Platform information: ChatGPT Ads are developing and platform requirements can change. This article was last reviewed on 11 September 2026. Shays Media rechecks material platform information before relying on it for client decisions.
A click is not a sale
That's the sentence worth holding onto through this entire article. A click tells you someone was interested enough to act on an ad. It tells you nothing about what happened next — whether they bought, enquired, abandoned the page, or simply changed their mind. Treating clicks, or click-through rate, as the measure of whether ChatGPT Ads is "working" is one of the most common and most avoidable mistakes in paid media generally, and it applies just as much here.
Metrics Ads Manager currently reports
Current official reporting guidance lists a defined set of fields inside the platform itself:
Impressions. How many times your ad was shown. Useful for understanding scale and reach, not for understanding whether it worked.
Clicks. How many people clicked. Measures whether your ad was compelling enough to act on — it says nothing about what happened after the click.
Spend. What you've actually paid. The baseline everything else gets measured against.
CTR (click-through rate). Clicks as a proportion of impressions.
Average CPC and average CPM. Cost per click and cost per thousand impressions — useful for understanding efficiency of delivery, particularly if you're running an impression- or click-based campaign, but neither is a commercial outcome in itself.
Conversions. Where conversion tracking is set up, the platform reports on the defined conversion event directly. This is where the picture starts to become commercially meaningful inside Ads Manager itself, and it depends entirely on having conversion tracking properly set up.
Not every commercial metric a business needs is a native Ads Manager field — the list above is what the platform itself currently reports.
Metrics you may calculate or connect to business data
Depending on your own tracking, CRM or ecommerce setup, you can build further metrics on top of what Ads Manager reports:
Click-to-conversion rate. Where the measurement basis is click-through activity, you can look at conversions attributed to clicks as a proportion of clicks. For impression-billed or view-through measurement, don't divide every reported conversion by clicks and call that a universal conversion rate — use the attribution basis that actually applies to the campaign.
CPA or CPL (cost per acquisition/lead). What each conversion actually cost you, calculated from spend and conversions. This is a genuinely useful number, but only as useful as the conversion tracking behind it — an unreliable tracking setup produces an unreliable CPA, however precise the figure looks.
Revenue or conversion value, and ROAS where meaningful. If you can attach a value to conversions — average order value, typical deal size — via your own systems, you can start comparing what you spent against what it generated. Return on ad spend is the most commercially direct of these numbers, but it's also the one most dependent on accurate underlying data connected from outside the platform, and it isn't meaningful for every business model (a long sales cycle or high-touch service, for instance, may not have a clean same-period ROAS figure at all).
Qualified-lead rate and downstream sale rate. For businesses where a "conversion" in Ads Manager is a lead rather than a sale, tracking what proportion of those leads turn out to be qualified, and what proportion eventually become paying customers, closes the gap between a platform-reported conversion and an actual commercial outcome. This lives entirely in your own CRM or sales process, not in Ads Manager.
The number underneath all of these: measurement confidence
Every metric above is only as trustworthy as the tracking behind it. If tracking is unreliable, the CPA figure is unreliable too — and so is everything derived from it. Before drawing any conclusion from performance data, it's worth being honest about whether you actually trust the conversion tracking generating it. This is why conversion-optimised campaigns require conversion tracking and a supported conversion event to be set up properly before you can optimise toward it at all — the platform is, in effect, insisting on the same discipline this article is arguing for.
Why sufficient data matters
A handful of clicks or one or two conversions will usually be too little evidence to support a confident general conclusion — though this isn't an absolute rule. There are cases where a single, very high-value conversion could matter commercially even from a small sample, particularly for a business with a small number of high-value customers. Generally, though, early data — particularly in the first days of a new campaign on a platform that's still developing its delivery and optimisation systems — tends to be noisy. Reacting to a bad first few days, or celebrating a good one, both carry the same risk: mistaking normal variation for a real signal.
This cuts against a natural instinct. When something isn't performing, the pull is to change it immediately. But a weak short-term period does not automatically justify a major change unless there's a clear fault, policy issue, tracking problem or commercial stop condition behind it — it might be exactly that: a short weak patch inside a longer pattern that would look different with more data. The discipline is to distinguish a genuine trend from noise before acting on it, which usually means waiting for a large enough sample, or identifying a specific cause, rather than reacting to the first sign of trouble.
Why a good CTR can coexist with poor commercial performance
This is worth stating directly because it surprises people who come from a mindset where a strong click-through rate feels like success. It's entirely possible to have an ad that generates a healthy CTR — people find it compelling enough to click — and still produce a poor commercial outcome, because the landing page doesn't convert, the offer doesn't match what was promised, or the audience clicking isn't actually the right audience for the offer. CTR measures the ad's ability to attract a click. It says nothing about what happens on the other side of it. Treating a good CTR as evidence the campaign is "working" skips the only part of the funnel that actually matters commercially.
The platform isn't CPC-only, and neither is measurement
It's worth noting that current OpenAI documentation supports more than click-based buying — CPM, CPC and conversion-optimised campaigns all exist, the latter available as either oCPC (conversion optimisation with click billing) or oCPM (conversion optimisation with impression billing, currently documented as a beta feature with access rapidly expanding). That matters for measurement too: if you're running a conversion-optimised campaign, the platform itself is already working toward a defined outcome rather than just clicks, and your own measurement should follow the same logic rather than defaulting to a CPC-only view out of habit.
Two related things worth checking on a live account rather than assuming: whether view-through conversion reporting is available for your campaign type, and what attribution settings actually apply to it. Both affect how a conversion gets counted and credited, which in turn affects every metric built on top of conversions — so it's worth confirming the attribution setup you're actually measuring against, rather than assuming it matches a different campaign or a different platform's defaults.
A commercial threshold, not a vanity metric
Before a campaign starts, it's worth deciding — honestly, and in commercial terms — what result would actually justify continuing to spend. Not "clicks are happening" or "CTR looks fine," but something tied to the business: a CPA the margin can absorb, a conversion volume that matters at your scale, a ROAS that makes sense against your economics. Deciding this in advance, before you're looking at real numbers and tempted to rationalise whatever they say, is what turns measurement into a decision-making tool rather than a running commentary.
Scale, Continue Controlled Testing, Hold-Fix, or Stop
Once there's enough reliable data to look at, the decision it should lead to falls into one of four categories:
- Scale — the evidence supports increasing budget or expanding activity, because the commercial threshold is being met with reasonable confidence.
- Continue Controlled Testing — the signal is promising but not yet conclusive; keep testing at current or similar levels rather than jumping to a bigger commitment.
- Hold-Fix — something specific and identifiable is holding performance back — landing page, offer, tracking, targeting — and the sensible move is to fix that one thing before deciding whether to continue.
- Stop — the evidence doesn't support continued spend, and continuing anyway would be spending on hope rather than on a case.
Shays Media treats Stop as a legitimate management outcome. If the evidence no longer supports continued advertising, continuing simply because a management service exists would not fit our approach. Measurement that never leads anywhere near a genuine Stop recommendation isn't really measurement — it's decoration around a decision that was already made regardless of the data.
The underlying principle
Measurement before scaling. Diagnosis before changing. Commercial outcome before vanity metrics. Those three together are the actual discipline behind everything in this article — the specific metrics matter less than the order you look at them in and what you let them tell you.
Check Your Fit if you're deciding whether to start, or get in touch about Setup + Management if you have a live campaign and want proper measurement and Scale/Continue/Hold-Fix/Stop decisions built into how it's run.
A few questions people ask at this stage
What's a good click-through rate for ChatGPT Ads? Shays Media does not publish a universal ChatGPT Ads CTR benchmark at this stage. A single benchmark without context can be misleading, and we do not yet have a sufficiently broad Shays Media dataset to present one as our own evidence. More importantly, a good CTR on its own doesn't tell you the campaign is commercially working, for the reasons covered above.
How long should I test ChatGPT Ads before deciding? Long enough to gather a sample you can actually trust, which varies by budget and conversion volume rather than being a fixed number of days. Shays Media generally recommends allowing up to 90 days for a controlled initial evaluation where the evidence continues to justify testing — though that's not a commitment to keep spending for 90 days regardless of what the data shows; a Stop recommendation can come sooner if the evidence supports it.
What happens if the ads aren't commercially viable? Then the honest recommendation is to stop, or to hold and fix a specific identified problem before continuing. We're not going to keep a client spending on a channel the evidence doesn't support, because that isn't a defensible way to manage someone else's budget.