Write Context Hints for the Role in the Decision, Not for the Product
By Ansh Khandelwal 8 min read
In brief
Structure ChatGPT Ads context hints around evaluators, users and budget holders without mixing different B2B buying-stage questions.
Last verified: 12 September 2026 | Version: 1.0 | Next scheduled review: 12 December 2026
Three people will touch a 40,000 dollar software decision. They will use ChatGPT during it. They will not type anything resembling each other.
The end user types about the problem. The evaluator types about the comparison. The person who signs types about the risk. One set of context hints written about your product describes none of those three conversations, because none of the three is having a conversation about your product. They are having a conversation about their own part of the decision.
Why product-shaped hints fail on long cycles
A hint written as "revenue attribution software for B2B marketing teams" is a description of a category, and category conversations happen mostly at one point in the cycle: the shortlist. Everything before and after it is shaped by role.
The mismatch is worse on long cycles than on short ones for a structural reason. In a two-week consumer purchase, the buyer and the user and the payer are one person, so one hint covers all three roles by accident. Stretch the same purchase across five months and three job titles, and the single hint now covers roughly one third of the conversations that decide the outcome, and you have no way of knowing which third, because ChatGPT Ads reports nothing at hint level.
The three roles, and what each one actually types
| Role | What they are trying to do | What the conversation sounds like | What the ad has to be |
|---|---|---|---|
| End user | Solve a task that is currently painful | Problem framing. "How do I stop rebuilding this report every Monday" | A better way to do the task, named concretely |
| Evaluator | Build and defend a shortlist | Comparison and criteria. "What should I be asking attribution vendors" | A credible option with a checkable differentiator |
| Signer | Not be wrong | Risk, cost, and consequence. "What usually goes wrong with these implementations" | Proof of low downside: contract terms, implementation length, references |
The evaluator is the only one of the three whose conversation contains your category name. That is why category-shaped hints feel like they work: they do work, for one role, and that role is the middle of the funnel.
Hints by role, one worked purchase
Take a mid-market company buying a revenue attribution tool. Same purchase, three ad groups.
End user group. The person who feels it.
- A demand generation manager at a 200 person B2B company who rebuilds a pipeline source report by hand in a spreadsheet every week, and cannot reconcile it with what the CRM says
- A marketer being asked which channel produced last quarter's closed revenue, who can answer for first touch but not for anything after it
Evaluator group. The person building the shortlist.
- A marketing operations lead drawing up evaluation criteria for attribution vendors, who needs to know which ones read Salesforce opportunity history rather than only web sessions
- Someone comparing attribution platforms after a failed implementation, who wants to know what the previous one could not do
Signer group. The person who says yes to the invoice.
- A VP of marketing deciding whether to approve an attribution platform, who has been through one analytics implementation that took nine months and wants to know what makes this one different
- A finance-adjacent leader asking what an attribution tool costs in total over a year once implementation and headcount are included
Read the three groups back and notice that the vocabulary barely overlaps. "Spreadsheet" and "Monday" and "reconcile" appear in the first. "Criteria" and "Salesforce opportunity history" in the second. "Nine months" and "total over a year" in the third. That non-overlap is the point. If your three groups share most of their nouns, you have written one group three times.
Each role needs its own ad, not just its own hint
The reason to split by role is not tidiness. It is that the ad has to change.
The end user ad sells the removal of a weekly task, and it can be concrete and small: "Stop rebuilding the source report". The evaluator ad has to be checkable, because the evaluator is going to verify it: "Reads Salesforce opportunity history, not just sessions". The signer ad sells the absence of a disaster: "Live in 30 days or you do not pay".
Those three lines cannot share an ad group, because an ad group has one default destination and one set of creative competing inside it. Splitting by role is how the creative gets to match the conversation you described. This is the same legible targeting argument that governs hint count: the ad group is the smallest unit you can read, so put one describable audience inside it.
The honest reckoning: your signer probably never sees the ad
Here is the part that undercuts the model above, and it is not small.
ChatGPT Ads serve to Free and Go plan users. They do not serve to paid plan users. Reported through 2026, and it has been stable.
Now think about which of the three roles is most likely to be on a paid plan. The VP who signs a 40,000 dollar contract works at a company that very likely pays for their AI tools, and is personally the least price sensitive of the three about a 20 dollar subscription. The evaluator is somewhere in the middle. The end user, the individual contributor rebuilding a report on a Monday, is the most likely of the three to be on a free tier, possibly on a personal account.
So the tidy three-persona model inverts the usual B2B priority. The role you would normally spend most on reaching is statistically the hardest to reach here, and the role most media plans treat as an influencer is the one the channel actually delivers.
Two consequences follow, and they are practical rather than rhetorical.
First, weight your budget toward the end user and evaluator groups until you have evidence otherwise. Not because signer hints are wrong, but because they are competing for the thinnest slice of the available audience.
Second, write the end user ad as if it will be forwarded. On this channel the realistic path to the signer is not an impression. It is an internal Slack message from the person who saw the ad. That changes the landing page more than it changes the ad: the page has to survive being read by someone who did not see the conversation that produced it.
We cannot size any of this. Nobody publishes the plan mix of ChatGPT users by seniority, and OpenAI reports nothing about who saw an impression. The direction of the bias is defensible. The magnitude is guesswork, including ours.
What we cannot tell you
- The plan mix of ChatGPT users by job seniority. Never published by OpenAI, and there is no third-party panel that measures it credibly.
- Which role a given impression reached. No demographic, firmographic or conversation reporting exists at any level of ChatGPT Ads.
- Whether role-split ad groups outperform product-split ones. No published test, and InPromptAds has no first-party campaign data to offer.
- How often a B2B purchase involves ChatGPT at all. Self-reported survey data exists in trade press; nothing observational.
- What share of your target audience has opted out of ads. Opt-out exists, take-up is unpublished.
Quick answers
Should I build one ad group per persona or one per product? Per role in the decision, when the purchase involves more than one person. The end user, the evaluator and the signer ask different questions, need different creative, and cannot share an ad group and still produce readable numbers.
Can I target job titles on ChatGPT Ads? No. There is no firmographic or professional targeting. Role targeting happens entirely through how the hint describes the situation, and it is approximate rather than deterministic.
Who actually sees ChatGPT ads in a B2B buying group? Ads serve to Free and Go plan users only, reported through 2026. That biases delivery toward individual contributors and away from senior budget holders, who are more likely on paid plans. The size of the skew is unpublished.
How do I reach the person who signs? Indirectly. Write the end user and evaluator ads so the landing page holds up when it gets forwarded internally, and treat direct signer reach as the least reliable of the three.
Does a longer sales cycle change how I write hints? It changes how many you need. A long cycle splits the buyer into separate people having separate conversations, so it needs more ad groups covering narrower situations, not broader hints covering all of them.
Should the signer ad group get equal budget? Not at the start. It is competing for the thinnest slice of the audience. Fund it only after the other two groups have shown the channel serves for your category.
Sources
| Claim | Source | Tier |
|---|---|---|
| Context hints are ad-group-level natural-language descriptions of relevant situations | OpenAI Help Center, Create Ad Groups for ChatGPT Ads, August 2026 | Confirmed, primary |
| Reporting exists at campaign, ad group and ad level only, with no hint-level or audience-level breakdown | OpenAI Ads Manager documentation, 2026 | Confirmed, primary |
| Ads serve to Free and Go plan users, not paid plans | Multiple trade write-ups of OpenAI's ads rollout, 2026 | Reported |
| No firmographic, demographic or job-title targeting exists | OpenAI documentation, by absence | Absent |
| Users can opt out of ads; take-up not published | OpenAI ads policy, 2026 | Confirmed, primary for the control; Absent for the rate |
| Role-split ad groups and the forwarded-landing-page argument | InPromptAds reasoning from the plan-tier constraint | Inference, ours |
Related reading
- Build Context Hints From Sales Calls, Not From Brainstorms
- How Many Context Hints Should One Ad Group Have?
- How to Write a ChatGPT Ads Context Hint
- Who Should Not Buy ChatGPT Ads Yet
- Context Hints Generator
Changelog
12 September 2026, v1.0. First publication. Establishes role-in-the-decision hint structure for multi-stakeholder B2B purchases, and records the Free and Go plan constraint as an inversion of normal B2B seniority weighting.
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.
