The operating system behind your ChatGPT Ads program.

1. Campaign strategy

We don't start with "run ChatGPT Ads." We start with one question: which customer decision do you want to be present for?

A furnace replacement decision is different from an emergency AC repair. A roof leak is different from a planned remodel. Each one requires different targeting, different creative, and a different landing page.

We build each campaign around:

• One service line with gross profit that supports paid acquisition
• One geography where you have capacity and want work
• One buying situation — the specific homeowner decision where your business is genuinely useful
• A clear economic threshold — the maximum you should pay to acquire a customer, calculated from your actual margins

This is not a media plan. It's a commercial test with a defined success criteria before a single dollar is spent.

2. Context hints and targeting

ChatGPT Ads don't use keywords. They use context hints — natural-language descriptions of when your business is relevant to a conversation.

Most advertisers treat context hints like a keyword dump: "HVAC, plumbing, furnace, repair." That gives the platform almost nothing to work with.

We write context hints that describe the actual buying situation:

"Homeowners in the Denver metro area with a furnace over 15 years old who are comparing another repair with full-system replacement."

That single sentence tells the ad system more than a list of 20 keywords. We build distinct ad groups around distinct decisions — not one giant group for "HVAC."

Geographic targeting uses OpenAI's location controls, not context hints. We enforce that separation because hints are relevance signals, not delivery controls.

3. Ad creative

The ad appears inside a conversation. It has to feel useful, not intrusive.

OpenAI gives you a title (max 50 characters), body (max 100 characters), image, and landing page URL. Every word has to earn its space.

We build 5+ distinct variations per ad group — not rewrites of the same headline, but genuinely different angles:

• Decision support ("Repair or replace your furnace?")
• Trust and credentials ("Licensed local HVAC company")
• Speed and availability ("Estimates this week")
• Financial clarity ("Financing options available")
• Specificity ("Serving the Denver metro since 2008")

OpenAI's system tests variations and allocates delivery to the creative that earns engagement in context. More distinct coverage gives the platform more to work with.

We do not use generic stock imagery. Every ad image should be a real job photo, a real truck, or a clean service-specific visual that makes the business immediately legible.

4. Landing pages

We do not send ChatGPT traffic to your homepage.

Your homepage serves ten audiences. A ChatGPT Ad serves one: a homeowner who just described a specific problem in a conversation and clicked a relevant ad.

We build a dedicated landing page per service that continues the conversation:

• Acknowledges the specific problem
• Establishes the company's relevance (service area, licensing, experience)
• Provides a clear next step (call or short form)
• Loads fast, works on mobile, requires no navigation

OpenAI includes landing-page relevance among the signals used for ad delivery. A focused destination is better for the customer, better for the ad system, and better for conversion rate.

No website rebuild required. The landing page runs alongside your existing site.

5. Measurement and attribution

This is where most agencies lose the thread. They report clicks and call it attribution.

We built our measurement stack directly on OpenAI's Measurement Pixel and Conversions API. We don't rely on a third-party wrapper or a spreadsheet.

What we instrument

• OpenAI Pixel — installed on every page, fires page_viewed and lead_created events with proper event IDs
• Conversions API — server-side event confirmation, deduplicated with the pixel, includes hashed email for conversion matching
• oppref preservation — OpenAI's click identifier is captured, stored in a first-party cookie, and passed through every form submission to the server
• UTM chain — source, medium, campaign, ad ID, and ad group ID flow from the click through to the lead record and the booking
• Call tracking — campaign-specific phone numbers that route to your office while preserving source attribution

What we report

The weekly scorecard tracks the full chain:

The owner sees the business result first. The diagnostic metrics are below it. Read our full measurement methodology.

6. Optimization

We review every campaign weekly. Not with a dashboard glance — with a diagnostic framework:

• Low delivery? → Bid, inventory, geography, creative coverage, ad review status
• Clicks but no leads? → Page relevance, ad-to-page mismatch, tracking failure, slow load, weak CTA
• Leads but no bookings? → Out-of-area traffic, wrong service, intake failure, slow response
• Bookings but no sales? → Lead quality, pricing, appointment quality, sales execution

The ad platform is not always the problem. Sometimes the bottleneck is speed-to-lead, phone answer rate, or estimate conversion. We diagnose the full funnel, not just the media metrics.

When performance justifies it, we recommend expanding: additional service lines, broader geography, or moving from click optimization to conversion optimization using OpenAI's supported standard events.

When performance doesn't justify it, we say so. We'd rather stop a campaign that isn't working than continue billing for one.

7. Platform infrastructure

We don't just use the OpenAI Ads Manager UI. We built programmatic integrations against the OpenAI Advertiser API to manage campaigns, ad groups, ads, targeting, budgets, creative, and reporting at scale.

This gives us:

• Faster campaign setup and iteration
• Structured audit logs for every change
• Automated performance data collection
• The ability to manage multiple client accounts with strict data isolation
• Direct access to platform capabilities as OpenAI ships them

We monitor the OpenAI platform roadmap and update campaigns as new features become available — visual formats, expanded measurement, conversion optimization improvements.

8. What we don't do

Clarity about scope matters as much as capability.

• We don't run your Google Ads. Keep your existing agency. We handle ChatGPT.
• We don't rebuild your website. The landing page is separate.
• We don't promise cheap leads. We promise controlled, measurable tests with honest reporting.
• We don't invent performance data. If a campaign isn't producing, we'll tell you.
• We don't lock you in. No long-term contracts. The work has to justify itself.
• We don't accept every company. If the economics or capacity aren't right, we'll say so before taking your money.

---

Start with a 15-minute call

We'll look at your service lines, market, capacity, and economics together. If there's a worthwhile pilot, we'll tell you exactly what it looks like. If there isn't, we'll tell you that too.