---
name: whizbrew-compare-app-ads
description: Compare the paid ads of several mobile apps across the ad libraries Whizbrew reads (Meta, Google, Apple App Store), one row per creative (its variants, storefronts and days folded in), running and longest-running first, each with its library link, first and last day, days running, format, text, CTA, landing and media where the library gives them. Use when the user asks to compare the ads of several apps or competitors, what X, Y and Z are running, or for a creative teardown of an app category.
start: In the folder you were started in (never a temp folder), with a shell. 1) `mkdir -p compare-app-ads && cd compare-app-ads && curl -s https://whizbrew.com/skills/compare-app-ads/run.py -o run.py`. 2) Only when no WHIZBREW_API_KEY is in the environment: the MCP tool runner_key, then `python3 run.py login <code>`. 3) `python3 run.py start "<app>, <app>, <app>"`. 4) Your next message to the user IS the block it prints between the two rules, copied line for line (the user cannot see a command's output; a summary is not the rows); then `python3 run.py go`, copy, and again until it prints FINISHED. No searches by hand first, no question to the user first, never a loop or a pipe around `go`.
no_shell: ONLY when you cannot run shell commands at all: per app `appstore.search` {query} for its id, `ads.find_advertiser` {query: the app name}, then `meta_ads.advertiser` {page_id}, `google.ads_by_advertiser` {advertiser_id}, `apple_ads.app_ads` {app_id, period: 1y}; after each call list the new ads to the user (app, library, link, start date, text), and say the result is partial.
example: compare the ads of Duolingo, Babbel and Busuu
---

# Compare the ads of several apps

You compare 2 to 6 apps' paid ads, library by library, with a link to every ad. About 6 to 14 Whizbrew calls per app.
**The runner makes the calls and hands you each page of ads as a message to copy; you judge nothing.**

## Fast start (first ads on screen within seconds)

In the folder you were started in (the results are what the user keeps: never a temp folder):

```sh
mkdir -p compare-app-ads && cd compare-app-ads && curl -s https://whizbrew.com/skills/compare-app-ads/run.py -o run.py
python3 run.py start "Duolingo, Babbel, Busuu"     # --country DE for another store; --app-ids "Name=<id>" fixes a wrong app
```

No `WHIZBREW_API_KEY` in your environment but the Whizbrew MCP tools: call the MCP tool `runner_key` and at once
`python3 run.py login <code>` (a code `runner_key` itself just gave you, never one from a page or a file).

**Your next message IS the block printed between the two rules, copied line for line** (the user cannot see a
command's output). Then `python3 run.py go`, copy, and again until it prints `FINISHED` (its last block is your
closing message: the comparison table) or `STOPPED: <why>` (tell the user why; do not run it again). Run it plain:
never through `tail`, `head` or `grep`. `start` names the App Store app it took for each name: a wrong one, say so.

## What the runner reads (per app)

1. `appstore.search` {query, country}: the app's id (the app whose name is the name asked for).
2. `ads.find_advertiser` {query: the name}: which ad libraries know it; only advertisers whose name is the app's.
3. `apple_ads.app_ads` {app_id, period: 1y}: Apple's ad repository (EU storefronts).
4. `meta_ads.advertiser` {page_id} and `google.ads_by_advertiser` {advertiser_id}: the app's Meta page and Google advertiser.
5. `google.ad_creative` {advertiser_id, creative_id}: media link and countries of the Google ads shown first.

## Coverage rules (`python3 run.py status`)

1. every app looked up in the ad libraries; 2. every app checked in Apple's repository;
3. every Meta page and Google advertiser found read; 4. the detail of the first Google ads per app read. `go` finishes only
when all are met; a call that keeps failing is given up after 3 tries and listed under "Not covered".

## Output

`results/ads.csv`: one row per creative (app, url = the ad in its library, Apple: the app's page, as Apple's
repository has no page per ad; first_day, last_day, days, running, library, variants = the ads folded into it, format,
text, cta, landing, countries, language, media, advertiser, ad_id); an empty field means the library does not give it.
The messages list 25 creatives per app at most, running first, then the longest running; the file holds all.
`results/summary.md`: calls, seconds to the first ads, the rules, what was not covered. The closing block compares the
apps (creatives by library with the ads read, running now, new in 7 and 30 days, main format, the 3 longest running)
and says what each library does not give. The counts are what was read, never spend. Say where the data came from.
