Competitor ad research

The deliverable is a strategy read, never the ad list itself. Read the copy, formats, targeting, and cadence, then explain how the competitor runs their account and what that means for the user's own ads.

How to reach Adwarp#

Prefer the Adwarp MCP tool inspect_competitor_ads if loaded. Otherwise the CLI: adwarp competitor ads <advertiser> (or npx -y adwarp competitor ads ...). No ad-account access is needed; this reads the public LinkedIn Ad Library.

Getting the right ads#

• Name plus company id. Name searches include unrelated advertisers. Supply both advertiser name and verified numeric companyId. The API searches by name and Adwarp filters the returned advertiser URL before queuing any creative work. Example: adwarp competitor ads <advertiser> --company-id <id> --max 100 --copy-max 10 --json. Total reported is the broad API query total, not an exact-company or active-ad count.
• Engines. api returns metadata. auto adds cached copy/screenshots from a remote creative worker or queues missing ads. Local scraping is disabled, including fallback. Configure ADWARP_CREATIVE_WORKER_URL and ADWARP_CREATIVE_WORKER_TOKEN on the CLI/MCP host.
• Coverage. Read up to 10 sampled creatives. Pending work is asynchronous: use get_competitor_creatives with IDs, or adwarp creative-status id1,id2, to check it. Status reads never visit LinkedIn or enqueue more work. Poll at reasonable intervals. Do not rerun discovery just to wait for copy. Report collection dates, sample coverage, missing copy, and whether a screenshot is a preview rather than a complete video. Screenshot paths require worker authentication; do not put tokens in URLs or reports.
• Cache. The worker adds a 7-day creative cache automatically. Existing temporary JSON exports are not imported automatically. Repeated ad IDs reuse jobs and cached results.
• Blocks and outages. If the worker is blocked, stop. Do not probe LinkedIn via another browser, curl, fetch, proxy or network. Never override worker limits, clear its state or resume it without explicit operator authorization. Return available metadata/cached data. If no worker is configured, say creative collection is unavailable; do not scrape locally.
• EU bonus data. Ads served in the EU carry run dates, impression ranges, per-country splits, and structured targeting facets. Use them; they are the closest thing to seeing a competitor's media plan.

Synthesis framework#

Work through these dimensions and ground every claim in specific ads (quote short copy snippets):

1. Messaging themes. The 2-4 recurring value props or narratives across the ads.
2. Offers and CTAs. What they ask for: demo, trial, report, webinar, event. The offer mix reveals which funnel stage they are buying.
3. Format mix. Single image vs video vs carousel vs document vs thought-leader ads, roughly proportioned.
4. Who they spotlight. Executives, customers, partners, product screenshots.
5. Cadence and scale. How many ads run concurrently, how often new ones ship (run dates where available), impression volume and geography.
6. Targeting (EU data where present): languages, locations, company and job facets, notable exclusions.

Comparing several competitors#

Run each pull separately, then produce 1 comparison: a table across the dimensions above, then a strategic read of who spends hardest, whose positioning overlaps the user's, and which offer types nobody in the set is running.

Report format#

• How they run it: 1 paragraph per competitor summarizing the account strategy.
• Dimension findings: themes, offers, formats, spotlight, cadence, targeting, each with example ads quoted.
• So what: implications for the user's own account. Over-used angles to avoid, gaps worth testing, offers worth copying, formats they are ignoring. Make these specific enough to brief an ad from.