Linkedin Ad Teardown

LinkedIn Ad Teardown — competitor paid-creative strategy read

LinkedIn's Ad Library is public (it exists for ad transparency). This skill reads a competitor's ads from it and turns them into a strategy read: what themes and offers they're pushing, in what formats, how often, who they're targeting (from the EU "who's shown this" disclosure), and — the part that matters most — where the gaps are that the client can move into.

Adapted from github.com/stan-default/liam's liam-competitors / inspect_competitor_ads (MIT), accessed 2026-07-14, via /steal — see .claude/discovery/0726-liam-steal-analysis.md. Concept port; the sourcing runs on our scraper stack, not Liam's Playwright engine.

Doctrine inherited

• quantitative-evidence-floors.md — no strategy read below the floor. State how many ads the read rests on.
• crawl-cost-discipline.md — free discovery first (the public Ad Library URLs), metered extraction only on the ads you keep.
• apify-credits.md — if an Apify actor is used for bulk, gate it and estimate the cost first.
• storage-policy.md + pii-redaction.md — the Ad Library is public transparency data (no member PII); route the teardown to {client}/competitors/, don't over-collect.

Sourcing — public, gated, cheapest-first

The library lives at https://www.linkedin.com/ad-library/ — searchable by advertiser, no login for the basic view. Source in this order:

1. Browser MCP (free). Navigate the public Ad Library, search the advertiser, read the results. The cheapest path — try it first.
2. Firecrawl scrape (metered). When the page is too JS-heavy to read cleanly via the browser.
3. Apify LinkedIn Ad-Library actor (metered, gated). Only for bulk across many competitors, and only after the apify-credits.md estimate + go-ahead.

Never scrape login-gated member data — the Ad Library only. It's transparency data, but stay polite (crawl-cost-discipline.md): respect rate limits, don't hammer.

What to extract per competitor

• Inventory — how many active + recent ads (this is the count the whole read rests on — cite it).
• Themes — the recurring messages, value props, and pains they lead with.
• Offers — demos, trials, gated content, events, discounts; what the CTA asks for.
• Formats — single image / carousel / video / document / thought-leader ad.
• Cadence — from launch dates, the posting rhythm (steady drip vs burst).
• Targeting signal — from the EU disclosure ("who's shown this": roles, seniorities, countries) where present.
• Gaps — the formats, angles, and segments they're not running. This is the opening.

The floor — voice-locked

No strategy read below the floor. Under ~5 ads, report the inventory only — not a "strategy." A competitor with 2 ads isn't a strategy you can reverse-engineer; it's two ads. State the count every time (quantitative-evidence-floors.md), and when the disclosure doesn't show targeting, mark it [UNAVAILABLE] — never infer who they're targeting from the creative alone.

Anti-patterns

• ❌ Reading a strategy from 2 ads — inventory only below the floor.
• ❌ Scraping login-gated member data — Ad Library only.
• ❌ An ungated Apify bulk run — estimate + gate per apify-credits.md.
• ❌ Inferring targeting the EU disclosure didn't show — mark [UNAVAILABLE].
• ❌ Design-cloning a competitor's creative — this is gap-finding + inspiration, not a copy job.

Integration with other skills

• Feeds /competitor-research — the paid-creative dimension its 11-dimension profile doesn't cover.
• Feeds /paid-campaign-strategy + /linkedin-ads-copy — the gap becomes our angle.
• Pairs with a1-gallery-protocol.md when you want design references for the formats you spotted.

Final ship gate

Run /premortem --output before ship. See /premortem skill for the 5 execution domains + output template.