Your first week of ChatGPT ads should establish whether the campaign is running correctly and producing useful evidence. It may take longer to judge acquisition cost, lead quality or repeat purchases. Seven days is a review appointment, not a guaranteed verdict.
Use this manual checklist to separate operational problems from performance questions. Fix broken tracking or a misleading offer promptly; avoid changing several working parts just because the first numbers are small.
Before launch: write the decision you want to make
Choose one primary outcome and define it precisely. For example, “a quote request from a business in our service area” is more useful than “more engagement”. Record your spending cap and the cost per outcome you can afford.
Then make a dated copy of the campaign: offer, hints, ads, destination pages and settings. Your notes should make it possible to explain what was running when a lead arrived.
Check the landing page on mobile. Complete the form or checkout, confirm the success message and verify that your team receives the result. Inspect account and ad review messages, billing setup, dates and location settings before activation.
Day 1: verify measurement end to end
OpenAI supports conversion measurement through the Pixel, Conversions API or both. Check that your data source, campaign event and attribution window are configured. Preserve the oppref click reference through redirects. OpenAI’s conversion measurement guide
Add consistent campaign tags to destination links for your own analytics. One possible convention is utm_source=chatgpt&utm_medium=paid&utm_campaign=office_cleaning_test. Use stable names that your team understands, and confirm the final destination retains the parameters.
For a complete naming example and website checks, use the UTM and GA4 measurement guide.
Test that the primary event happens only after the intended action succeeds. Opening a contact form is not the same as submitting it. Refreshing a confirmation page should not create another customer in your business records.
If browser and server tracking report the same conversion, use the same event ID for deduplication. OpenAI’s Pixel documentation
Days 2–3: investigate delivery before creative performance
Use the messages in your actual ad account to guide diagnosis. Do not assume that an active label means every prerequisite is satisfied.
| What you observe | What to check next |
|---|---|
| No impressions | Review messages, account eligibility, dates, limits, geography and bid guidance |
| Impressions but few clicks | Whether the ad clearly offers something relevant; whether volume is large enough to compare |
| Clicks but no measured visits | Destination errors, redirects and analytics configuration |
| Visits but no recorded outcomes | Event setup, form or checkout operation, and visitor suitability |
Document the issue before changing a setting. Zero impressions do not identify one particular cause. A low click-through rate does not establish a universal creative failure threshold.
Day 4: read the whole journey
Review spend, clicks, primary conversions and the quality of actual enquiries together. Read the totals before percentages. One conversion from ten clicks and two from fifty are both tiny samples, even though their rates look different.
As an illustrative calculation, a 3% conversion assumption applied to twenty clicks produces only 0.6 expected conversions. Seeing none would not, by itself, justify a firm channel verdict. That 3% is an example, not a ChatGPT benchmark.
Ask what happened after the click. Were visitors offered an appointment they could actually book? Did a product sell out? Did sales respond while the enquiry was still relevant? These operational details can matter more than another headline variation.
Days 5–6: make one interpretable improvement
If you have identified a clear issue, change the part responsible. A broken form should be repaired immediately. An offer mismatch may call for clearer page copy. A speculative creative change can wait until there is enough evidence to learn from it.
Log the time, reason and expected effect. Leave other variables stable where practical. If the budget is too small to give several ad groups useful exposure, simplify the test rather than adding more variants.
Day 7: choose a next step from the evidence
| Decision | When it makes sense |
|---|---|
| Continue within the agreed cap | Delivery and measurement work, but outcomes need more time or volume |
| Revise the test | You found a specific offer, audience or page problem to address |
| Pause | Tracking is unreliable, the cap is reached, or the current economics are unacceptable |
Check attribution windows, reporting time zones and event configuration before reconciling different dashboards. Reporting differences do not automatically mean tracking is broken. OpenAI explains measurement differences
Write the decision and the next review date. AthillyAds’ campaign preparation tools can help organise the draft and preflight checks; this playbook remains a practical review process for you and your team. Start with the campaign guide if the offer itself still needs work.