competitor-ad-tracker

For marketers and founders: every new Facebook, Instagram and Google ad your competitors launch lands in a Notion database, with a morning digest in Slack or Discord that also flags their long-running ads, the ads they stopped and the new landing pages they send traffic to.

It runs on a GitHub Actions schedule. You edit one YAML file. There is no server to host.

What actually arrives

The first run is a summary per competitor (real output, 12 Nov 2026, as it reads in Slack):

Competitor ads · 12 Nov 2026 · US

Notion · now tracking 93 active ads (Meta 41, Google 52) · 66 running 30+ days
• Google image · started 9 Nov
• Meta video (4 variants) · started 7 Nov · "Meet Your AI-Powered Workspace" → notion.com/enterprise
• Meta dynamic (4 variants) · started 7 Nov · "Knowledge and work. Connected." → notion.com/enterprise
• Meta dynamic (5 variants) · started 7 Nov · "Knowledge and work. Connected." → notion.com/enterprise
• Meta dynamic (7 variants) · started 7 Nov · "Knowledge and work. Connected." → notion.com/enterprise
• Meta dynamic (5 variants) · started 7 Nov · "Knowledge and work. Connected." → notion.com/enterprise
• …and 87 more in Notion
Top landing pages: notion.com/enterprise (5 ads)

Perplexity · now tracking 142 active ads (Meta 68, Google 74) · 75 running 30+ days
• Meta dynamic (2 variants) · started 10 Nov · "Stop paying for AI separately." → perplexity.sng.link/A6awk/lplb
• Google text · started 10 Nov
...

After that, a message arrives only when something changed. A normal day looks like this (built from the same real ads):

Competitor ads · 12 Nov 2026 · US

Notion · 1 stopped
• Stopped · Google text · ran 30 Apr to 5 Nov (189 days)

Perplexity · 4 new · 12 stopped
• New · Meta dynamic (2 variants) · started 10 Nov · "Stop paying for AI separately." → perplexity.sng.link/A6awk/lplb
• New · Google text · started 10 Nov
• New · Google text · started 10 Nov
• …and 13 more in Notion

Wispr Flow · 4 running 30+ days
• 30+ days · Google video · running since 13 Oct (30 days)
• 30+ days · Google video · running since 13 Oct (30 days)
• 30+ days · Google video · running since 13 Oct (30 days)
• …and 1 more in Notion

Every ad label (Meta video, Google text ...) links to the ad in the Meta Ad Library or the Google Ads Transparency Center. Nothing new means no message.

In Notion, every ad is one page in a database you own, so it becomes a swipe file you can filter, sort and comment on:

Column	Example
Name (the title)	Meet Your AI-Powered Workspace
Competitor / Platform / Format	Notion / Meta / Video
Status	Active, Stopped, or Not seen (see How it works)
Started / Last seen / Days running	2026-11-07 / 2026-11-12 / 5
Long-running	checked once the ad passes 30 days
Landing page / CTA	https://notion.com/enterprise / Learn more
Ad text	the full ad copy
Variants / Placements	4 / Facebook, Instagram, Audience Network, WhatsApp, Threads
Ad link / Ad ID	the ad in the public library / its id

Setup in three steps

1. Use this template. Click Use this template > Create a new repository at the top of this page. A private repository works the same way.

2. Add secrets. In your new repository, open Settings > Secrets and variables > Actions > New repository secret:

   Secret	Needed	Where to get it
   SCRAPEUNBLOCKER_KEY	yes	your API key from ScrapeUnblocker
   SLACK_WEBHOOK_URL	one or more of these	a Slack incoming webhook for your channel
   DISCORD_WEBHOOK_URL		Discord channel > Edit Channel > Integrations > Webhooks > New Webhook > Copy URL
   WEBHOOK_URL		any URL that takes a JSON POST (Zapier, Make, n8n, your own endpoint)
   NOTION_TOKEN + NOTION_DATABASE_ID		see Notion below

   Without any destination the digest still shows up on each run's summary page in the Actions tab.

3. Edit config.yml. Replace the three example competitors with yours. For each one, add the Facebook Page ID and the website domain their Google ads point to:

   country: US
   competitors:
     - name: Notion
       meta_page_id: "825120160874613"
       google_domain: notion.com

   To find a Page ID, search the advertiser at facebook.com/ads/library, pick their Page from the suggestions, and copy the number after view_all_page_id= in the address bar.

Then open Actions > Track competitor ads > Run workflow for the first run, or wait for the daily schedule (07:17 UTC). Tick Dry run to see the digest in the run log without sending anything.

Notion

1. Go to notion.so/profile/integrations, create an internal integration and copy its secret. That is NOTION_TOKEN.
2. Create an empty full-page database (for example "Competitor ads"). Open its ... menu > Connections and add your integration.
3. Copy the database link (... > Copy link). The whole link works as NOTION_DATABASE_ID.

The first run adds the columns it needs. Columns you add yourself (Notes, Owner, Tags) are left alone.

How it works

config.yml ──► collect ──────────────► process ──────────────► deliver
               Meta Ad Library          merge into state         Notion database (one page per ad)
               Google Ads Transparency  new / 30+ days /         Slack, Discord, webhook (one digest)
               (1 request each)         stopped / landing pages  Actions run summary
                                              │
                                     state/ads.json (committed)

• Collect. For each competitor, one call to the ScrapeUnblocker Meta Ad Library endpoint (the Page's active ads in your country) and one to the Google Ads Transparency endpoint (ads pointing at their domain, shown in the last lookback_days). Both go through the official scrapeunblocker Node SDK, which retries temporary failures. If one call fails, the other competitors and platforms still run, and nothing is marked stopped because of the error.
• Process.
  • Meta lists every variant of an ad as its own row. Rows of one collation group become one ad with a Variants count, so a 7-variant test is one line, not seven.
  • New: an ad that was not there on earlier runs and started within new_within_days. An older ad seen for the first time (it was beyond max_ads before) goes to Notion quietly.
  • 30+ days: an ad still running after long_running_days. Advertisers rarely keep paying for an ad that does not work, so these are the ones worth studying.
  • Stopped: a Google ad not shown for stopped_after_days, or a Meta ad missing from a complete answer. When an advertiser has more active ads than one answer holds (the API says has_more), a missing ad proves nothing. It is reported as Not seen after unseen_days instead, never as stopped.
  • New landing page: a page on a recently started Meta ad that none of their earlier ads used. Tracking parameters are stripped, so ?utm_source=fb variants are one page.
  • First run: one "now tracking" summary per competitor instead of hundreds of "new" lines.
• Deliver. Notion mirrors the state: missing pages are created, and a page is updated when its ad stops or passes the long-running mark. Chat destinations get one digest per run. Each event is remembered per destination: if Slack is down, Slack gets it on the next run, and Discord is not sent it twice. A failed delivery turns the run red and is retried for retry_days.
• State. Everything lives in state/ads.json. The workflow commits it after each run, so the next run knows what was already delivered, and the git history doubles as a timeline of your competitors' ads. A second run with nothing new sends nothing. --dry-run prints what would be sent and saves nothing.

Running it yourself

You need Node.js 22 or newer.

git clone https://github.com/<you>/competitor-ad-tracker && cd competitor-ad-tracker
npm ci --omit=dev
cp .env.example .env            # then fill in your key and destinations
node --env-file=.env bin/cli.js check            # validates config.yml, prepares Notion
node --env-file=.env bin/cli.js run --dry-run    # prints the digest, sends nothing
node --env-file=.env bin/cli.js run

Cron, every morning at 07:17:

17 7 * * * cd /opt/competitor-ad-tracker && /usr/bin/node --env-file=.env bin/cli.js run >> tracker.log 2>&1

Docker, with the state kept in the mounted folder:

docker run --rm -v "$PWD":/app -w /app node:22-slim \
  sh -c "npm ci --omit=dev && node --env-file=.env bin/cli.js run"

CLI options: run (default) or check; --config <file>, --state <file>, --dry-run, --only <competitor>, --version, --help.

Config reference

Everything is in config.yml. Secrets never go there.

Key	Default	What it does
country	US	Meta Ad Library country and Google Ads Transparency region (2 letters).
competitors[].name		Shown in messages and Notion. Also the key in the state file, so renaming starts that competitor fresh.
competitors[].meta_page_id		Numeric Facebook Page ID. Leave out to skip Meta.
competitors[].google_domain		Website their Google ads point to, e.g. notion.com.
competitors[].google_advertiser		Instead of the domain: an advertiser name or id (AR...) from adstransparency.google.com.
meta.max_ads	200	Active ads read per advertiser (1-500), in one request.
meta.media_type	all	all, image, video or meme.
google.max_results	100	Ads read per advertiser (1-1000). Each block of up to 100 is one request.
google.lookback_days	30	Only ads shown in the last N days.
google.stopped_after_days	3	A Google ad not shown for N days counts as stopped.
google.platform	all	Optional comma list: search, youtube, maps, play, shopping.
signals.new_within_days	7	An ad counts as new when it started within N days.
signals.long_running_days	30	Flag ads still running after N days. 0 turns it off.
signals.stopped	true	Report stopped (and not seen) ads.
signals.landing_pages	true	Report new landing pages.
signals.unseen_days	14	Capped answers: report an ad as not seen after N days. 0 turns it off.
signals.retry_days	3	How long a failed delivery is retried.
delivery.max_items_per_competitor	6	Ads listed per competitor in the digest; the rest are counted.
delivery.notion_max_writes	150	Notion writes per run. A big first run finishes over the next runs.

Cost

ScrapeUnblocker requests per run = for each competitor, 1 Meta Ad Library request plus 1 Google Ads Transparency request per 100 ads in google.max_results.

With the default config.yml (3 competitors, google.max_results: 100): 6 requests per run. The schedule runs once a day, so that is about 180 requests a month. Each run prints its request count. Notion, Slack and Discord calls go to their own APIs and cost nothing here.

With no SCRAPEUNBLOCKER_KEY secret, the workflow prints a notice and exits green without calling anything. That is why this template's own Actions runs do nothing.

Adding another destination

A chat destination is an object with a name (the key in the state file for what it already received), a style (plain, markdown, slack or discord) and an async send(digest, payload) that throws on failure. For example, Google Chat in src/sinks.js:

export class GoogleChatSink {
  constructor(url, { fetchImpl = fetch } = {}) {
    this.name = "google_chat";
    this.style = "plain";
    this.url = url;
    this.fetch = fetchImpl;
  }

  async send(digest) {
    const res = await this.fetch(this.url, {
      method: "POST",
      headers: { "Content-Type": "application/json" },
      body: JSON.stringify({ text: digest }),
    });
    if (!res.ok) throw new DeliveryError(`HTTP ${res.status}`);
  }
}

Register it in destinationsFromEnv() (if (env.GOOGLE_CHAT_WEBHOOK_URL) chat.push(new GoogleChatSink(env.GOOGLE_CHAT_WEBHOOK_URL));), then pass the secret in .github/workflows/run.yml. payload holds the same events as JSON (type, competitor, platform, format, started, days_running, text, landing, link). The webhook destination sends exactly that, if you would rather do the formatting in Zapier, Make or n8n.

Links

• ScrapeUnblocker: the API this runs on
• Documentation
• Meta Ad Library endpoint and Google Ads Transparency endpoint
• Node.js SDK
• Related: website-change-monitor (watch competitors' pricing and changelog pages) and google-rank-tracker (your Google positions against theirs)

The Meta Ad Library and the Google Ads Transparency Center are public transparency tools. This project reads that public ad data and is not affiliated with Meta or Google.

Development

npm ci
npm run lint     # ESLint
npm test         # node:test, offline: the SDK, Slack, Discord and Notion are all faked
npm run dry-run  # a real run that sends and saves nothing (needs SCRAPEUNBLOCKER_KEY)

The code is small and split by job: src/collect.js (API calls and shaping), src/process.js (state merge and events), src/format.js (the digest), src/sinks.js and src/notion.js (destinations), src/state.js, src/run.js and bin/cli.js.

License

MIT. See LICENSE.