# See which ads your competitors keep running. And the offers behind them Competitor ad research finds the ads your competitors keep paying for, which are the ones that work for them, and the offers they lead to. With Agentik, your AI pulls their public Meta and Google ads with AI breakdowns from Adspy.sh, reads their landing pages with Firecrawl, and compares those pages with earlier copies from the Wayback Machine. - For: Founders, performance marketers and agencies - Providers: 3 (Adspy.sh, Firecrawl and Internet Archive), 6 tools - Example (3 competitors, 20 active ads each, 15 ads broken down, 6 landing pages read today and six months ago): at most $0.67, computed from the live catalog - Updated: 2026-10-07 ## Each provider does one part No single provider covers this job. Through Agentik your AI calls each one for its part, from one connection and one prepaid balance. - **Adspy.sh** (4 tools, at most $0.61): Competitors’ public Meta and Google ads, how long each has run, hook and format breakdowns, and each brand’s strategy. - **Firecrawl** (1 tool, at most $0.06): The landing pages the ads send to: offer, price, proof and call to action. - **Internet Archive** (1 tool, free): The same landing pages months ago, to see how the offer moved. ## Give this to your agent ```text Use my connected Agentik MCP server for this job: competitor ad research. Research my competitors’ ads. First ask me for three competitors and a short description of my product and offer. Find each brand with adspy/search-brands. Pull their longest-running active ads with adspy/search-creatives, break down the top five each with adspy/get-creative-analysis, and compare the three brands’ strategies with adspy/compare-brands. Read the landing pages their ads link to with firecrawl/scrape, find each page’s copy from six months ago with wayback/availability, and read it too. Finish with what keeps running and why, how each offer changed, and a brief for my next five ads: hook, format, message and offer. Before running anything, inspect each tool with inspect_tool, ask me for any input you don’t have instead of inventing it, and tell me the total price. Wait for my go. At the end, tell me what Agentik charged. ``` ## You provide - Three competitors - Your product and offer, for the brief ## What you get: a brief for your next five ads - Each competitor’s longest-running ads, with format and days active - The hooks, messages and formats that keep running, brand by brand - Their landing pages: offer, price, proof and call to action, then and now - A brief for your next five ads ## How to research your competitors’ ads in Claude or ChatGPT Prices are the most each call can reserve for this example: 3 competitors, 20 active ads each, 15 ads broken down, 6 landing pages read today and six months ago. ### 1. Find the brands Adspy.sh finds each competitor in its library of public ads. - [Find brands in the ad library](/tools/adspy/search-brands.md) `adspy/search-brands` (Adspy.sh, 3 brands): at most $0.003 ### 2. Pull the ads they keep running The longer an ad runs, the likelier it pays for itself, so their active ads are sorted by days running. - [Search ad creatives](/tools/adspy/search-creatives.md) `adspy/search-creatives` (Adspy.sh, 20 ads each): at most $0.12 ### 3. Break down why they work Adspy.sh’s AI analysis gives each ad’s hook, message, visuals and format, and compares the three brands over 90 days. - [AI analysis of ad creatives](/tools/adspy/get-creative-analysis.md) `adspy/get-creative-analysis` (Adspy.sh, 5 ads each): at most $0.12 - [Compare brands' creative strategies](/tools/adspy/compare-brands.md) `adspy/compare-brands` (Adspy.sh, 3 brands, 90 days): at most $0.36 ### 4. Read where the ads lead Firecrawl reads each landing page: the offer, the price, the proof and the call to action. - [Scrape a page to markdown](/tools/firecrawl/scrape.md) `firecrawl/scrape` (Firecrawl, 6 pages): at most $0.03 ### 5. See how the offer changed The Wayback Machine finds each page’s copy from six months ago, and Firecrawl reads it for comparison. - [Archived page lookup](/tools/wayback/availability.md) `wayback/availability` (Internet Archive, 6 pages): free - [Scrape a page to markdown](/tools/firecrawl/scrape.md) `firecrawl/scrape` (Firecrawl, 6 archived copies): at most $0.03 ### 6. Brief your next ads Your AI turns what keeps running into a brief for your own ads, with a reason for each choice. Example total: at most $0.67. Each call is billed for measured usage; unused reservations are released. ## Ask it your way - Which of Competitor X’s Meta ads have run the longest, and what do they promise? - Compare the hooks of our three main competitors’ video ads this quarter. - Did Competitor Y switch from a free trial to a demo on its ad landing page? - Write five ad briefs that answer Competitor Z’s best-running ad. ## Why through Agentik - The ad library, landing pages and archive come from three providers, without an Adspy.sh plan or a Firecrawl account. - Adspy.sh credits and Firecrawl pages come out of the same balance, priced before each run. ## Good to know - These are public ads from the Meta Ad Library and Google Ads Transparency. Spend and results aren’t public, so days running is the signal. - The Wayback Machine may not hold a copy of every landing page. ## When to use something else To monitor many brands with alerts and a shared board, use Adspy.sh’s own app. Agentik suits research your AI runs and turns into briefs. ## Competitor ad research: questions ### Which platforms are covered? Meta, including Instagram, and Google, including YouTube, from Adspy.sh’s library of public ads. ### Couldn’t I just use Adspy.sh? For the ads, yes. Reading the landing pages and comparing them with earlier copies needs Firecrawl and the Wayback Machine. Through Agentik your AI uses all three from one connection; the example costs at most $0.67. ### Can it see their ad spend? No. Spend and results aren’t public; how long an ad keeps running is the best public signal of what works. ## Keep going - [Product photo ads](/product-photo-ads.md): Your product page for the copy, your category’s longest-running ads as the brief, then new scenes, headlines and a clip. - [AI video ads](/ai-video-ads.md): Hooks from your niche’s longest-running video ads, objections from YouTube reviews, then frames, clips, a voiceover and music. - [Social listening](/social-listening.md): Posts on X, Hacker News threads, YouTube comments and news coverage, read together and sorted into themes. - [Advertising tools for your agent](/tools/advertising.md) - [Scraping tools for your agent](/tools/scraping.md) - [Knowledge tools for your agent](/tools/knowledge.md) [All playbooks](/playbooks.md) · [Connection guides](/docs/clients.md) · [Billing](/docs/billing.md) --- [HTML page](/competitor-ad-research) · [Agent index](/llms.txt)