If your ChatGPT ads are serving and nobody is clicking, the cause is usually one of five things, and four of them are visible from outside your account. Here is how to check each against real captured ad data.
Impressions without clicks almost always means a relevance mismatch, not a budget problem: your ad is surfacing on conversations it does not belong in, or it is competing against advertisers who fit those conversations better.
Four of these five are visible from outside your account, which is why an Ads Manager alone rarely explains a weak result.
This is the most common cause and the easiest to miss. A wide hint wins impressions on conversations your product does not actually answer, so the impressions are unwinnable from the moment they serve. Narrow the hint to the situation you genuinely solve, even though the impression count will fall.
See the real prompts that surface ads in your category →When several advertisers compete for the same conversations, attention splits and the best-fitting card wins the click. This is invisible inside your own Ads Manager — you can only see it by looking at who else appears in the category.
Check how crowded your industry is →The card appears inside somebody's answer. Copy written for a display placement reads as an interruption, and interruptions do not get clicked. Compare your title and description against the shape real advertisers are shipping.
Compare against real captured creatives →Some categories already have dozens of established advertisers. Others have almost none. If a well-resourced competitor owns the conversations you are bidding into, a nearby, less contested category often converts better for the same budget.
Find categories with demand but little competition →Impressions are the least meaningful number on a conversational surface, because eligibility is not relevance. Judge the campaign on clicks and downstream actions, and compare against the category rather than a generic cross-platform benchmark.
What you can and cannot measure →Relative competition observed in our captured sample. The more advertisers competing for the same conversations, the further attention splits.
| Most crowded categories | Competition |
|---|---|
| Business Consulting | High |
| SaaS & Productivity | High |
| Email Marketing & CRM | High |
| Cloud & DevOps | High |
| Translation & Interpretation Services | High |
By contrast, Reverse Mortgage & Senior Home Equity Conversion, Credit Cards, Custom Home Builders & Spec Home Construction currently show far fewer competing advertisers in our sample. If your product plausibly fits a lighter category, the same budget goes further there.
What we can and cannot tell you. We do not have your competitors' click-through rates, conversions or spend, and nobody outside OpenAI does — that data lives inside each advertiser's own account. What we observe is who advertises, how often they appear, what their creatives say, and which prompts surface them. That is enough to diagnose relevance and competition, which is where most underperformance originates.
The questions advertisers actually ask once the spend has started.
An impression only means your ad was eligible for a conversation, not that it fitted it. The usual cause is a context hint that is too broad, so the ad surfaces on conversations where your product is not the answer the person wanted. The second most common cause is category crowding: if several advertisers are competing for the same conversations, attention splits and the best-fitting card takes the click.
Publicly reported figures from the beta vary widely, and OpenAI does not publish an official benchmark, so treat any single number with caution. What is verifiable is that the ad sits inside an answer rather than on a results page, so relevance to the specific question matters more than it does in search. We publish observed competition levels by industry so you can at least tell whether your category is crowded or open.
Compare against the category rather than against a generic benchmark. We track Thousands advertisers and Tens of thousands creatives across Hundreds niches, which shows how many advertisers you are competing against and what they are saying. A weak result in a crowded category is a different problem from a weak result in an empty one.
No, and neither can anyone else outside OpenAI. Click-through and conversion data live inside each advertiser's own account. What is observable is which advertisers appear, how often they appear, what their creatives say, and which prompts surface them. That is enough to diagnose relevance and competition, which is where most underperformance comes from.
Before pausing, check whether the problem is the category or the campaign. If your niche is crowded, narrowing the context hint to the conversations you genuinely win is usually more effective than increasing budget. If your niche has few advertisers and you still get no clicks, the creative or the landing page is the more likely cause.
Across tens of thousands captured creatives the median title runs about 28 characters and the median description about 63, and 100% carry an image. These are observed patterns from live ads, not official limits. The consistent trait is that they read like a recommendation inside an answer rather than like a banner.
Questions about the data, a brand you expected to see, partnerships, or access to the intelligence layer. We read every message.