Ads Prompts

Ad copy lives or dies on the first five words. Every prompt you write should be optimised for one job: producing a hook that survives the scroll. Everything else — body, CTA, social proof — is downstream of the hook.

The mistake most teams make is asking for "five ad variations." The model produces five variations of the same hook. Instead, ask for five different hook types and let the strongest one earn the body copy.

• Pattern-interrupt hook — a sentence that breaks the visual rhythm of the feed.
• Specific-number hook — "We tested 47 landing pages…"
• Direct-callout hook — "If you run paid social at a 12-person company…"
• Contrarian-claim hook — "Most A/B tests are a waste of money."
• Story-cold-open hook — "It was 11pm on a Sunday when the CMO Slacked me."

• One hook, five variations — the model gives you five rewrites of the same idea.
• No character-count constraint — body copy gets truncated mid-sentence.
• Asking for "engaging" copy — "engaging" is meaningless. Specify the emotion.
• Skipping the offer line — without it the model defaults to soft features-and-benefits.
• Generic CTAs — "Learn more", "Click here". Replace with the actual next step.

Facebook / Meta primary-text prompt

You are a senior performance copywriter for [product] (1-line description).
Audience: [persona, pain, awareness level — cold / warm / retargeting].
Offer: [the exact offer + price point + risk reversal].
Constraint: primary text under 125 chars before "See more". Body under 90 words. One CTA.

Produce 5 ads, each using a DIFFERENT hook type:
1. Pattern-interrupt
2. Specific number
3. Direct callout
4. Contrarian claim
5. Cold-open story

For each: hook (one line), body (≤ 90 words), CTA (3-5 words), headline (≤ 40 chars).
Voice reference: [paste 2 of your best-performing ads].

Google Search RSA prompt

Generate 15 headlines (≤ 30 chars) and 4 descriptions (≤ 90 chars) for a Google Responsive Search Ad.

Product: [name + 1-line]
Primary keyword: [keyword]
Secondary keywords to weave in: [list]
Unique selling points: [3-5 bullets]
Match-type intent: [exact | phrase | broad]

Headlines must include:
- 3 with the exact primary keyword
- 3 with a number or stat
- 3 with a benefit
- 3 with a CTA
- 3 with social proof

Avoid: superlatives without proof, generic CTAs ("Learn more"), repeated phrasing.

1. Reverse-prompt your winners: paste your top 5 ads and ask the model to extract the structural pattern. Reuse the pattern for new campaigns.
2. Audience pasting: paste 50 real customer reviews and ask the model to write hooks using only words customers themselves used.
3. Multi-platform translation: write one strong ad, then ask for native rewrites in TikTok, Reels and Google Search voice — same offer, very different cadence.
4. Awareness-stage ladders: produce one ad per Schwartz awareness level (unaware, problem-aware, solution-aware, product-aware, most-aware) and match to retargeting audiences.