AI Brand Monitor (MCP)

What does AI Brand Monitor do?

Pass a brand name and get back a structured report on how LLM-search platforms describe it. The actor queries Perplexity, You.com, Phind, and Brave AI in parallel, parses each platform's answer + citations, then computes brand-mention rate, competitor share-of-voice, sentiment per platform, and most-cited source domains. Returns a deterministic LLM-ready markdown report. Zero LLM API cost server-side - we hit the public answer pages directly.

Designed for AI-SEO / brand monitoring teams: track if your brand surfaces when users ask LLM-search engines about your category, who competes against you in the AI answer panel, and which third-party sources LLMs cite.

Why use it from an AI agent?

• Multi-platform in one call - 4 LLM-search engines queried in parallel.
• Deterministic - no LLM call inside, same input today = same metrics tomorrow.
• Citation-aware - returns the exact URLs each platform cites, so you can investigate why your brand isn't surfaced.
• Share of voice - explicit metric for brand vs competitor when you supply both.
• Sentiment per platform - lexicon-based, no LLM, picks up positive/negative skew across the answer text.

Modes

A. Brand Pulse (default)

Default question template: "What do people say about [brand] for [category]?". Use this for routine brand health monitoring.

{
    "mode": "brand_pulse",
    "brand": "Linear",
    "category": "project management"
}

B. Competitor Compare

Asks each platform: "[brand] vs [competitor] for [category]". Returns explicit share-of-voice between the two.

{
    "mode": "competitor_compare",
    "brand": "Linear",
    "competitor": "Asana",
    "category": "project management"
}

C. Custom Query

Ship your exact question. The actor queries each platform with the verbatim string and parses the response.

{
    "mode": "custom_query",
    "brand": "Linear",
    "customQuery": "What's the best project management tool for engineering teams in 2026?"
}

Output

{
    "meta": {
        "brand": "Linear",
        "competitor": "Asana",
        "mode": "competitor_compare",
        "query": "Linear vs Asana for project management",
        "scrapedAt": "2026-05-08T15:00:00Z"
    },
    "intelligence": {
        "platformsQueried": 4,
        "platformsAnswered": 3,
        "brandMentionCount": 3,
        "competitorMentionCount": 2,
        "shareOfVoice": 0.6,
        "totalCitations": 28,
        "uniqueDomains": 19,
        "topCitedDomains": [
            { "domain": "linear.app", "count": 5 },
            { "domain": "asana.com", "count": 4 }
        ],
        "sentimentDistribution": { "positive": 2, "neutral": 1, "negative": 0, "unknown": 1 },
        "perPlatform": [
            { "platform": "perplexity", "answered": true, "brandMentioned": true, "competitorMentioned": true, "sentiment": "positive", "citationsCount": 8 }
        ]
    },
    "responses": [
        {
            "platform": "perplexity",
            "query": "Linear vs Asana for project management",
            "answer": "Linear is generally praised for its speed and modern UI...",
            "citations": [{ "url": "https://linear.app/...", "domain": "linear.app" }],
            "answerLength": 423,
            "fetchedAt": "2026-05-08T15:00:00Z"
        }
    ],
    "llm_ready": {
        "summary": "Linear: queried 4 LLM-search platforms (3 answered)...",
        "markdown_report": "## AI Brand Monitor Report\n..."
    }
}

Pricing (PPE)

Mode	Price per call
All modes	$0.001 actor start + $0.005 brand monitor card = $0.006/call

The Actor uses Apify residential proxy because all 4 platforms sit behind Cloudflare Bot Management. Proxy bandwidth (~100-300 KB per platform call x 4 platforms) adds operator cost ~$0.001-0.002 per call. Net margin remains positive.

Platforms covered

Platform	What we extract	Reliability
Perplexity (perplexity.ai/search)	answer text + citation URLs	High - clean public page
You.com (you.com/search)	AI answer (from __NEXT_DATA__) + result links	Medium - schema changes
Phind (phind.com/search)	answer + citation URLs	Medium - dev-focused, narrower coverage
Brave AI (search.brave.com/search?summary=1)	AI summary + organic result URLs	Medium - summary not always present

ChatGPT and Claude.ai are not covered in v0.1 because their public chat interfaces require auth and have aggressive bot detection. Future v0.2 may add ChatGPT Atlas (shared chat URLs) and Google AI Overviews (via SERP scraping).

Limits & caveats

• Anti-bot variance: any of the 4 platforms can return blocked-page HTML on a given run. The actor never aborts on a single platform's failure - the analyzer treats missing answers as 'no answer' and the report flags which platforms succeeded.
• Sentiment is lexicon-based, not LLM. It picks up direction (positive vs negative) but misses sarcasm and domain-specific jargon.
• Mention detection is exact-word-boundary regex on the brand string. "Linear" matches "Linear" but not "Linear app"-typos.
• Schema drift: each platform's HTML structure changes occasionally. The parsers fall back to og:description meta as a last-resort answer source.

Calling from an MCP agent

Via Apify MCP server (mcp.apify.com) as call-actor with name ai-brand-monitor-mcp. Direct Standby HTTP:

curl -X POST -H 'Content-Type: application/json' \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -d '{"mode":"brand_pulse","brand":"Linear","category":"project management"}' \
  https://<actor-standby-url>

Pairs naturally with

• web-company-intelligence-mcp to enrich the brand's homepage tech stack
• reddit-intelligence-mcp for community sentiment beyond LLM-search platforms
• github-org-intelligence-mcp if the brand is a developer tool

Sources & legality

• All requests hit publicly accessible search/answer pages. No auth tokens, no API keys, no LLM provider billing.
• Browser-like User-Agent, residential IP rotation - mimics a normal user looking up the brand.
• Citation URLs and answer text are returned verbatim from the public page; no derivative work or republication.