AI Competitor Page Monitor with Diff + Slack Alerts in n8n

Eleven nodes, one branching If, one sentence-level differ, one AI summarizer — wired and saved on n8n cloud. The most interesting node is the Code that hashes against $getWorkflowStaticData so you get alerts only when it matters. Real execution captured below.

⚠️ Before activating: the Post to Slack node ships with a placeholder webhook URL (https://hooks.slack.com/services/REPLACE/WITH/YOUR_WEBHOOK). Until you replace it, the diff and AI summarization still run — but the alert step fails with a 404. See Step 5 for the replacement procedure.

What You'll Build

A daily n8n workflow that:

• Runs at 9 AM every day
• Fetches a list of competitor pages you care about (pricing, features, changelogs)
• Computes a SHA-1 hash of each page's stripped content
• Compares against the previous hash stored in n8n workflow static data
• If unchanged: silent no-op (no AI cost)
• If changed: computes sentence-level diff (new sentences, removed sentences) → gpt-5-mini summarizes the material change → Slack Block Kit alert

Skip the Build — Import the Workflow

Prerequisites

Step 1 — Import the Workflow

Create a new workflow. Paste the JSON onto the blank canvas. Eleven nodes load in a branching layout — straight line until the If Changed node, then splits into a "Summarize & Alert" path and a "No Change Log" path.

Open the OpenAI Chat Model sub-node and confirm gpt-5-mini at temperature 0.3.

Step 2 — List the Pages You're Watching

Double-click Competitor URLs. It's a Set node with a single assignment:

"urls": "[\"https://openai.com/pricing\", \"https://www.anthropic.com/pricing\"]"

It's JSON-stringified because n8n expression fields don't render raw arrays cleanly. The next node parses it.

To add URLs, edit the string:

"urls": "[\"https://openai.com/pricing\", \"https://www.anthropic.com/pricing\", \"https://docs.n8n.io/changelog/\", \"https://www.zapier.com/pricing\"]"

No max — the Fan Out node below parallelizes them.

Step 3 — Fetch + Diff Against Static Data

The Fan Out URLs Code node does a JSON.parse + map to produce one item per URL. Then Fetch Page runs once per item in parallel (n8n's automatic parallelism).

The Diff Node Is the Star

Double-click Diff vs Last Snapshot. Key parts:

1. Strip noise, hash what's left:

const stripped = html
  .replace(/<script[^>]*>[\s\S]*?<\/script>/gi, ' ')
  .replace(/<style[^>]*>[\s\S]*?<\/style>/gi, ' ')
  .replace(/<nav[^>]*>[\s\S]*?<\/nav>/gi, ' ')
  .replace(/<footer[^>]*>[\s\S]*?<\/footer>/gi, ' ')
  .replace(/<header[^>]*>[\s\S]*?<\/header>/gi, ' ')
  .replace(/<[^>]+>/g, ' ')
  .replace(/\s+/g, ' ')
  .trim();

const currentHash = crypto.createHash('sha1').update(stripped).digest('hex');

Stripping nav/footer/header kills 95% of false positives. Those regions are where cookie banner IDs, timestamps, and build hashes live.

2. Read previous snapshot from static data:

const wfStatic = $getWorkflowStaticData('global');
wfStatic.snapshots = wfStatic.snapshots || {};
const prev = wfStatic.snapshots[url] || null;
const prevHash = prev?.hash;
const prevText = prev?.text || '';

$getWorkflowStaticData('global') returns an object that persists across workflow runs. Reads are free, writes are persisted automatically when the workflow run completes successfully.

3. Compute sentence-level diff:

const toLines = t => t.split(/(?<=[.!?])\s+/).filter(l => l.length > 30);
const newLines = new Set(toLines(stripped));
const oldLines = new Set(toLines(prevText));
const added = [...newLines].filter(l => !oldLines.has(l)).slice(0, 12);
const removed = [...oldLines].filter(l => !newLines.has(l)).slice(0, 12);

Splits on sentence punctuation (period/exclaim/question), filters tiny fragments (< 30 chars = nav links, dates), takes Set difference. Capped at 12 added + 12 removed to keep the AI prompt bounded.

4. Persist the new snapshot:

wfStatic.snapshots[url] = {
  hash: currentHash,
  text: stripped.slice(0, 20000),
  capturedAt: new Date().toISOString()
};

Stores the current version so tomorrow's run has something to diff against. Cap at 20k chars to avoid bloat.

5. Return everything the downstream branch needs:

return [{ json: {
  url,
  changed: !prevHash || prevHash !== currentHash,
  isFirstRun: !prevHash,
  addedLines: added,
  removedLines: removed
}}];

isFirstRun is crucial — we don't want to alert on the very first run (every URL would fire since there's nothing to compare against).

Step 4 — Branch: Summarize or Skip

Double-click If Changed. The condition:

{{ $json.changed && !$json.isFirstRun }}

Must be true AND not a first run. Output branches:

• True → Summarize the Change → Format Alert → Post to Slack
• False → No Change Log (a noOp node — does nothing, just a visual terminator)

The Summarize the Change Prompt

Double-click Summarize the Change. The prompt is deliberately cautious:

You are a competitive-intelligence analyst. A competitor's page changed
overnight. Summarize what is materially different, based strictly on the
added/removed sentences below. Do not speculate beyond what is shown.

PAGE: {{ $json.url }}

ADDED SENTENCES:
{{ ($json.addedLines || []).map(l => '+ ' + l).join('\n') }}

REMOVED SENTENCES:
{{ ($json.removedLines || []).map(l => '- ' + l).join('\n') }}

RULES
- Output a punchy 3-5 bullet Slack message.
- Lead with the most commercially-interesting change (pricing, feature,
  tier, customer, policy).
- Call out specific numbers, tier names, or feature names from the diff.
- If the change is cosmetic/copy-only, say that in one line and stop.
- No preamble. No "As an AI...". No hashtags.

The rule "If the change is cosmetic/copy-only, say that in one line and stop" prevents the AI from inflating trivial word-swaps into panic-inducing bullet lists.

Step 5 — Format + Send to Slack

Format Alert is a Code node that builds Slack Block Kit JSON with a header, a 2-field section (URL + diff counts), the AI summary, and a context footer.

Post to Slack is an HTTP Request node. Replace the URL placeholder:

https://hooks.slack.com/services/REPLACE/WITH/YOUR_WEBHOOK

…with your real Incoming Webhook URL. See the Multi-Source News Digest guide for the exact steps to create one.

The rendered Slack alert looks like:

🚨 Competitor page changed
─────────────────────────
Page: openai.com/pricing     Diff: +3 new, -2 removed
─────────────────────────
• Flex pricing tier now public: $1/M input tokens for gpt-5-nano
• Enterprise tier removed minimum commitment language
• "Priority processing" replaced with "Batch processing" as tier label

Clickable URL, clear diff counts, specific bullets. Exactly what on-call marketing wants.

Step 6 — Turn On Production Schedule

The Schedule Trigger fires at 9 AM once the workflow is toggled to Active. By default, new workflows are inactive.

Activate the Workflow

Click the Publish button top-right of the canvas, then toggle the workflow to Active. The Schedule Trigger starts ticking. n8n handles the queueing; your workflow runs at 9 AM in the timezone set in your profile (UTC by default).

First Production Run

On the first run after activation, every URL will be classified as isFirstRun: true — so no Slack alert. The second day, every URL gets compared against the first snapshot — changes fire alerts.

Test the pipeline before waiting: click Execute workflow to run it manually. First run captures the baseline. Edit one of the competitor pages (not really — but you can simulate by editing the stored text via a manual static data write), then run again to verify the Slack alert fires.

In our live test against openai.com/pricing and anthropic.com/pricing, the workflow correctly fetched both, computed SHA-1 hashes, and routed both URLs through the No Change Log branch because isFirstRun: true (they had no prior snapshot). On the next scheduled run, any pricing-page change would now fire the AI summarizer + Slack alert.

⚠️ Cloudflare-protected sites note: during our live test, both pricing pages returned a "Enable JavaScript and cookies to continue" interstitial instead of full content. This means the diff captures the interstitial, not the real prices. For sites with bot protection, swap the Fetch Page node for a Browserless or ScrapingBee call (see Extensions) — they return the post-JS HTML.

Extensions: Screenshots, Teams, Notion

Include a Screenshot of the Page

Add a Browserless or ScreenshotOne HTTP call in parallel with Fetch Page. Feed the returned image URL into the Slack Format Alert node as an image block:

{ "type": "image", "image_url": "{{ $json.screenshotUrl }}", "alt_text": "Before / after preview" }

Lets your team see the visual change without clicking through.

Post to Teams Instead

Swap the Post to Slack Webhook node for an HTTP Request to Teams' Incoming Webhook URL. Convert Block Kit to Adaptive Card JSON in the Format Alert node. See adaptivecards.io for schema.

Archive Every Snapshot to Notion

After Diff vs Last Snapshot, add a Notion Append Database Row node. Create a database with columns: URL, Captured At, Changed (bool), Diff Summary, Full Text. You get a versioned archive of every competitor page — great for quarterly competitive reviews.

Conditional Summarization by Importance

Add another If node between Fetch Page and Diff: if URL contains "/pricing" or "/changelog", always alert. For other pages, alert only when > 5 sentences changed. Reduces noise on heavy content-marketing pages that change copy daily.

What's Next

Pair this with AI Lead Enrichment to automatically qualify leads from companies whose pricing just changed, or Multi-Source News Digest to combine competitive changes with industry news into one morning digest.