The AI Ad Copy Workflow We Run for Clients

To turn AI-written copy into usable ads, give it a specific buyer problem, an approved claim, and a clear reason to respond. Then review the wording before spending money to test it. Generating more lines helps only when they make different, credible arguments.

The short version: Our workflow tests five kinds of argument: pain, proof, a contrarian view, price, and social norms. We generate versions for buyers at different stages, then check claims, ad policies, and brand voice. Approved copy runs with matching visuals so we can compare which argument works. One strategist usually completes a sprint in about a day.

In a sprint of 40 copy variants, we may cut about 15 during review. Below, we explain the prompts, selection rules, and how results shape the next batch.

This is the copy process within our AI performance creative work. It runs alongside the visual workflow, so the words and images make the same case.

What is the AI ad copy workflow?

The workflow has five steps. First, define the angles and generate variants for each angle and awareness level. Review every line for evidence, policy, and voice before pairing it with matching creative. Then read the results by angle.

An older Nielsen Catalina analysis attributed about 47% of sales lift to creative and 9% to targeting. The 2017 finding came from consumer-goods campaigns, so it is evidence of creative’s importance, not a current universal split. The words still carry the hook, offer, and objection handling.

How do you build a hook and angle taxonomy?

Start with five angle families: pain, proof, contrarian, price, and social norm. Give every hook one family and one awareness level. The taxonomy turns a pile of lines into a structured test. It shows which argument won, not only which sentence won.

Five families cover most of our paid-social work. Larger taxonomies often blur. When two strategists classify the same hook differently, angle-level analysis becomes less useful.

Write one sentence that defines each angle before generation. Add one positive and one negative example. This small reference keeps tagging consistent when several people review the batch.

Angle family	What it sounds like	When it wins
Pain	"You have re-downloaded this app four times"	Cold, problem-aware traffic that has never heard of you
Proof	"4,000 reviews and the same sentence keeps showing up"	Solution-aware buyers comparing options, retargeting
Contrarian	"Skip the 10-step routine. You need two."	Saturated categories where every ad makes the same promise
Price	"That is $0.40 a day, less than the tip on your coffee"	Product-aware audiences stalling at checkout, promo windows
Social norm	"Half your group chat already switched"	Broad consumer products with word-of-mouth loops

The taxonomy also improves briefs. When a client asks for more ads, ask which argument lacks coverage. The table gives both teams a shared vocabulary.

How do you generate variants per angle and awareness level?

With the angles defined, generate small batches: one prompt per angle and awareness level, with eight hooks per prompt. Add word limits, a spoken-register note, approved claims, and examples from past winners. A broad request for 40 hooks usually returns minor variations of one idea.

Here is the hook prompt, shortened:

For [product], write 8 ad hooks in the PAIN angle for a
PROBLEM-AWARE audience. The reader knows the frustration
([specific frustration]) but not the product category.
Rules: under 12 words each, no brand name in the hook,
no exclamation points, spoken register, each hook names
a different concrete moment of the pain.
Output as a numbered list, one line each.

Winners from the hook pass get expanded into primary text with a second prompt:

Take hook #3 and write 3 versions of primary text, 40 to
80 words each. Voice reference: [3 pasted lines of past
winning copy]. Stay inside this claims inventory:
[approved claims list]. Do not invent statistics, awards,
guarantees, or customer quotes.

The claims inventory is a client-specific doc: every claim the founder and legal have signed off on, with the source next to it. The model never gets to argue from outside it.

Keep a version history and source dates for that inventory. Product terms, prices, and evidence change, so a line approved last quarter may need another review before launch.

What gets killed in the human filter step?

We remove unsupported claims, policy risks, and voice drift. Roughly one-third of a generated batch fails this review in our sprints. That is an internal observation, not an industry benchmark. One strategist usually finishes the pass within an hour.

Unsubstantiated claims. The model writes confident numbers with no source behind them. “Clinically proven to double retention” is a great hook and a legal problem. If a claim is not in the inventory, the line dies or gets rewritten around a claim that is.

Platform policy risk. Meta restricts ads that assert or imply personal attributes. Pain-angle copy often pairs “you” with a named insecurity. We rewrite those lines to describe the situation instead. “Your acne” becomes “the third concealer this month.”

Brand voice drift. By variant 30, the model slides toward the category-average voice. We paste past winners into every prompt and still catch it. The test is blunt: if a line could run under a competitor’s logo unchanged, it dies.

Record why each line was removed. Those rejection tags improve the next prompt and reveal recurring policy or voice problems. The review then produces reusable guidance, not only a shorter list.

How do you pair copy with creative without exploding the test matrix?

Once the copy passes review, pair it with creative by angle. Testing all six copy options against all six visuals would create 36 cells, too many for most budgets. Use one cell per angle, with 4 to 5 variants.

The pairing rule holds because a mismatched cell is unreadable anyway. Contrarian copy over a testimonial visual is two arguments interrupting each other. When that ad loses, you cannot say which half failed, so the result teaches you nothing.

For one consumer subscription client, this reduced 48 proposed combinations to 5 angle cells. The final plan had about 22 ads. It kept the same learning agenda and required less budget.

Why read results at the angle level, not the variant level?

Individual variants often receive too few conversions for a reliable conclusion. Pooling results by angle gives the test more data. It also answers the question that shapes the next batch: which argument is working?

Meta’s current performance guidance recommends creative diversification so its system can match messages to people. It does not assign a universal share of performance to creative. Our angle read makes the recommendation practical. The next sprint gives the winning argument 16 fresh hooks and retires weak angles.

Angle-level reads also change how you handle fatigue. When a winning angle’s numbers sag, you rotate fresh variants inside that angle instead of scrapping the argument. Our guide to detecting ad creative fatigue covers the warning signals; the taxonomy is what makes the refresh cheap.

Where this fits in a full creative system

The same principle carries through the full creative sprint: copy and visuals share one set of angles. We generate, review, and assess them together so each result informs the next brief.

We can help you see which arguments your current ads cover and which buyer objections still need an answer. Book a Free Strategy Call and bring a recent batch. Together, we will identify the ideas worth testing next.