Tech • AI • Robotics
Claude Code Scraped 3 Months of Winning Meta Ads & Rebuilt Them
5/10TL;DR
Claude Code can now automate much of the Meta advertising workflow, from competitor research and campaign creation to asset uploads and performance review, reducing reliance on traditional agency media buying.
KEY POINTS
Automation targets agency work
AI-driven ad operations are increasingly replacing tasks once handled by marketing agencies and specialist media buyers. The approach focuses on using Claude Code with scripted rules to research markets, build Facebook and Instagram campaigns, and push them into Meta Ads Manager with less manual work. Some operators still keep a human approval step, especially when campaigns use a company’s own ad budget.
Meta Ads Manager complexity is a key pain point
Meta Ads Manager remains difficult for first-time users because it combines technical setup with strategic decisions on targeting, budgets, creatives, and optimization. AI agents are being used to reduce both forms of overload by identifying suitable audiences, recommending campaign structures, and later reading campaign data to guide adjustments.
Business prerequisites still matter
Automation does not remove the need for a functioning commercial setup. Advertisers still need a defined offer, a landing page, pricing, a payment processor, lead nurturing systems, onboarding processes, and a testing budget. One campaign example used a budget of $2,000 for a two-day intensive priced at $599 per seat.
Technical setup requires Meta infrastructure
The workflow depends on a Facebook account, a Meta developer account, a Meta app, and access to Business Manager assets such as pages, ad accounts, and Instagram profiles. A system user with assigned permissions and a generated access token is required so an external tool can act on behalf of the ad account. Without that infrastructure, the automation cannot create or manage campaigns.
A Meta Ads MCP connector enables control
Inside VS Code, the system used a local Meta Ads MCP with 95 tools connected to the Meta app. That connector allows AI to inspect ad accounts, create campaigns, build ad sets, and later review campaign performance. A common safeguard is to have campaigns created in a paused state so they can be reviewed before spending begins.
Research is pulled from the Meta Ad Library
The process begins with competitive research in the Meta Ad Library, filtering for ads that have stayed active for at least two to three months. Long-running ads are treated as evidence of likely conversion success. The system extracts creatives, copy, links, and positioning cues, creating a report of patterns worth imitating rather than starting from a blank page.
Scraping and analysis rely on additional tools
The research workflow used Apify to scrape ads from the Meta library and return them to Claude for analysis. A custom repository called advertising ops then organized the findings, pulled both image and video examples, and generated recommendations for copy and creative direction. The setup was positioned as a “chief marketing officer in a box” for early-stage campaign planning.
Creative generation remains imperfect
After research, the system generated ad images through Higgsfield, producing multiple rounds before selecting stronger options. Initial outputs were described as inconsistent, repetitive, or poorly framed, showing that AI-generated creative still needs human review and likely post-production edits. Even so, the final assets were considered usable enough for testing.
Campaign buildout can be completed end to end
In the demonstrated workflow, the system created one campaign, three ad sets, and three ads per ad set. It applied budget settings, audience choices, website links, and attached creatives after the Meta app was properly published. Earlier attempts created the campaign structure but failed to push assets until app publishing and privacy settings were completed.
Human review remains essential
Even when the automation succeeds, advertisers still need to inspect descriptions, targeting, image quality, and brand fit before launch. The core benefit is speed: AI can build the groundwork, surface options, and monitor results, while the final judgment stays with the operator. That makes the model less about eliminating marketers entirely and more about shifting execution power back to businesses willing to manage campaigns directly.
CONCLUSION
AI tools are making Meta ad execution faster and more accessible, but they work best as force multipliers rather than fully autonomous replacements. The biggest shift is that campaign research, setup, and iteration can now be handled in-house with far less agency dependence.
Explain this