ChatGPT Ads Context Hints: The Complete Guide
Context hints are the closest thing ChatGPT Ads has to keyword targeting — and the most misunderstood. This guide explains what they actually are, how to write them with a repeatable formula, real examples by vertical, and the targeting traps that waste budget.
Primary sources: OpenAI: target your ads with context hints ↗ · OpenAI campaign targeting docs ↗
What Context Hints Are
A context hint is a short, plain-language description of the conversations your ad is relevant to, written at the ad group level. When a user talks to ChatGPT about something that matches your hint, your ad becomes eligible to appear.
OpenAI evaluates hints against the full conversational context — not just isolated words. That is the entire point: ChatGPT understands intent better than a keyword match ever could. A hint like “people comparing tax software for self-employed income” reaches the same intent a pile of keywords would, but without match-type guessing.
Definition in one line
“A context hint tells OpenAI which conversational situations your ad belongs in.”
Why They Are Not Keywords
- No match types. Exact, phrase, broad — none exist. Hints are intent signals, not match syntax.
- No literal matching. OpenAI matches meaning, so synonyms and paraphrases are on the table.
- No negative hints (yet). You exclude by writing precise positive hints rather than adding exclusions.
- Conversation-level relevance. A hint fires based on the whole thread, not a single search query.
- Quality-weighted. Delivery still depends on the auction, budget, and page quality.
The 5-Element Hint Formula
The highest-performing hints read like a buyer's thought, not a keyword pile. Use this repeatable structure:
- 1
Who
The person: "small business owners", "first-time home buyers", "parents planning a family trip".
- 2
Task or intent
What they are trying to do: "judging between two CRM platforms", "estimating a kitchen remodel".
- 3
Stakes or context
Why it matters now: "before tax season", "during an overseas move", "comparing quotes".
- 4
Objections to speak to
The hesitation your ad can preempt: "worrying about hidden fees", "nervous about switching providers".
- 5
Outcome they want
The destination: "want a same-day estimate", "need a 30-day free trial first".
You do not need all five in every hint — three strong ones beat five weak ones. Start with Who + Intent + Outcome and layer in stakes as you learn which conversations convert.
Examples by Vertical
Home Services
"Homeowners getting quotes for roofing or HVAC replacement and wanting to compare contractors before committing."
Car Rental
"Travelers comparing rental cars for a family trip, weighing size, insurance, and total cost with pickup location."
B2B SaaS
"Marketing leads at small agencies comparing all-in-one tools versus point solutions and asking what a per-seat plan covers."
Legal
"People who had an accident and are researching whether to talk to a personal injury lawyer and what a realistic settlement looks like."
International Real Estate
"Investors and expats researching buying property abroad, asking about financing rules and or taxes for foreign buyers."
Notice the pattern: each hint describes a person in a moment of comparison — not a product name repeated three times. That is what makes the AI match it to real conversations.
Targeting Mistakes to Avoid
- Writing keyword lists instead of descriptions. "roofing, roof repair, roof replacement" is a list, not a hint.
- One giant hint. Overloading dilutes relevance; keep each hint focused on one conversational job.
- Forgetting the landing page. A hint that targets research-stage intent must send to a page that answers that research, or conversions stall.
- Ignoring the reporting window. Judging hint performance within minutes (instead of 24–48 hours) produces false conclusions.
- No A/B. Shipping a single hint and never refining leaves cheap clicks on the table.
How to Test and Refine
- 1
Create two or three ad groups, each with a single focused hint, and give every group its own themed landing page.
- 2
Run a small, conversion-optimized test budget for at least a week so the algorithm learns from attributed events.
- 3
Compare by weighted metrics: CPA and qualified-lead quality, not raw clicks or CTR.
- 4
Keep winners, split broad hints in two, and retire hints that attract irrelevant conversations.
- 5
Fold winning hint language back into ad copy — matching the conversation increases click quality further.
Refinement is only trustworthy when conversions are credited correctly. If the pixel and CAPI are not deduplicated, you will optimize against phantom numbers. Setup clean tracking before you start A/B hint — order matters.
Frequently Asked Questions
What are context hints in ChatGPT Ads?
Context hints are plain-language descriptions you write at the ad group level that tell OpenAI which conversational situations your ad is relevant to. They act as intent signals, not exact-match keywords: OpenAI evaluates the whole conversation and matches ads to contextually relevant moments.
Are context hints like Google keywords?
Partially. Like keywords, they scope who sees your ad and drive relevance. Unlike Google, they are not matched literally. OpenAI uses them as signals to understand conversational context, so they work best when written like real buyer language rather than keyword lists.
Where do I add context hints?
In the OpenAI Ads dashboard, inside your campaign at the ad group level. An ad group contains your ads plus one or more context hints. You can test multiple hints per ad group and refine them based on performance.
Can context hints be negative target?
Not as such. OpenAI Ads does not support negative keywords the way Google does. Context hints scope in relevant conversations, but for excluding irrelevant topics you rely on hint phrasing, campaign-level exclusions where available, and monitoring where impressions actually land.
Do hints guarantee my ad shows?
No. Hints determine relevance and eligibility, but whether your ad appears depends on the auction (bid, budget, quality) and on the conversation actually being a good contextual match. Hints are the targeting layer, not a guarantee of delivery.
Target smarter, measure honestly
Context hints find the right conversations; only clean tracking shows you which ones convert. Get both with the Ridoway Pixel + CAPI pair.