Monitoring brand mentions across the web with n8n and AI: Google Alerts, press, forums

A negative mention of your brand on a forum or in trade press can circulate for days before a team member stumbles on it by chance — and by the time that happens, the damage is often done. Conversely, a positive mention in an article or a Reddit thread goes unnoticed when it deserves to be amplified. Unlike customer review monitoring, which covers channels you control (Google Business Profile, Trustpilot, internal forms), brand mention monitoring covers the rest of the web: press, blogs, forums, social media — anywhere someone can talk about you without ever telling you directly. This guide builds an n8n pipeline that centralizes these mentions, eliminates false positives, scores sentiment and urgency with AI, and alerts your team only when it matters.

Why automate this monitoring instead of relying on internal word-of-mouth

A study by Judith Chevalier and Dina Mayzlin published in the Journal of Marketing Research (2006) shows, on book sales at Amazon and Barnes & Noble, that a shift in online reviews directly translates into relative sales for a product — and that the effect of a very negative review outweighs that of an equally strong positive one. That asymmetry alone justifies not letting a negative mention circulate unanswered for days.

Timing matters just as much as the mention itself. Duygu Istanbulluoglu showed in a study published in Computers in Human Behavior (2017) that a shorter response time to a social media complaint increases customer satisfaction, regardless of what the customer actually expects in return — a quick acknowledgment alone already shifts perception. A pipeline that detects a negative mention within minutes rather than days therefore doesn't just save time: it changes how the incident is perceived, both by the person concerned and by anyone reading the thread.

Pipeline architecture

The workflow has five steps: multi-source collection, deduplication, false-positive filtering, AI scoring, then routing by severity.

Step 1 — Set up collection sources

Google Alerts as an RSS feed remains the simplest, free starting point: at google.com/alerts, create one alert per brand name variant (legal name, product name, founder's name if relevant), then under Edit alert, change "Deliver to" from your email address to RSS feed. Each alert generates a feed URL you paste directly into an n8n RSS Feed Trigger node — see our RSS node guide for the basic setup.

Google News as an RSS feed complements press coverage specifically: the URL https://news.google.com/rss/search?q=%22brand+name%22&hl=en&gl=US&ceid=US:en returns an RSS feed filtered on your query, also usable in an RSS Feed Trigger. Links in the feed pass through a news.google.com/rss/articles/... redirect before reaching the original article — add an HTTP Request node with redirects followed if you need the final URL rather than the redirect link.

Both sources share the same limitation: an indexing delay ranging from a few hours to several days depending on the site, and incomplete coverage of forums and communities poorly indexed by Google. The section below shows how to fill that gap.

Step 2 — Deduplication and false-positive filtering

As in our competitive monitoring pipeline, a Supabase table indexed on the mention's URL avoids reprocessing the same article twice if a feed republishes it. Insert the URL on its first appearance; an IF node downstream of the RSS trigger checks its absence before continuing.

False-positive filtering deserves a dedicated AI pass, before any sentiment scoring. A brand name that overlaps with a common word, a common surname, or another company with the same name in a different industry otherwise produces a feed polluted with unrelated mentions. A simple prompt is enough:

Does this text actually mention [Brand name], a company in the
[industry] sector? Answer only true or false.

Text: {{ $json.content }}

A Structured Output Parser forces a boolean output directly usable by an IF node, without having to parse free text.

Step 3 — Sentiment and urgency scoring with AI

On mentions that pass the filter, a second AI call — or the same call, with a richer structured output — assigns a sentiment (positive, neutral, negative) and an urgency score, following the pattern described in our Sentiment Analysis node guide:

{
  "sentiment": "negative",
  "urgency": 4,
  "summary": "A user reports a recurring billing bug on a specialized forum, several replies echo the issue.",
  "source_type": "forum"
}

Urgency doesn't reduce to sentiment alone: an isolated negative mention on a low-traffic blog carries less weight than a neutral mention on a high-traffic outlet that's starting to draw replies. The prompt should make that criterion explicit rather than letting the model infer urgency from sentiment alone.

Step 4 — Routing: immediate alert or weekly digest

A Switch node routes based on the urgency score: above a threshold (4 or 5 out of 5), an immediate Slack alert goes to the relevant channel, following the same pattern as the negative review alert already described for customer reviews:

🔴 Urgent mention detected ({{ $json.source_type }})
Sentiment: {{ $json.sentiment }}
Summary: {{ $json.summary }}
Link: {{ $json.url }}

Below the threshold, the mention is simply logged to Supabase, aggregated into a weekly digest by a second scheduled workflow — the same logic as for competitive monitoring or AI-generated summary reports.

Step 5 — Draft a reply, never publish automatically

On a mention that calls for a public reply (forum, social network), the same AI call can draft a professional response, shown in the Slack message as a proposal to validate — never published directly. It's the same safeguard used for automated review replies: a badly calibrated tone on a public, permanent channel costs more than the time saved by skipping human review. For the most sensitive cases, a human approval flow with Slack buttons pauses the workflow until explicit validation before sending anything.

Going further: closing Google Alerts' delay gap with Firecrawl Search

For a brand that needs faster or broader coverage than Google Alerts provides — particularly on poorly indexed forums — the Search operation of the Firecrawl node lets you run a scheduled web search (every few hours) on the brand name, retrieving the actual content of the results rather than just a link. This isn't a replacement for Google Alerts — the credit cost doesn't justify a permanent scan — but a targeted complement for the windows or sources where Google's indexing latency becomes a problem.

Common pitfalls

• A single feed for every brand name variant: "FlowKit" and "Flow Kit" don't return the same results; create one alert per plausible variant rather than a single overly restrictive query.
• Skipping false-positive filtering: for a common brand name, half the feed can be about something else entirely; without this filter, the team mentally tunes out the alerts within days.
• A poorly calibrated urgency threshold: too low, and every neutral mention triggers a Slack alert until the team ignores the channel; too high, and a negative mention that deserved a reaction gets buried in the weekly digest. Tune it on a week of real results before rolling it out.
• Publishing a generated reply without review: on a forum or social network, a badly calibrated response stays visible indefinitely and can escalate the incident rather than defuse it.

In summary

A brand mention monitoring pipeline in n8n combines free sources (Google Alerts and Google News as RSS feeds) with, if needed, a complementary Firecrawl search, AI-based false-positive filtering, sentiment and urgency scoring via structured output, and routing between an immediate Slack alert and a weekly digest. No single piece is complex on its own — it's the assembly that turns a stream of scattered mentions into an actionable signal, with a response time that, per research on online complaint handling, weighs as heavily on customer satisfaction as the reply itself. The AI Inbox Pack (€79) already provides the AI classification and Slack/Telegram alerting foundation reusable for this pipeline; the RAG Assistant Pack (€119) takes it further by making the mention history queryable through a chatbot instead of just archived.