ad-copy-writer

Generate ad copy variations for Google, Meta, and LinkedIn Ads

SKILL.md

GitHub

Ad Copy Writer

Generate high-converting ad copy for Google, Microsoft, Meta, LinkedIn, TikTok, Reddit, Snapchat or X Ads. Works with no accounts connected.

Usage

/ad-copy-writer google "project management tool" — Google Ads copy /ad-copy-writer meta example.com — Meta Ads copy from website /ad-copy-writer linkedin "B2B SaaS" — LinkedIn Ads copy /ad-copy-writer reddit|tiktok|snapchat|x|microsoft example.com — other platforms

Works with

Job Primary Also answers it What differs
Product and competitor research WebFetch / WebSearch — Always available
Angles already running (enhanced when connected) The target platform's ad MCP Any other connected ad platform Read-only; used to avoid repeating angles, never to edit ads

Pulling live ads per platform (tool names omit mcp__cogny__mcp__<Prefix>__):

Platform Prefix Read current copy
Google Google_Ads tool_execute_gaql(customer_id, query: "SELECT ad_group_ad.ad.responsive_search_ad.headlines, ad_group_ad.ad.responsive_search_ad.descriptions, metrics.ctr FROM ad_group_ad WHERE ad_group_ad.status = 'ENABLED' AND segments.date DURING LAST_30_DAYS")
Microsoft Bing_Ads tool_get_ad_groups → tool_get_ads(account_id, ad_group_id)
Meta Meta_Ads tool_get_ads(ad_account_id, status_filter: ["ACTIVE"], include_creative: true)
LinkedIn LinkedIn_Ads tool_get_creatives(ad_account_id, status_filter: ["ACTIVE"])
TikTok TikTok_Ads tool_get_ads(advertiser_id, status_filter: [...])
Reddit Reddit_Ads tool_get_ads(ad_account_id)
Snapchat Snapchat_Ads tool_list_organizations → tool_list_ad_accounts → tool_get_creatives(ad_account_id)
X X_Ads tool_get_tweets(account_id, tweet_type: "PUBLISHED"), tool_get_cards(account_id) for card headlines
Apple Search Ads Apple_Search_Ads tool_list_acls → tool_get_campaigns → tool_get_ad_groups → tool_get_targeting_keywords (keywords, not copy)

Steps

1. Research the product

If a URL is provided, WebFetch it and extract:

  • Value proposition
  • Key features / benefits
  • Target audience
  • Pricing (if visible)
  • Unique differentiators

If a description is provided, use that directly.

2. Research the competition

WebSearch for competitors to understand:

  • What messaging angles they use
  • Common claims in the space
  • Price anchoring

2b. Enhanced when connected: read what's already running

If the target platform is connected (check mcp__cogny__list_integrations), pull its active ads with the tool above. List the angles, claims and CTAs already in use, and mark the top performers by the platform's own CTR where available. Then write new copy that tests different angles, and say which existing angle each new variation is meant to beat. If nothing is connected, skip this step.

3. Generate copy by platform

Google Ads (RSA format)

Generate 15 headlines (max 30 chars each) and 4 descriptions (max 90 chars each):

Headlines — mix of:

  • Benefit-led: "Cut Ad Spend 40%"
  • Feature-led: "AI-Powered Optimization"
  • Social proof: "Trusted by 1,000+ Agencies"
  • CTA: "Start Free Trial Today"
  • Keyword-match: "[Primary keyword] Tool"

Descriptions — mix of:

  • Problem/solution
  • Feature list
  • Social proof + CTA
  • Urgency/offer

Meta Ads

Generate 3 variations:

  • Primary text (50-150 chars recommended for single-image Feed ads, per Meta's Ads Guide; up to 500)
  • Headline (27 chars recommended for single-image Feed ads; longer headlines truncate)
  • Description (30 chars)
  • CTA button recommendation

Angles:

  1. Pain point → solution
  2. Social proof → aspiration
  3. Direct value prop → CTA

LinkedIn Ads

Generate 3 variations:

  • Intro text (150 chars recommended to avoid truncation; 3,000 max)
  • Headline (70 chars recommended; 200 max)
  • Description (100 chars recommended; 300 max; shown only on LinkedIn Audience Network placements)

Angles:

  1. Professional credibility
  2. ROI / efficiency focus
  3. Industry-specific insight

Microsoft Ads (RSA format)

Same limits as Google: 15 headlines (30 chars), 4 descriptions (90 chars), 2 display paths (15 chars). Reuse the Google set, but check the audience skews older and more desktop/B2B (LinkedIn profile targeting is available), so favour clarity over slang.

TikTok Ads

Generate 3 variations:

  • Ad text (1–100 chars; keep under ~80 so it isn't truncated)
  • Display name (brand, up to 40 chars)
  • CTA button recommendation Angles: native hook in the first line ("POV:", "I tried…"), demo/result, creator testimonial. Copy supports the video; it never explains it.

Reddit Ads

Generate 3 variations:

  • Headline / post title (max 300 chars; aim for 60–120)
  • CTA button recommendation Angles: plain-spoken problem statement, honest comparison, "we built this because…". No hype words; Reddit users downvote ad-speak.

Snapchat Ads

Generate 3 variations:

  • Brand name (max 25 chars)
  • Headline (max 34 chars)
  • CTA recommendation Angles: instant benefit, offer/urgency, playful/social.

X Ads

Generate 3 variations:

  • Post text (max 280 chars; a URL counts as 23)
  • Website card headline (max 70 chars) Angles: hot take / contrarian, number-led proof, thread-style teaser.

Apple Search Ads

There is no free-text ad copy: ads are built from the App Store listing (app name 30 chars, subtitle 30 chars, screenshots) and custom product pages. Instead deliver: subtitle options (30 chars), 3 custom product page themes matched to keyword groups, and the keyword groups themselves. If Apple_Search_Ads is connected, read live keywords with tool_get_targeting_keywords(org_id, campaign_id, ad_group_id) first.

4. Output format

Present each variation clearly labeled with character counts:

=== Google Ads RSA ===

Headlines (30 char max):
1. "Cut Your Ad Spend by 40%" (25)
2. "AI Marketing Automation" (24)
...

Descriptions (90 char max):
1. "Stop wasting budget on underperforming ads. Our AI optimizes your campaigns 24/7." (83)
...

=== Meta Ads ===

Variation 1 (Pain → Solution):
Primary: "..."
Headline: "..."
...