How to Structure a ChatGPT Ads Account When There Are No Keywords
By Ansh Khandelwal 8 min read
In brief
Structure a ChatGPT Ads account around readable targeting decisions. Learn how campaigns, ad groups, context hints, and budgets fit together.
Last verified: 12 September 2026 | Version: 1.0 | Next scheduled review: 12 December 2026
Structure your account around the questions you intend to answer, not around your product catalogue, your team, or your website's information architecture. On a channel with no keywords, no search terms report and no reporting beneath the ad group's targeting inputs, the shape of the account is the only instrument you have.
That sounds abstract. It has a very concrete consequence, and it is the thing most first accounts get wrong.
Why structure carries more weight here than on Google
On Google Ads, structure is largely an organisational convenience. If you build it badly, the search terms report still tells you which queries brought which clicks, and negatives still let you correct course inside an existing group. You can recover from a bad structure without rebuilding.
ChatGPT Ads removes both of those recovery tools. Targeting lives in context hints, set at ad group level, described in natural language. There is no hint-level reporting, no search terms report, no exclusion layer. Ads Manager reports impressions, clicks, spend, CTR, average CPC, average CPM and conversions at campaign, ad group and ad level.
Read that list carefully, because the interesting boundary is not where reporting stops. It goes down to the ad. The interesting boundary is what each level can tell you about. Ad-level numbers vary your creative while holding targeting constant, so they answer questions about headlines and images. The targeting input, the hint, only varies between ad groups. So the ad group is the smallest unit at which a targeting decision produces a readable answer.
Everything below follows from that one sentence.
The rule: one question per ad group
An ad group should exist because you want to know something, and it should be built so that its numbers answer that thing and nothing else.
Put two buying situations in one ad group and its cost per lead has two possible causes and no way to separate them. The spend is the same either way. The information is not. We have written about this as the ceiling on hint count; at account level it becomes the ceiling on how much you can consolidate.
The practical test is not whether two themes are related. It is whether, if this ad group returned a bad number, you would know what to change. If the answer is "one of two or three things, and I could not tell you which", the group is doing two jobs.
What belongs at campaign level
Campaign level holds the settings you are not trying to learn from: objective, budget, dates, locations, custom audiences. These are constraints, not experiments.
That gives you a clean rule for when a second campaign is warranted. Split campaigns when a setting that lives at campaign level genuinely needs to differ. Do not split campaigns to organise themes, because themes are an ad group concern and splitting them into campaigns fragments your budget across pacing pools for no analytical gain.
Concretely, you need a second campaign when:
- The objective differs. A Reach campaign and a Conversions campaign are not comparable and cannot share a budget sensibly.
- The geography differs in a way you want to control spend on separately, rather than merely report on.
- One set of ad groups needs a budget ring-fenced from another, usually a proven set versus an exploratory one.
- A custom audience applies to some ad groups and not others, since audiences attach at campaign level.
You do not need a second campaign because you have two products, two personas or two seasons. Those are ad groups.
A worked skeleton for a B2B advertiser
Take a company selling compliance software to mid-market finance teams, at 100 USD a day, US only, running a first proper test.
Campaign 1: Core, Conversions objective, 75 USD a day, US, desktop web
- Ad group A: first-time audit failure. Hints describe a finance lead who has just had an audit sample come back with missing evidence. Creative names the trigger.
- Ad group B: outgrown spreadsheets. Hints describe a controller running quarterly evidence collection manually across a growing team.
- Ad group C: predecessor left. Hints describe someone who has inherited a compliance process nobody documented.
Three groups, three genuinely different triggers, three different headlines. Each one can be read and paused on its own.
Campaign 2: Exploration, Conversions objective, 25 USD a day, US, desktop web
- Ad group D: one situation the company believes exists but has never sold into.
The second campaign exists for a single reason: to stop an unproven theme borrowing budget from proven ones when pacing reallocates. It is the same objective and the same geography, so it would otherwise be an ad group. The budget ring-fence is the justification.
Note what is absent. No branded campaign, because there are no keywords to capture branded intent with and no query to route. No match-type tiering, for the same reason. No campaign per product line unless the products need different budgets. The account is smaller than its Google equivalent, and it should be.
The argument against the tidy account
Here is the honest reckoning, because the advice above has a cost.
Splitting for readability divides your budget, and a divided budget on a young channel can mean no ad group accumulates enough spend to produce a number worth reading. Three ad groups at 75 USD a day is roughly 25 USD each. At the three to five dollar CPC that agency write-ups report OpenAI guiding towards through 2026, that is somewhere around five to eight clicks per group per day. You will wait weeks for a conversion signal, and in the meantime you are optimising on noise.
So the structural rule is bounded by an arithmetic one. The number of ad groups you can support is set by budget divided by a daily spend floor that produces a readable result, not by how many themes you can articulate. Where those two rules conflict, the arithmetic wins and you consolidate, accepting that you have traded attribution for signal. That is a real trade, and it is not resolved by preferring cleaner structure. We work through the numbers separately in our piece on ad group count.
What we cannot tell you
- Whether campaign count affects delivery or pacing behaviour. Undocumented. OpenAI describes budget pacing at campaign level and does not describe interaction between campaigns.
- Whether ad groups compete with each other in the auction. Undocumented. On other platforms this is a known effect with published guidance. Here there is none.
- What a structural change costs in learning-phase terms. OpenAI does not document a learning phase for ChatGPT Ads campaigns, so the standard advice about avoiding edits has no published basis on this platform.
- Which hint inside an ad group drove a result. No hint-level reporting exists.
- Whether any of this structure outperforms a flat one-campaign account. No published test exists, and InPromptAds has no first-party account data to offer.
Quick answers
How should I structure a ChatGPT Ads account? Around what you want to measure. One buying situation per ad group, since the ad group is the smallest unit at which a targeting decision produces a readable result. Campaigns hold objective, budget, dates, location and custom audiences.
How many campaigns do I need? As few as your campaign-level settings allow. Split only when objective, geography, budget ring-fencing or custom audience genuinely differ, never to organise themes.
Should I build a branded campaign like I do on Google? No. There are no keywords and no query to capture, so there is nothing for a branded campaign to intercept.
Can I just put everything in one ad group? You can, and if your budget is small you may have to. Understand the cost: the group's numbers will have several possible causes and no way to separate them.
Does ad-level reporting solve the attribution problem? Only for creative. Ads vary headline, description and image while targeting stays constant, so ad-level numbers answer creative questions, not targeting ones.
What replaces account structure as the optimisation lever? Ad group structure, creative variants and the landing page. The Google loop of reading search terms and adding negatives does not exist here.
Sources
| Claim | Source | Tier |
|---|---|---|
| Campaign holds objective, budget, dates, locations and custom audiences; ad group holds bid, destination URL and context hints; ad holds creative | OpenAI Help Center, Create Campaigns and Create Ad Groups for ChatGPT Ads, 2026 | Confirmed, primary |
| Objectives are Reach (CPM), Clicks (CPC) and Conversions (oCPC) | OpenAI Help Center, Create Campaigns for ChatGPT Ads, 2026 | Confirmed, primary |
| Reporting covers impressions, clicks, spend, CTR, avg CPC, avg CPM and conversions at campaign, ad group and ad level | OpenAI Ads Manager documentation, 2026 | Confirmed, primary |
| No hint-level reporting, no search terms report, no exclusion layer | OpenAI documentation, by absence | Absent |
| Minimum daily budget of 25 USD | OpenAI Help Center, Create Campaigns for ChatGPT Ads, 2026 | Confirmed, primary |
| Three to five dollar CPC starting bid guidance | Multiple agency write-ups citing OpenAI guidance, 2026 | Reported |
| One question per ad group as the structural rule | Newtation reasoning from reporting granularity | Inference, ours |
Related reading
- How Many Ad Groups Should One ChatGPT Campaign Have?
- How to Write a ChatGPT Ads Context Hint
- Context Hints Are Not Keywords: The Translation Table
- How to Measure ChatGPT Ads, and What You Cannot Measure Yet
- Daily Budget or Total Campaign Budget on ChatGPT Ads
Changelog
12 September 2026, v1.0. First publication. Establishes the ad group as the smallest readable unit of a targeting decision, and sets the arithmetic limit on splitting against it.
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.
