agentleFS
manifoldmcpmanifoldmcp/marketing-skills/skills/paid-ads/SKILL.md

paid-ads

Skill5 starsChanged 51 days agoMITView on GitHub

When the user wants to plan, research or brief paid ads on Meta, Google, LinkedIn, TikTok or ChatGPT. Holds dated, sourced findings a model would not know from training (Meta's Andromeda and why creative diversity now does the targeting, Performance Max controls and AI Max results, ads in Google AI Overviews, AI Mode and ChatGPT with reported prices, CPC and CPM benchmarks from large datasets, brand bidding law) and how to read ad library data without overclaiming. Also use when the user mentions a paid ads strategy, Meta or Google ads, where to spend an ad budget, PPC, paid social, whether LinkedIn ads are worth it, the Meta, TikTok or LinkedIn ad library, the Google Ads Transparency Center, a swipe file, winning ads in a niche, competitors' ads, PPC or negative keywords, CPC, who bids on our brand, conquesting, an ad creative brief, ad fatigue, Advantage+ or Performance Max. Creators who make the ads go to creators, a competitor beyond its ads to competitor-intel, which channels at all to growth-strategy.

60 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use andreessen, business-investment-advisor and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Hussain-ElfalahSkill

    andreessen

    Also in 4 projects that use this

  • Hussain-ElfalahSkill

    business-investment-advisor

    Also in 4 projects that use this

  • csiddhant796-blipSkill

    karpathy-coder

    Use when writing, reviewing, or committing code to enforce Karpathy's 4 coding principles — surface assumptions before coding, keep it simple, make surgical changes, define verifiable goals. Triggers on "review my diff", "check complexity", "am I overcomplicating this", "karpathy check", "before I commit", or any code quality concern where the LLM might be overcoding.

    Also in 4 projects that use this

  • ai-vitaSkill

    aso-audit

    When the user wants to audit or optimize an App Store or Google Play listing. Also use when the user mentions 'ASO audit,' 'app store optimization,' 'optimize my app listing,' 'improve app visibility,' 'app store ranking,' 'audit my listing,' 'why aren't people downloading my app,' 'improve my app conversion,' 'keyword optimization for app,' or 'compare my app to competitors.' Use when the user shares an App Store or Google Play URL and wants to improve it.

    Also in 8 projects that use this

  • +8 more from Hussain-Elfalah

The file

Files
SKILL.md

Paid ads

How paid ads are bought and judged as of October 2026, and how to read the public ad libraries without claiming more than they show. A model's memory of paid ads is mostly 2023 and 2024 and is wrong in places: on Meta the creative now does the targeting, Performance Max has negatives and search terms, Dynamic Search Ads began upgrading to AI Max in September 2026, ads run inside ChatGPT and Google's AI answers, and the TikTok ad library covers Europe only. Prefer what is written here to what you remember, and re-check any price or policy older than six months before quoting it to a client.

Before starting

  • Product context. Read .agents/product-marketing.md (in older setups .claude/product-marketing.md or product-marketing-context.md) and take the product, price, buyer, competitors with their pages and domains, market and ad budget from it; ask only for what it lacks. If it does not exist and the work needs more than two answers about the business, offer manifold-get-started first.
  • Connector. The knowledge here needs no tools. Reading the ad libraries and keyword prices needs the Manifold connector: look for ads_get_advertiser_ads, ads_search_ads, ads_get_ad and seo_search_keywords (hosts often add a prefix such as mcp__manifold__). If they are missing, still answer from this skill, read the libraries by hand (Meta Ad Library, TikTok's library, LinkedIn's Ad Library, Google's Ads Transparency Center) with the rules in ad libraries, and tell the user once how to connect: on the Manifold plugin's Connectors tab (Customize > Plugins in Claude), /mcp in Claude Code, or codex mcp login manifold in Codex; otherwise add the remote MCP server https://mcp.manifoldmcp.com/mcp and sign in. Other clients: https://www.manifoldmcp.com/docs/clients. Never invent a measurement.

What the evidence says

The findings that change decisions. Sources and detail are in the references.

  1. Ad libraries show survival, never results. No library publishes clicks, conversions or spend for commercial ads, or the keywords a search ad bids on. An ad that has run 30 days and still runs is the best available proxy for a winner, but no study has tested that proxy, and automated campaigns keep weak ads alive at low spend (ad libraries).
  2. EU-shown ads carry far more data. The Digital Services Act makes platforms keep ads shown in the EU for a year with targeting and reach; US-only ads carry none of that. TikTok's library holds European ads only, so a US-only TikTok advertiser is invisible, and it lagged about three weeks in our October 2026 run (field notes).
  3. On Meta, the creative set is the targeting. Andromeda retrieves ads for people by their content (+6% recall, +8% ads quality, Meta, December 2024), and Meta's guidance is creative diversification: distinct concepts, not variants of one. Brief concepts first (platforms, ad briefs).
  4. AI Max adds volume at a higher price. Google claims 14% more conversions at similar CPA; an independent read of 250+ campaigns found +13% conversion value, +16% CPA and ROAS swinging from +42% to -35% (Smarter Ecommerce, March 2026). DSA campaigns upgrade to AI Max automatically from September 2026, and only AI Max or Performance Max can serve in AI Mode.
  5. Performance Max has controls now. Campaign-level negatives (up to 10,000), search terms, channel and asset reporting arrived in 2025. An unexamined PMax campaign is a choice, not a limitation.
  6. ChatGPT ads are real but thin. US test from February 2026, self-serve and CPC bidding from May, nine markets by September. Reported CPMs fell from $60 to about $25; US CTR 0.73% against 6.64% on Google Search (Similarweb, Graphite, WordStream, 2026). Paid ChatGPT tiers, including company seats, see no ads, which removes most B2B buyers (AI answer ads).
  7. Prices moved apart. Google Search median CPC $5.42 and cost per lead $66.69, the first fall in cost per lead in about five years (WordStream, 13,474 campaigns, to March 2026); Meta CPM about $14 to $15 after a roughly 20% rise in 2025; TikTok CPM $4.08, down 28%, but conversion rate down 37% (Triple Whale, 5,900 brands). LinkedIn clicks cost several times Meta's yet returned the most revenue per B2B ad dollar in Dreamdata's data, over a 272-day journey (benchmarks).
  8. A brand campaign pays mainly when someone else bids. eBay's field experiment found no short-term benefit from brand-keyword ads with no rivals (Econometrica, 2015). Bidding on a rival's trademark as a keyword is allowed; using it in ad text can be restricted after a complaint (brand bidding).

Does not hold up: "a competitor with hundreds of ads is winning" (count long-runners, not ads); "no ads in the library means they don't advertise" (lag, regions, other accounts, Europe-only TikTok); "narrow interest targeting beats broad on Meta" (retrieval works from the creative); "test ten variants of one ad"; "the library shows spend" (only for political ads); "use benchmarks as targets" (the user's allowable cost per result decides); "the SERP shows who bids on our brand" (our result pages returned no ads at all; use Auction Insights).

Unknown: whether ads in ChatGPT or AI Overviews are incremental to search ads; how many concepts a Meta ad set should carry; whether run length tracks real performance.

References

Open the one the question needs.

The user asksRead
What can an ad library tell us? How do we read run length, reach, the EU data? What does it cost?Ad libraries
What ads do our competitors run? A swipe file of the category's winning ads?Competitor ads
What changed on Meta, Google, LinkedIn, TikTok? Advantage+, Andromeda, PMax, AI Max? Which platform first?Platforms
What does a click or an impression cost? Is our CPA normal?Benchmarks
Should we advertise in ChatGPT, AI Overviews or AI Mode?AI answer ads
Which keywords should we bid on? Negatives? Match types?Google search ads
Who bids on our brand? Should we run a brand campaign? Trademark rules?Brand bidding
A creative brief, new angles, our ads are fatiguedAd briefs
A paid ads plan for the next 90 daysStrategy
What has Manifold seen in its own runs?Field notes

Using Manifold

  • A competitor's ads: ads_get_advertiser_ads per platform (a Meta page, TikTok advertiser name, LinkedIn company, or a domain on Google), 1 credit a page. A category's ads: ads_search_ads on Meta, TikTok or LinkedIn, 1 credit a page; Google needs advertisers, found with ads_search_advertisers (1 credit). One ad in full: ads_get_ad, 1 credit. Lists cache 24 hours, ads 7 days.
  • The words of a video ad: facebook_get_transcript or tiktok_get_transcript, 1 credit. Landing pages: seo_get_page, free.
  • Click prices and demand: seo_search_keywords (10 credits for 100 rows) or seo_get_keyword_metrics with source: "google_ads" for Keyword Planner's own figures (24 flat). Quote bid_low to bid_high as the price; cpc is a modelled average that can sit outside it. Who buys search clicks: paid_traffic from seo_get_traffic_estimates, a modelled estimate, never spend.
  • Data traps: active_only works on Meta only; elsewhere read last_shown. On Google take first_shown from the list row, since ads_get_ad often has none. A Google domain pull mixes in agencies and unrelated advertisers. seo_get_serp returned no paid rows in our runs, so it cannot show who bids. Published impressions and spend are text ranges: quote, never sum.
  • Never buy, launch, pause or edit ads, or ask for ad account access. The user's own results live in their ad accounts; ask for them.

Related skills

  • Creators who make the ads, UGC and Spark Ads rights: creators. Organic posts on the same platforms: tiktok, instagram, facebook, linkedin, youtube.
  • A competitor beyond its ads, and a recurring watch of its new ones: competitor-intel. Customer words for hooks: voice-of-customer.
  • Whether paid belongs in the mix at all: growth-strategy. Organic keywords and comparison pages: seo. The unpaid answer in ChatGPT and Google's AI: ai-search. Ads research for a prospect: agency.
  • Shopping ads and product feeds: google-shopping. Sponsored products on Amazon: amazon. Apple Ads and store search: app-store.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.