, not on engagement signals that have not settled.

What metrics should I track to benchmark ChatGPT Ads? Track cost per click, conversion rate, and blended ROAS on a rolling window, and run first-party tracking rather than leaning on the native dashboard. The platform reports cost per acquisition but not ROAS, and it under-reports, so the number that decides whether the channel pays only exists if you measure it yourself. Impressions and click-through rate are worth logging, but this early they say less than the revenue your spend returned.

1. Digiday, “OpenAI turns on cost-per-click ads inside ChatGPT” (April 21, 2026). https://digiday.com/marketing/openai-turns-on-cost-per-click-ads-inside-chatgpt/
2. Search Engine Land, “OpenAI adds CPC ads to ChatGPT” (April 22, 2026). https://searchengineland.com/openai-adds-cpc-ads-to-chatgpt-475148
3. The Next Web, “OpenAI’s ChatGPT ads just went cost-per-click, and the AI advertising war has its battle lines” (April 22, 2026). https://thenextweb.com/news/openai-chatgpt-cpc-ads-launch