The first week of a ChatGPT ad campaign: what to do, day by day
Days one to three: nothing. Day four: one honest read. Day seven: the verdict. A week-one playbook for new ChatGPT advertisers, and the panic-edits to avoid.
The first week of a new ChatGPT campaign has one job: produce a trustworthy answer to “does this channel work for my business?” Most first weeks fail at that job — not because the ads were bad, but because the advertiser kept poking them. Here’s the week as it should go.
Before day one: the checks that save the week
Three things to verify before submitting, each worth a full day of review latency if you skip it:
- Every card has an image, at least 256 × 256, with no text baked into it.
- Every landing URL loads, doesn’t redirect somewhere weird, and doesn’t block OpenAI’s crawler.
- Every claim on the card is visible on the landing page. “On site in 60 minutes” on the ad and nowhere on the site is a rejection.
And one thing that gates everything: a fresh ad account can’t serve until its brand review passes, and a missing favicon is the usual silent blocker. Approved ads plus zero impressions? Check that first, then check your bid.
Days 1–3: change nothing
Your campaign has a few hundred impressions and single-digit clicks. Every instinct says do something. The correct move is almost always nothing — at these volumes, CTR is a coin flip, and every edit muddies the only read you’re waiting for. The one exception: zero impressions by end of day two means a delivery problem (bid too low, brand review pending), not a performance problem. Fix that; touch nothing else.
Our coach’s day-two message is literally: “Not enough data yet (312 impressions). Change nothing.” We put restraint in the product because it’s the hardest advice to follow alone.
Day 4: one honest read
By now there’s enough to read directionally. Three questions, in order:
- Are we serving? Impressions growing daily? If an ad group has reached nobody, its hints are off — too narrow, or describing conversations that don’t happen. Rewrite that group’s hints (broader scenario, same intent); it’s the one early edit that’s usually right.
- Are we relevant? CTR meaningfully below ~1% suggests the card doesn’t match the conversations it’s entering. Check the weakest group’s hints against its card copy — they should tell one story.
- Is the click landing? Watch the tagged traffic. High bounce on ad clicks means the landing page breaks the promise the card made.
One read, at most one intervention. Then hands off again.
Days 5–6: let the difference emerge
If you ran two or three cards per ad group (do this — it’s the only A/B testing this platform has), a gap between them starts showing now. Don’t crown a winner yet; note it.
Day 7: the verdict
Now the numbers mean something. Sit down with three of them:
- Cost per enquiry (or per order), against what an enquiry is worth to you.
- The spread between ad groups — usually one intent (acute need vs. comparison vs. brand) clearly outworks the rest. That’s next week’s budget map.
- The spread between cards — keep the winner, rewrite the loser toward what the winner does differently.
Week one rarely delivers profitability; it delivers a slope. Improving CTR, sane cost per enquiry, one ad group clearly working — that’s a channel worth week two. Flat everything at healthy bids and honest hints — that’s an answer too, and it only cost you a week.
This day-by-day rhythm — restraint, one honest read, verdict — is exactly what AthillyAds’ coach walks you through automatically, and week one of coaching is free on every plan. However you run it: the discipline is the strategy. The advertisers who lose week one are the ones who couldn’t leave it alone.