Can You Exclude Conversations on ChatGPT Ads?
By Ansh Khandelwal 6 min read
In brief
ChatGPT Ads has no negative keywords, so exclusion moves to the creative layer: the ad itself has to disqualify the reader you do not want.
Last verified: 12 September 2026 | Version: 1.0 | Next scheduled review: 12 October 2026
No. ChatGPT Ads has no negative keywords, no exclusion list and no way to block a topic, a competitor comparison or a category of conversation. OpenAI's ad group documentation describes context hints only as descriptions of relevant situations and provides no inverse.
The useful question is what happens to the job negatives were doing.
What negatives were actually for
On Google, a negative keyword list does three separate jobs, and they get bundled together because one control handles all of them.
It stops you paying for clearly wrong intent, the job hunters and the students. It stops you paying twice for the same person across two campaigns. And it enforces your own account structure, keeping the branded terms inside the branded campaign.
On ChatGPT Ads the second and third jobs mostly evaporate, because there is no query to route and no overlap to arbitrate. The first job does not evaporate. It moves.
Where exclusion moved to
Three layers are available, and they work in this order.
The hint. A constraint inside the hint is the closest thing to a preventive exclusion. "A first-time buyer in England whose offer has just been accepted" carries an implicit exclusion of remortgaging, commercial property and Scottish conveyancing, without an exclusion field existing. This is cheap because it happens before the impression.
The ad group. Separating a low-fit theme into its own ad group does not exclude anything, but it makes the theme pausable. If a second group is spending at a bad cost per lead, you can switch it off, which is exclusion arrived at after the fact and paid for once.
The creative. This is the layer nobody writes about, and on this channel it is the important one. With no exclusion layer and no search terms report, your ad copy is the only negative keyword you have.
Writing an ad that repels the wrong reader
Standard paid-media instinct says maximise appeal, get the click, sort out qualification on the landing page. That instinct is calibrated for channels where you can see and exclude what you matched. Here you can do neither, so the filtering has to happen before the click, in the only asset the user reads.
A disqualifying line is a specific, checkable condition placed in the headline or description that the wrong reader will recognise as not applying to them.
- Price floor: "From £2,400 a year" removes the free-tool hunter.
- Size qualifier: "Built for teams of 20 to 200" removes the solo founder and the enterprise evaluator.
- Geography: "UK limited companies only" removes half the match surface.
- Stage: "For teams already running payroll" removes the pre-revenue reader.
The arithmetic is what makes this worth doing rather than merely tidy. Agency write-ups through 2026 report OpenAI guiding CPC bids into the three to five dollar range. At four dollars, a hundred clicks is four hundred dollars. If thirty of those clicks were never qualified and a price line in the description removes twenty of them, you have recovered eighty dollars and, more importantly, you have stopped feeding unqualified traffic into a measurement system too young to tell you it was unqualified. Treat that as illustration of the mechanism rather than a measured result: the CPC figure is reported, not published, and the thirty percent is an assumption.
The reckoning
Disqualifying copy lowers click-through rate. That is the point, and it is also a cost.
The auction is relevance-weighted, and click-through is one of the few relevance signals a new account can generate. Suppress CTR deliberately and you may be paying more per click on the clicks you do want, or serving less often. Nobody outside OpenAI can currently size that trade, because relevance is never surfaced as a number and no advertiser has published a controlled test of it.
So the honest position is a conditional one. Disqualifying copy is clearly right when your cost per unqualified lead is high and your sales team is the thing being wasted, which describes most B2B. It is not obviously right when volume is scarce and you are still trying to prove the channel serves at all. In the first two weeks, bias toward reach. Once there is spend to read, add the disqualifier and watch what happens to cost per qualified lead rather than to CTR.
What we cannot tell you
- Whether OpenAI will ship an exclusion layer. Not on any published roadmap.
- What a lower CTR costs you in the auction. Relevance weighting is documented, its magnitude is not.
- Whether constraints inside a hint genuinely narrow matching. Not measurable without hint-level reporting.
- How much unqualified traffic a typical account receives. No advertiser has published it and there is no search terms report to measure it with.
Quick answers
Can I add negative keywords to ChatGPT Ads? No. There is no exclusion layer of any kind.
How do I stop my ad showing in irrelevant conversations? Put constraints inside the hint, split low-fit themes into separate pausable ad groups, and put a disqualifying condition in the ad copy.
Can I exclude competitor comparison conversations? No. You can only decline to describe them in your hints and write creative that a comparison-stage reader is less likely to click.
Does splitting themes into separate ad groups exclude anything? No, it makes them pausable, which is exclusion after you have already paid once.
Will a disqualifying line hurt my performance? It lowers CTR by design. Judge it on cost per qualified lead, not on CTR, and hold it back until there is enough spend to read.
Sources
| Claim | Source | Tier |
|---|---|---|
| Context hints describe relevant situations; no inverse or exclusion field is documented | OpenAI Help Center, Create Ad Groups for ChatGPT Ads, August 2026 | Confirmed, primary |
| No negative keyword or exclusion capability exists | OpenAI Ads Manager documentation, by absence | Absent |
| Hints are not negative keywords; separate low-fit themes instead | Lapis, context hints best practices, 2026 | Reported, third party |
| Auction is relevance-weighted | OpenAI documentation, 2026 | Confirmed, primary |
| Three to five dollar CPC starting bid guidance | Multiple agency write-ups citing OpenAI guidance, 2026 | Reported |
Related reading
- How to Write a ChatGPT Ads Context Hint
- Context Hints Are Not Keywords: The Translation Table
- How the ChatGPT Ads Auction Works
- Who Should Not Buy ChatGPT Ads Yet
Changelog
12 September 2026, v1.0. First publication. Establishes creative-layer exclusion and the disqualifying line, with the CTR and relevance trade stated as unsized.
Field kit
Tools
Site
OpenAI (primary)
Ansh works across GEO strategy, B2B research, and execution. At InPromptAds, he translates new AI advertising products into clear operating advice, tests, and measurement questions for marketing teams.
