How to Write Context Hints That Convert

Tarun Kapoor Updated May 21, 2026 14 min read

A good context hint reads like a one-paragraph briefing for the auction. It names the audience, the intent, and the topic with enough specificity that the relevance scorer cannot mistake it for a generic competitor's. This guide is the writing field manual: the Audience–Intent–Topic framework, six templates organized by funnel stage, twelve do/don't pairs, and the pre-launch checklist that decides whether your ad group ships or gets rewritten one more time.

Why how you write the hint determines what you pay

OpenAI's auction is relevance-weighted. The relevance score is computed across four named inputs: context hints, landing page, ad title, ad copy. Of those four, the context hint is the only input that describes the conversation rather than the offer. That makes it the highest-leverage thing you write in the entire campaign. Strong hint = high relevance = lower CPC and higher delivery. Weak hint = commodity bidder = either you outspend or you don't show up.

The discipline this guide teaches is borrowed from two places: single-keyword ad group (SKAG) campaign architecture from paid search, and proposal-writing from B2B sales. The output is a paragraph; the rigor behind it is closer to writing a sales proposal than building a keyword list.

The Audience–Intent–Topic framework

Every strong hint names all three layers, in this order:

Audience — who is having the conversation

Role, seniority, company size or stage, industry, geography. Be concrete. "Founders" is weak. "Solo founders of pre-seed B2B SaaS companies in the US" is strong. The richness here is what lets the relevance scorer match conversations where the user has stated similar facts about themselves.

Intent — what they are trying to do right now

The action verb does the work: evaluating, switching, comparing, replacing, choosing, shortlisting. Each implies a funnel stage. "Wondering what a CRM is" is research. "Evaluating CRMs after outgrowing HubSpot Starter" is decision.

Topic — the category, sub-category, and constraint

Category is the noun ("CRM"). Sub-category sharpens it ("modern CRM with usage-based pricing"). Constraint disqualifies the generic competition ("HIPAA-eligible CRM with EHR integration"). The constraint is what makes your hint specific to your offer.

Templates by funnel stage

Template 1 — Research (top of funnel)

[Audience: who] who are researching [Topic: category] and want to understand
[specific question or sub-topic] before they evaluate vendors.

Template 2 — Evaluation (mid funnel)

[Audience] evaluating [Topic: category] alternatives because [trigger event
or pain]; they care about [specific criteria] and want [outcome].

Template 3 — Comparison (mid funnel)

[Audience] comparing [Vendor A] vs [Vendor B] for [use case] because they
need [specific capability] and want [outcome].

Template 4 — Switching (mid-to-bottom funnel)

[Audience] switching away from [Incumbent] because [specific failure] and
looking for [Topic: category] that supports [specific must-have].

Template 5 — Decision (bottom funnel)

[Audience] who have decided to buy [Topic: category] and need
[specific feature, integration, or constraint] to close.

Template 6 — Vertical-specific decision

[Audience in a specific vertical] facing [vertical-specific problem]
considering [Topic: category] that handles [vertical-specific constraint].

Twelve do/don't pairs

Use the buyer's vocabulary, not yours

The fastest way to weaken a hint is to write it in the language of your product spec sheet. Buyers don't ask for "AI-powered revenue intelligence platform with multi-touch attribution." They ask "how do I figure out which channel actually drives revenue when my pipeline is so noisy."

Three places to harvest real buyer vocabulary:

• Sales call transcripts. Pull the first 60 seconds of any qualified-discovery call. The buyer's first sentence describing their problem is almost always usable as a hint backbone.
• Reddit and niche communities. Search for your category on relevant subreddits — buyers ask their actual questions there, in their actual words.
• Customer support tickets. The phrase a customer used to describe what they were trying to do when they signed up is the highest-fidelity hint vocabulary in your company.

How many variants per ad group

Five to fifteen, written as paraphrases of the same audience-intent-topic combination. Variants exist to cover the natural language variation buyers use to describe the same situation. "Switching away from HubSpot," "leaving HubSpot," "outgrowing HubSpot Starter," "HubSpot got too expensive" — these are four hints for one ad group.

Below five and you're under-covering the language space. Above fifteen and you're either splitting the ad group too coarsely or you're padding with hints that don't add new coverage.

How to test and iterate

1. Launch with five hints per ad group. Watch impressions for 5–7 days.
2. Add five more hints if delivery is under-pacing. Most under-delivery is a coverage problem before it's a bid problem.
3. Pause underperformers once you have enough volume per hint to judge — typically 500+ impressions per hint. If you can't get to that volume, your splits are too granular.
4. Try a second ad group with a deliberately different framing of the same audience — switch from intent verb "evaluating" to "switching" and see whether one is materially cheaper.
5. Pull the hint variants that win and feed them into landing-page tests on the same audience. The relevance score is computed across both, and matching them compounds.

Pre-launch checklist

• ☐ Every hint names audience, intent, and topic explicitly.
• ☐ Every hint contains at least one constraint your product wins on.
• ☐ Hints are full sentences in the buyer's vocabulary, not your category taxonomy.
• ☐ Five to fifteen variants per ad group, covering the same audience-intent.
• ☐ Ad copy mirrors the hint's claim and constraint.
• ☐ Landing page mirrors the hint's audience and intent (a vertical or use-case page, not the homepage).
• ☐ One ad group per audience-intent combination, not per product line.
• ☐ Max bid set near your true-value-of-a-click, not deliberately shy.
• ☐ UTMs on the landing page, generated via your UTM Builder.
• ☐ Conversion measurement configured in Ads Manager Beta before launch.

Tarun Kapoor

Founder & CEO, Context Hints

Twelve years of media buying across GroupM, WPP, Ogilvy & Mather, and Neil Patel Digital — where he was directly responsible for $3.5M+ added to the bottom line of a $6.5M ARR agency.