ChatGPT Ads buyer questions: how to find, group and write for them
How to plan ChatGPT Ads around buyer questions: where to find them, how to group them into ad groups, how to write context hints and cards for each type, and how to prioritise.
In ChatGPT Ads, the buyer question replaces the keyword. Collect the real questions your buyers ask from sales calls, support and site search, group them by intent into a handful of ad groups, describe when your offer fits in plain language, and write a card that answers the next thing the buyer will want to know.
Why questions, not keywords
On Google, a buyer types "AP automation software". In ChatGPT, the same buyer writes: "We're a 200-person manufacturer on NetSuite, invoices take two weeks to approve, what tools should we look at?" That one prompt carries the company size, the stack, the problem and the stage.
OpenAI matches ads to the conversation, the ad, its landing page and the context you provide. It doesn't sell keywords in the classic sense, and it doesn't let you target companies or job titles. That makes your understanding of buyer questions the main lever you control. Better questions lead to better context, better cards and better landing pages.
Where to find real buyer questions
Don't invent questions in a workshop. Collect them.
| Source | What you get | How to use it |
|---|---|---|
| Sales call notes and recordings | The exact words buyers use, objections, competitor names | Strongest source for alternatives and validation questions |
| Support tickets and onboarding | Integration and "how do I" questions | Integration and problem clusters |
| Site search and chat logs | What visitors couldn't find | Gaps in your pages, new clusters |
| Lost-deal notes | Why buyers chose someone else | Comparison angles |
| Search Console queries | Long, question-shaped searches | Validation that a question has volume |
| Testing AI answers | What ChatGPT and other assistants say today | Where competitors are named and you aren't |
Aim for fifty to a hundred real questions before you group them. Write each one the way a buyer would type it, including context like company size, industry or tools.
The anatomy of a B2B question
Most B2B prompts combine four parts:
- Role or team: "our finance team", "as the IT lead".
- Company context: size, industry, region, current tools.
- Problem or goal: what hurts, or what they need to achieve.
- Ask: a shortlist, a comparison, a how-to, a yes-or-no check.
When you group questions, the ask tells you the stage. The context tells you the segment. The role tells you how to write the card.
Group questions into ad groups
Group by ask first, then by segment if needed. Five clusters cover most B2B categories:
| Cluster | Typical ask | Stage | Card angle |
|---|---|---|---|
| Shortlist | "What are the best tools for X for a company like ours?" | Solution | A clear fit statement: who it's for and the outcome |
| Alternatives | "Alternatives to Vendor A?" "Vendor A vs Vendor B?" | Vendor | The specific difference that matters to switchers |
| Integration | "Does any X tool work with our ERP/CRM?" | Solution to vendor | The integration, how deep it goes, how fast it's live |
| Problem | "How do we fix Y?" | Problem | A useful resource or a quantified outcome |
| Validation | "Is X compliant with Z?" "How long does X take to implement?" | Late vendor | Proof: certification, timeline, a direct answer |
Three to five ad groups per campaign is enough to start. Each ad group gets its own landing page where possible.
Prioritise: gap, defend, expand
Not every question deserves budget. Check what AI assistants say today for each question and sort:
- Gap: the answer names competitors and not you. Highest priority. The ad is your only presence in that moment.
- Defend: the answer names you, alongside others. An ad keeps you visible as competitors start bidding.
- Expand: problem-stage questions where nobody is named yet. Good for reach, lower intent.
Track this weekly. AI answers change, and a question can move from gap to defend once your AI visibility work lands.
Write context hints in plain language
Context hints tell the ad system when your offer is relevant. Write them like a short brief, not a keyword list. A useful hint answers three things:
- What it does: "Automates invoice approval and supplier payments."
- Who it suits: "Finance teams at manufacturers and distributors with 100 to 2,000 employees."
- When it helps: "When invoices sit in email for days, or month-end close runs late because of AP."
Avoid superlatives and claims you can't back up. "The best AP tool" helps nobody, including the matching system.
Write the card for the next question
The buyer has just read an answer. The card should address the question they'll ask next.
- After a shortlist answer, the next question is "which one fits us?" Lead with fit: "AP automation for NetSuite manufacturers."
- After a comparison, it's "what's the real difference?" Lead with the difference: "Approval in Teams, no extra logins."
- After an integration question, it's "how hard is it?" Lead with effort: "Two-way NetSuite sync, live in about a week."
- After a validation question, it's "can I trust that?" Lead with proof and link to the evidence page.
These examples are illustrative; use your own facts. Keep titles and text within the current Ads Manager limits, and make sure every claim appears on the landing page.
Match the landing page to the cluster
A question-shaped click deserves a question-shaped page. For each cluster:
- Shortlist: a use-case page for that segment.
- Alternatives: an honest comparison page.
- Integration: the integration page, with setup steps.
- Problem: a guide that solves the problem, with a soft next step.
- Validation: the security, compliance or implementation page.
The same pages help organic AI visibility, because AI engines cite pages that answer questions directly.
Refresh the question set
Buyer questions drift. New competitors appear, regulations change, AI answers shift. A monthly routine:
- Add new questions from sales and support.
- Re-check AI answers and update gap, defend and expand labels.
- Retire clusters that bring clicks from companies outside your ICP.
- Promote problem questions that turn out to bring target accounts.
How IntentBid helps
IntentBid tracks your buyer questions across AI engines every day and labels the ones where AI names competitors and not you. From those it builds ChatGPT Ads ad groups with plain-language context hints, checks the card against the limits, and creates the campaign paused in your own account, with 0% fee on media spend. The pixel shows which company clicked on which question. The GEO agent can publish page fixes to your CMS so the same questions start naming you organically.
FAQ
How many buyer questions do I need to start?
Fifty to a hundred real questions is enough to find three to five clusters. Quality matters more than quantity.
Are context hints the same as keywords?
No. They describe when your offer is relevant, in plain language. They guide matching but don't guarantee your ad shows for any specific prompt.
Should I advertise on competitor-name questions?
Alternatives and comparison questions are often high-intent. Keep cards factual, avoid claims about competitors you can't support, and follow OpenAI's ad policies.
Can I see which prompts triggered my ad?
No. OpenAI doesn't share conversations with advertisers. Reporting is aggregated, for example at campaign level, which is why question-based ad groups matter.
How do buyer questions relate to AEO and GEO?
They're the same questions. Ads cover the paid placement below the answer; AEO and GEO work on being named in the answer itself.
Turn your buyer questions into ChatGPT Ads campaigns. Start a 14-day trial, no card needed.