Why Your ChatGPT Ads Campaign Is Not Spending
By Keshav Parsai 9 min read
In brief
Diagnose a ChatGPT Ads campaign that is not spending by separating zero delivery, underdelivery and sudden delivery loss.
Last verified: 12 September 2026 | Version: 1.0 | Next scheduled review: 12 October 2026
Start by classifying the symptom. Zero impressions, impressions without clicks, and partial spend against budget are three different faults with almost no shared causes, and treating them as one problem is why most underdelivery investigations go in circles.
This article is about underdelivery specifically. Rejections, disapprovals, policy failures and account-level blocks are covered in the ChatGPT Ads troubleshooting post and are not repeated here.
Step one: which symptom do you have
| Symptom | What it rules out | Where the cause lives |
|---|---|---|
| Zero impressions | Not a creative problem, not a bid efficiency problem | Eligibility: verification, dates, review status, targeting breadth, or a bid under the clearing price |
| Impressions accruing, no clicks | Not an eligibility problem, the campaign is in the auction | Creative and match quality, or a very small sample being misread |
| Spending, but well under daily budget | Not an eligibility problem | Inventory available to your targeting, bid position, or genuinely thin matching surface |
Get this classification right before touching anything. The most expensive mistake in a stalled campaign is rewriting creative to fix a verification problem.
Zero impressions: the eligibility chain
Nothing serves until every link holds. Check them in this order, because the early ones are binary and cheap to confirm while the later ones require judgement.
Business verification and billing. ChatGPT Ads accounts are postpay and require a legal business name, business details, a payment method, account branding and completed verification. An incomplete verification does not throw a campaign-level error; it stops delivery. Advertiser eligibility follows the legal entity country. OpenAI currently lists 52 self-service countries, including India, while the UAE is not listed. Confirm the buyer location in the campaign picker.
Campaign dates. Obvious, routinely wrong. Check that the start date has passed, the end date has not, and that the dates are being interpreted in the timezone the account uses rather than yours.
Ad review status. Ads enter review. An ad group whose only ad is still in review is a live ad group with nothing to serve. OpenAI does not publish a review duration.
OAI-AdsBot access. OpenAI's ads crawler must be able to reach your destination URL. If it is blocked in robots.txt, by a firewall rule, by bot-protection tooling or by a staging password, ads can be rejected. This one is worth checking even when you are confident, because the block is frequently added by someone else: a security vendor, a CDN default, or a WAF rule shipped in an update.
Targeting breadth. Location targeting can go to country, region, DMA or postal code, and platform targeting splits into Android app, Android web, desktop web, iOS app and iOS web. A campaign targeting three postal codes on iOS app only has combined two narrow filters multiplicatively. Widen one at a time so you learn which one was binding.
The bid. Last, not first. A bid below the clearing price produces zero impressions in a relevance-weighted auction. Guidance circulating in 2026 puts CPC starting bids at three to five dollars with a 60 USD default max CPM, reported in agency write-ups citing OpenAI rather than published as a floor. If everything above is clean and impressions are still zero, this is the remaining candidate.
Spending under budget: the inventory question
This is the symptom most advertisers misdiagnose, because underspend looks like a settings problem and is often a supply problem.
Delivery on this channel has been uneven and independently reported as such. Digiday reported in 2026 that early advertisers saw severe underdelivery, including one that spent 2,500 USD against a 250,000 USD commitment over four weeks for roughly 200 impressions, and an agency executive describing OpenAI as able to push roughly 100 USD per client per week until mid-April. Fill rates were reported to have climbed 30 to 50 percent from launch levels by late April. One summit attendee put it as less than 10 percent of the eligible base being in the pilot. All anonymous, all reported, and all pointing the same way: this channel has had less inventory than advertisers expected.
Two structural facts compound it. Ads serve to Free and Go plan users, not to paid plans, which removes a segment of the user base before any targeting is applied. And users can opt out of ads, with take-up unpublished. Your addressable population is smaller than the ChatGPT user count by two factors you cannot measure.
So the checks for partial spend are different from the zero-impression checks.
- Is your hint theme genuinely narrow? A theme describing a rare situation competes for few conversations. This shows up as steady but small delivery rather than as a fault.
- Is your targeting stacking filters? Postal-code location plus a single platform plus a narrow theme multiplies down fast.
- Is the bid holding you in a low position? Raise it one increment and watch impressions against average CPC. If impressions rise and CPC stays flat, you were bid-constrained. If both rise together, you have bought into auctions you were losing on relevance.
- Has anything changed on the destination URL? A URL that stopped resolving cleanly for OAI-AdsBot can degrade delivery without producing a visible rejection.
If all four are clean and spend still sits well below budget, the honest conclusion is that the available inventory for your theme is smaller than your budget. That is not a fault you can fix in the account.
What not to do while diagnosing
Do not change more than one thing per day. With no query data and small daily volumes, two simultaneous changes give you one outcome and two explanations.
Do not rewrite context hints first. Nothing is reported at hint level, so a hint change is the least readable intervention available. It also resets any sample you were accumulating. Exhaust the binary checks before touching anything interpretive.
Do not raise the budget to fix zero impressions. A campaign that is not eligible to serve will not serve harder with more money behind it. Budget affects how much a serving campaign can spend, not whether it can enter.
Do not conclude no clicks means bad creative after 200 impressions. At a CTR anywhere in the reported range for this channel, 200 impressions buys you an expected click count near enough to one that zero clicks is an unremarkable outcome.
The reckoning: some of this is unfalsifiable from the account
The uncomfortable part is that the final diagnosis on the list, thin inventory, cannot be confirmed from inside Ads Manager.
There is no impression share metric, no lost impression share, no auction insights and no inventory forecast. A campaign delivering 40 percent of its budget looks identical whether the cause is a supply shortage, a bid position problem or a theme nobody discusses. You can distinguish bid position from the other two by raising the bid and watching, but you cannot separate supply from theme narrowness at all with the instrumentation available.
That leaves one workaround, and it is crude. Run a deliberately broad ad group alongside your narrow one, same bid, same period, same targeting. If the broad group delivers fully and the narrow one does not, the constraint is your theme. If neither delivers, the constraint is upstream of you. This costs a second ad group's budget to answer a question the platform should answer for free, which is a fair description of operating on this channel in 2026.
What we cannot tell you
- How much inventory exists for your theme. No forecasting tool, no impression share, no auction insights.
- How long ad review takes. OpenAI publishes no review duration.
- What proportion of users have opted out of ads. The opt-out exists; take-up is unpublished.
- The clearing price in your category. Not reported and not inferable from the account.
- Whether delivery has stabilised. Reported fill rates improved through 2026, but no current figure is published and the reported ones are anonymous.
Quick answers
Why is my ChatGPT Ads campaign showing zero impressions? Work the eligibility chain in order: business verification and billing complete, campaign dates live, ads out of review, OAI-AdsBot able to reach your destination URL, targeting not stacked too narrowly, and only then the bid. Creative is never the cause of zero impressions.
My campaign is spending far less than its daily budget. Why? Usually inventory rather than settings. Ads serve only to Free and Go plan users, users can opt out, and reported fill rates were still improving through 2026. Check theme breadth and stacked targeting filters, then test a bid raise.
Will raising my budget fix a campaign that is not delivering? No. Budget limits how much a serving campaign spends. It does not make a campaign eligible. A campaign at zero impressions has an eligibility problem, and more budget behind it changes nothing.
Could OAI-AdsBot be the reason my ads are not running? Yes. OpenAI's ads crawler must be able to reach your destination URL or ads can be rejected. Blocks are often added by a CDN, firewall or bot-protection tool rather than by your robots.txt.
Is my bid too low? Possible, and it is the last thing to check rather than the first. Circulating guidance suggests three to five dollars CPC and a 60 USD default max CPM, reported rather than published as a floor. Raise in one increment and watch impressions against average CPC.
How long should I wait before treating underdelivery as a fault? One full day for zero impressions, since that is a binary state. For partial spend, four days, which is long enough for pacing to stabilise and short enough that you have not wasted a fortnight.
Sources
| Claim | Source | Tier |
|---|---|---|
| Account setup requires legal business name, business tax ID, payment method, favicon, logo, business verification; billing is postpay | OpenAI Ads Manager account setup, 2026 | Confirmed, primary |
| OAI-AdsBot must be allowed or ads can be rejected | OpenAI advertiser documentation, 2026 | Confirmed, primary |
| Location targeting to country, region, DMA, postal code; platform targeting across five surfaces | OpenAI Ads Manager, 2026 | Confirmed, primary |
| Ads serve to Free and Go plan users, not paid plans; users can opt out | OpenAI, 2026 | Confirmed, primary |
| Reported serving markets: US, UK, Canada, Australia, New Zealand, Japan, South Korea | Reported, 2026 | Reported |
| Three to five dollar CPC guidance, 60 USD default max CPM | Agency write-ups citing OpenAI guidance, 2026 | Reported |
| 2,500 USD spent of a 250,000 USD commitment, roughly 200 impressions over four weeks; roughly 100 USD per client per week until mid-April; fill rates up 30 to 50 percent by late April; under 10 percent of the eligible base in the pilot | Digiday, 2026, anonymous advertiser and agency sources | Reported, third party |
| No impression share, lost impression share, auction insights or inventory forecast | OpenAI Ads Manager, by absence | Absent |
| Broad-group control test to separate theme narrowness from supply | InPromptAds | Inference, ours |
Related reading
- ChatGPT Ads Troubleshooting: Documented Failure Modes
- Raise the Bid or Improve Relevance?
- Is OAI-AdsBot Reaching Your Site? (checker)
- The First Fourteen Days of a ChatGPT Ads Campaign, Decision by Decision
- ChatGPT Ads Eligibility: Who Can Advertise
Changelog
12 September 2026, v1.0. First publication. Splits underdelivery into three symptoms with separate cause lists, orders the eligibility chain from binary checks to judgement calls, and states that supply and theme narrowness cannot be separated from inside Ads Manager.
Field kit
Tools
Site
OpenAI (primary)
Keshav studies how AI systems retrieve, verify, and cite brand information. At InPromptAds, he leads source research and turns platform documentation into practical guidance for advertisers.
