Claude Skills for Google Ads: 6 Downloadable Skills for Copy, Campaigns, and Audits

If you spend your days inside Google Ads, the LLM workflow most teams default to is exhausting: open a chat, paste your brand voice, paste your USPs, paste the landing page, paste the keyword, paste the audience, then write a one-paragraph prompt explaining what you actually want. Three ads in, you’ve spent 40 minutes on context-pasting alone.

Claude skills fix this by encoding the prompt once, in a file. You install the skill, run it with a one-line invocation, and the skill follows its own internal rulebook, character limits, pinning logic, hard rules about landing-page alignment, the lot. Pair them with Claude Memory and your brand voice and USPs are loaded automatically too.

We’ve published six free Claude skills for Google Ads. This post is the operator’s guide to all six: what each does, when to reach for it, and how they chain together for a real campaign workflow.

Download the bundle →

The six skills, at a glance

All six are free. All six work on the standard Claude plan (no API billing on top). And all six are designed to be boring in the right way, they enforce policy, count characters, pin headlines, and avoid banned claims so you don’t have to.

1. google-ads-rsa-headlines, the everyday workhorse

Every other skill on this list is occasional. This one runs almost every day in an active Google Ads account.

What it does: Generates 15 Responsive Search Ad headlines (each ≤30 characters) and 4 descriptions (each ≤90 characters), labelled by angle (KEYWORD / BENEFIT / FEATURE / PROOF / CTA / URGENCY / SPECIFICITY), with pinning suggestions tailored to the offer.

What makes it different from a one-off prompt:

• It enforces character counts as hard limits. If a headline lands at 31 characters, it rewrites it. No “close enough”.
• It hits a target angle distribution across the 15 headlines (3 KEYWORD, 4 BENEFIT, 2 FEATURE, 2 PROOF, 2 CTA, 1 URGENCY, 1 SPECIFICITY) so Google has variety to test.
• It pins for you. Headline 1 and 2 to position 1 (strongest keyword-forward). Headline 3 to position 2 if it carries a non-negotiable offer.
• It bans superlatives (“best”, “#1”, “world’s leading”) unless you’ve supplied proof.
• It re-checks against the landing page (if you give it a URL) so headlines don’t promise something the LP doesn’t deliver.

One-line invocation (with memory configured):

“Use google-ads-rsa-headlines for our merino wool socks ad group, LP outerknown.com/merino-socks, primary keyword ‘merino wool socks’.”

Output is paste-ready into Google Ads Editor or the UI.

See the full skill →

2. pmax-asset-pack, the campaign-launch heavyweight

Performance Max needs more assets than any other campaign type and most operators under-supply. Google’s own recommendation is “max out every field” and the skill is built to hit that target without filling the gaps with junk.

What it does: Generates a complete PMax asset group:

• 15 short headlines (≤30 chars)
• 5 long headlines (≤90 chars)
• 5 descriptions (≤90 chars)
• 4 sitelinks (title ≤25 chars + 2 desc lines ≤35 chars each)
• 10 callouts (≤25 chars)
• 8 structured snippet values for the best-fit header
• 3–5 audience signals (in-market, custom segment, customer match)

When it earns its keep:

• New asset group at campaign launch, the 90-minute write becomes a 10-minute review of generated output.
• Quarterly asset refresh, most accounts let assets stagnate and slowly bleed Quality Score. The skill makes a refresh a small task instead of a project.
• Asset group splits, when you split a PMax theme into two (e.g., “hiking” and “running”), the skill lets you stand up the second asset group in minutes.

Hard rules the skill enforces:

• No two assets carry the same hook (variety is what PMax is built on).
• No promo claims with silent expiries (if you give an end date, it’s included).
• Skipped categories on the callouts list (don’t write “Free Returns” if you don’t offer them).

See the full skill →

3. google-ads-campaign-builder, the blueprint generator

This is the skill that turns “I want to launch a campaign for our hiking sock line” into a full Google Ads Editor blueprint.

What it does: Takes a brief (product, LP, geo, budget, AOV, goal) and returns:

• Campaign settings (name, type, networks, location, language, bidding strategy, ad schedule)
• 3–7 ad groups, each with: name, LP, match-type strategy, keywords (broad/phrase/exact), and ad-group-level negatives
• 3 RSAs per ad group (delegated to google-ads-rsa-headlines)
• Campaign-level sitelinks (4–6), callouts (10), and structured snippets
• 30–50 campaign-level negatives organised by theme
• Budget split with explicit math
• A pre-launch checklist (conversion actions, enhanced conversions, audience signals, remarketing tag, GMC link)

Where it’s especially useful:

• Replatforming from another channel manager or in-house build.
• Standing up a new geography or language for an existing campaign.
• New product line that doesn’t fit existing campaigns.
• Agency takeovers where you want a clean rebuild rather than inheriting the previous operator’s structural choices.

What it intentionally does not do:

• It doesn’t write to your account. The output is a blueprint.
• It doesn’t recommend SKAGs by default (Google’s matching has shifted; the skill recommends themed ad groups).
• It doesn’t enable Display Expansion or Search Partners (off by default, always).

One-line invocation:

“Use google-ads-campaign-builder for premium-bottles.com, US-only, $50/day, Max Conv goal, $40 AOV, conversion tracking live.”

See the full skill →

4. google-ads-account-auditor, the takeover and refresh tool

The most common audit format LLMs produce is generic, “improve Quality Score, add more negatives, use Smart Bidding”. This skill enforces the opposite: every finding cites a specific campaign, ad group, keyword, or metric, and every recommendation has a measurable expected outcome.

What it does: Walks an 8-section audit in this order:

1. Conversion tracking (always first, bad tracking poisons everything)
2. Account structure
3. Ad copy quality
4. Keyword hygiene
5. Bidding + budget
6. Audience layering
7. Extensions + assets
8. Shopping / PMax (if present)

It outputs every issue with severity (Critical / High / Medium), evidence, and fix. The full report is then ordered by severity, not by section, so a Critical conversion-tracking issue floats above a Medium audience- overlap finding regardless of which section surfaced it.

Where the skill bites hardest:

• New client takeovers. The first-day audit becomes a 60-minute task instead of three days.
• Pre-Q4 prep. You want to enter peak with no Critical issues in the account. The auditor finds them all at once.
• ROAS-lifting projects. You need a prioritised punch list, not a wall of recommendations.

Input format: Paste a Google Ads Editor export, or describe your account structure + share 90 days of conversion data by campaign.

See the full skill →

5. search-terms-to-negatives, the monthly hygiene skill

Most accounts leak budget on the same 10–30 query patterns every month. Manually reviewing the search-terms report is brutal at any scale beyond a small account, and the typical LLM approach (“here’s a list, suggest negatives”) produces a flat list of mostly-useless candidates.

What it does: Takes a search-terms CSV and returns negatives organised into themed groups:

• Job-seeker intent (jobs, careers, salary, hiring)
• DIY / free intent (free, diy, how to make)
• Wrong-product intent (specific mismatches based on your category)
• Competitor brands you don’t want to bid on
• Inappropriate context (vertical-specific)

For each negative, it suggests the right match type (exact, phrase, or broad), most operators add too many as broad and end up blocking valid queries.

Where it earns its keep:

• Monthly negatives review (replaces a 2-hour task with 15 minutes).
• Post-launch cleanup in the first 30 days when search terms are wild.
• Quality Score recovery projects.

One-line invocation:

“Use search-terms-to-negatives on this CSV [paste]. Account is a premium water bottle store, exclude competitor terms.”

See the full skill →

6. low-ctr-ad-rewrite, the diagnostic copywriter

You see an ad with a 0.6% CTR while the ad group baseline is 2.1%. The typical move is to pause and add another headline variation. The skill goes deeper: it tags each rewrite with the hypothesis it tests.

What it does: Takes a losing ad (paste headlines + descriptions) and returns 5 rewrites. Each rewrite is tagged with a hypothesis like:

• “Hypothesis: hook is too feature-led. Trying benefit-led.”
• “Hypothesis: too generic. Trying SKU-specific.”
• “Hypothesis: no urgency. Trying time-bound CTA.”
• “Hypothesis: superlative-heavy and fluffy. Trying numbers-first.”
• “Hypothesis: ignores the keyword. Trying keyword-front.”

Why hypotheses matter: If you ship 5 rewrites without hypotheses and two improve CTR, you’ve learned nothing. With hypotheses, you’ve learned which framing your audience responds to, and that compounds across future ads.

When to use:

• Any ad under group baseline by > 30% with > 1000 impressions.
• Pre-Q4 stress-testing your hero campaigns.
• After a big shift in the LP or offer (old ads might be off-brand now).

See the full skill →

The chained workflow: how the six skills compose

Here’s the workflow a real campaign launch looks like, end-to-end:

Brief (from client or marketing)
      │
      ▼
[google-ads-campaign-builder]  ──>  Blueprint (campaigns, ad groups, keywords, negatives, budget)
      │
      ▼
[google-ads-rsa-headlines]     ──>  3 RSAs per ad group (called by builder, or run separately)
      │
      ▼
[pmax-asset-pack]              ──>  PMax asset group if PMax is in the mix
      │
      ▼
Build in Google Ads Editor, upload, launch.
      │
      ▼ (Day 7+)
[search-terms-to-negatives]    ──>  First cleanup pass
[low-ctr-ad-rewrite]           ──>  Rewrite any ad underperforming baseline
      │
      ▼ (Day 30+)
[google-ads-account-auditor]   ──>  Full audit, repeat monthly or quarterly

The whole chain runs inside one Claude conversation (or several if you want each skill in isolation). Memory entries are inherited, so your brand voice, banned claims, USPs, and audience profile show up in every output without you re-supplying them.

If you want to see the chain run end-to-end with real prompts, see the chaining skills walkthrough.

What the skills don’t do (and what to pair them with)

Skills generate copy and audit findings. They don’t push changes to your account. For that, you need a server that lets Claude read and write live account data. The relevant servers are:

• Google Ads MCP, gives Claude read/write access to your Google Ads account. Pair with google-ads-account-auditor to audit live, or google-ads-campaign-builder to push the blueprint straight to the account.
• Google Shopping API, for catalog-side workflows (pair with product-feed-quality-auditor and gmc-disapproval-fixer, see the feed management skills guide).

The skills work standalone. They get better with MCPs underneath them.

What’s next

If you’ve installed all six skills, here’s the recommended order to adopt them in:

1. Start with google-ads-rsa-headlines. Run it on three ad groups you know well. Compare the output to what you’d write manually.
2. Run google-ads-account-auditor on one client account. See if it surfaces issues you’d missed.
3. The next time you launch a new campaign, run google-ads-campaign-builder instead of writing from scratch.
4. Add search-terms-to-negatives to your monthly recurring tasks.
5. Use low-ctr-ad-rewrite whenever an ad goes off-baseline.
6. Use pmax-asset-pack on the next PMax asset group refresh.

If you also run Meta, the parallel set of skills lives in the Claude skills for Meta Ads guide. If you also manage product feeds, see the product feed management skills.

Download all 16 skills →