AI Brand Visibility Monitor — ChatGPT, Perplexity, Copilot, Google AI Overview & AI Mode

Analyst-grade AI brand visibility tracking across 5 major AI search engines, with ground-truth validated data — no subscription, no API keys, no month‑long commitments. Pay $0.30 per query. Each query automatically expands into 15–25 real AI interactions across all engines, so you cover buyer intent, not just three guesses.

Co-designed with a senior SEO / GEO analyst. Every metric, formula, validation layer and output field reflects what a professional brand-visibility report needs to look like — the kind that holds up in a board deck, an agency client review, or a Looker dashboard for a CMO. This is not a one-prompt scraper rebadged as "AI monitoring".

Built for Generative Engine Optimization (GEO), AI Engine Optimization (AEO), SEO agencies, brand managers, content strategists, RevOps teams — and AI agents consuming the data via the Model Context Protocol (MCP) server we're building (see roadmap below).

---

🎯 Why this is different

AI search is replacing classic Google. When a buyer asks ChatGPT, Perplexity, or Google AI Overview "what's the best CRM for a marketing agency?" — does your brand appear? In what position? With what framing (leader, alternative, just-mentioned)? How do competitors perform on the same queries on the same engines?

Most AI‑visibility tools give you one prompt on one or two engines and call it a day. That's a lucky-guess audit. We cover buyer intent. One query you type becomes 3–5 semantically related variants (sourced from Google's People‑Also‑Ask and related searches), each run across 5 engines — so one query buys you 15–25 validated data points instead of one.

On top of that, every metric in the output is ground‑truth validated against the raw AI response text. Brand extraction LLMs routinely hallucinate up to 5.6× more mentions than actually appear. We catch that inflation at three layers before it touches your numbers.

---

💰 Pricing — $0.30 per query, no subscription

Every paid plan on Apify charges the same flat $0.30 per query, with automatic volume discounts via Apify tier:

Apify tier	Price per query	Discount
Free	$0.30	—
Bronze (Starter)	$0.27	−10%
Silver (Scale)	$0.23	−23%
Gold (Business)	$0.18	−40%

Minimum order: 3 queries ($0.90 on Free tier, $0.54 on Gold).

What you actually get for $0.30

One query is not one question with one answer. It's the full fan-out pipeline:

1 query you typed
   └── 3-5 semantically related variants (People-Also-Ask + related searches)
        └── × 5 AI engines (ChatGPT, Perplexity, Copilot, Google AI Overview, AI Mode)
              └── = 15-25 real AI responses
                    └── Each validated against the raw markdown (3-layer ground truth)
                          └── Scored, classified, cross-compared with competitors

15–25 real AI datapoints for $0.30. At the Gold tier that's $0.007–$0.012 per datapoint.

Sample costs

Scenario	Queries	Free tier	Gold tier
Quick brand check	3	$0.90	$0.54
Monthly brand scan	10	$3.00	$1.80
Deep competitive audit	30	$9.00	$5.40
Agency client report (5 clients)	100	$30.00	$12.00
Heavy power user	500	$150.00	$60.00

---

🆚 How we compare to the alternatives

vs. SaaS AI‑visibility subscriptions

	Typical enterprise AI‑visibility SaaS	This actor
Entry price	€85–€425 / month (annual contracts common)	$0.30 per query, pay as you go
Setup	Account, onboarding call, contract	Zero setup — paste brand, run
AI engines covered	3–5	5
Fan‑out per query	None or 1–2 variants	3–5 automatic variants
Ground‑truth validation	Marketing copy says yes, reality varies	3‑layer markdown gating, documented
Data format	Dashboards, rarely API	Structured JSON in your Apify dataset
Kill switch	Cancel before renewal	No recurring bill to cancel

Bottom line: a 10‑query scan here costs $1.20–$3.00. An enterprise subscription covering the same scope costs €85/month, minimum.

vs. DIY using raw OpenAI / Anthropic / Google APIs

	Build it yourself	This actor
Cost per validated datapoint	$0.30–$0.50 (LLM tokens + proxies + orchestration)	$0.005–$0.020
Engineering time	40+ hours (scraping, rotation, schema, validation)	0 hours
API keys to manage	5+	None
Proxy rotation, captcha handling	On you	Handled
Hallucination validation	On you	Built in

vs. other Apify brand‑visibility actors

We cost more per "item" on paper than some bare scrapers. Here's what the per‑item price hides:

What you actually get per scan	Cheapest Apify alternatives ($0.008–$0.10 per item)	This actor ($0.30/query)
AI engines covered	1–3 (often Perplexity only, or ChatGPT only)	5 (ChatGPT, Perplexity, Copilot, Google AI Overview, AI Mode)
Real AI datapoints per unit you pay for	1	15–25
Fan‑out across related queries	No — one prompt, one answer	3–5 variants per query, sourced from Google PAA + related
Ground‑truth validation	No — LLM says brand X was mentioned, you trust it	Yes — literal markdown substring match, 3 layers
Hallucination inflation control	No — up to 5.6× inflated mention rates	Yes — validated numbers only
Real browser sessions vs. API	Often API‑only (different answers than users see)	Real browser sessions
Scoring (AIS) with consistency gating	No	Yes — composite 0–100 AI Visibility Score
Framing enum (leader / compared / …)	No	Yes — 7‑class structured
Gap analysis	No	Yes — 5‑class structured with severity
Entity resolution (dedup of "HubSpot", "HubSpot CRM")	No	Yes — 3‑phase: normalize → prefix → LLM grouping
Per‑platform × per‑brand matrix	No	Yes — full matrix with platform‑local SoV
Owned vs. earned citation split	No	Yes
Sentiment breakdown (positive / neutral / negative)	No or only positive	Full three‑bucket
Top‑cited domains ranking	No	Yes — with engine coverage + citation roles
Output format	Flat CSV or plain text	Structured JSON with typed enums
Cost per validated datapoint	$0.008–$0.10	$0.005–$0.020 (at Gold tier)

Cheaper per‑item actors give you one scrape. We give you a full brand‑visibility audit for the same dollar amount, with data you can actually trust in a board deck.

---

🔬 The technical moat — what makes our data reliable

1. Ground‑Truth Validation (3 layers)

When a downstream LLM is asked "which brands are mentioned in this AI response?" it routinely invents brands that were never actually in the response. In our internal benchmarks the worst offender claimed a brand was mentioned 5.6× more often than it actually appeared in the raw text.

Our fix:

• Layer 1 — markdown evidence. Every mention rate is computed by literal lowercase substring match against the raw response markdown. Not LLM extraction rows. Real text, or nothing.
• Layer 2 — validated metrics. Sentiment, position, prominence are computed only for response rows whose markdown actually contains the brand name. Hallucinated rows are discarded before scoring.
• Layer 3 — consistency from evidence rates. Our consistency score uses markdown evidence rates, not inflated LLM counts.

Every response carries ground_truth_validated: true so you know the gate ran.

2. Query fan‑out that matches real buyer intent

Real buyers don't search "best CRM" once and stop. They ask 3–5 variations — "CRM for small business", "HubSpot vs. Salesforce", "affordable CRM with AI". If your tool checks just the seed query and your brand happens to miss it, your report says you're invisible, and you're wrong.

Our fix: every query you submit goes through Google's People‑Also‑Ask and related‑searches panel. An LLM picks the top 3–5 most semantically related variants. All variants run across all 5 engines.

One query you pay for = 15–25 real AI interactions.

3. Five engines covering every major AI ecosystem

• Google AI Overview — the AI answer box above Google organic results. Most user‑facing AI surface in 2026.
• Google AI Mode — Google's standalone conversational mode, separate from AI Overview.
• Microsoft Copilot — Bing‑grounded, default for 400M+ Windows users.
• OpenAI ChatGPT — the consumer AI, search‑enabled by default in this actor.
• Perplexity — the "AI search engine" of choice for technical buyers.

Want Gemini or Grok too? Pass platforms: [...] explicitly and override the default mix.

4. Entity resolution (no double‑counting)

"HubSpot", "HubSpot CRM", "HubSpot Inc." would triple‑count your mention rate in a naive pipeline. We run a 3‑phase resolution pipeline — L1 normalize → L1.5 prefix merge → L2 LLM grouping — so these all collapse into one canonical brand before scoring.

5. Platform‑local Share of Voice

Per‑platform × per‑brand matrix is computed with tracked‑only denominators — your target plus the competitors you defined, excluding third‑party mentions the AI drops in. SoV of 28% on ChatGPT means 28% of tracked brand mentions on ChatGPT, not 28% of every brand the AI ever mentioned (which would be meaningless).

---

🧑‍💼 Built with a senior SEO / GEO analyst

This actor isn't a "we threw a prompt at ChatGPT and called it brand monitoring" project. The metric set, formulas, validation gates, sort orders, tier semantics and output schema were co-designed with a senior SEO / GEO analyst who runs brand-visibility audits for paying agency clients. Concretely:

• AIS formula (with consistency gating) and 5-class gap analysis were ported from a battle-tested production reporting stack used in the field, then adapted to ground-truth validation.
• Per-platform × per-brand matrix matches the structure agencies already use in client decks — you can drop the JSON straight into Looker / Sheets / a slide and it makes sense without re-shaping.
• Tracked-only Share of Voice is the version that survived analyst scrutiny — global SoV (denominator = every brand the AI ever hallucinated) was rejected as misleading.
• Aggregate strengths / weaknesses and top-cited domains with role classification are pulled into the output because that's what content strategists ask for first when they get a brand audit.

Net result: the JSON we ship is what an analyst would build in Looker after a week of querying raw data, except you get it in 3–15 minutes for $0.30 per query, with hallucination control already applied.

---

📊 What you get in the output

Section	What's inside
Summary	Your brand's AIS, mention rate, share of voice, consistency score, sentiment breakdown, owned / earned citation split, top‑10 aggregate strengths and weaknesses, dominant framing.
Competitors	Same aggregate metrics, one row per tracked competitor.
Per‑platform	Each engine's coverage %, mention rate, AIS, average position, top‑3 rate, sentiment breakdown. Sorted by AIS desc so [0] is your strongest engine.
Per‑platform × per‑brand matrix	One row per (engine × tracked brand). Answers "where am I winning, where am I losing" in one table.
Per‑query	Every (query × engine) cell with framing, sentiment, position, gap class, opportunity type, markdown excerpt, strengths, weaknesses, and full citation list with domain types and roles.
Top cited domains	Ranked list of domains AI engines trust on your topic. Flags your own domain as is_brand_owned: true.
Upgrade CTA (demo and free tiers)	Counts of platforms, queries, domains, matrix rows hidden at this tier.

Full OpenAPI contract is in the source repo.

---

🚀 Input — what you pass in

{
  "brand": "HubSpot",
  "category": "CRM software",
  "competitors": ["Salesforce", "Pipedrive", "Zoho CRM"],
  "language": "us",
  "queries": [
    "best CRM for marketing agencies",
    "HubSpot vs Salesforce for small business",
    "AI features in HubSpot CRM"
  ]
}

• brand — required. The brand you're auditing. Unicode letters supported (e.g. Żabka, Müller, L'Oréal).
• queries — required for paid scans (min 3, max 15). Write them in the language your customers use; each becomes 3–5 fan‑out variants × 5 engines = 15–25 real AI calls. Total cost = N_queries × $0.30. Omit only when tier="demo" (cached preview, no charge).
• category — optional. Plain‑English product/market category (e.g. "CRM software", "opony do samochodu"). Backend treats it as "general" if omitted; paid scans use your queries verbatim and don't strictly need it.
• language — optional. ISO‑3166 country code (lowercase, 2 chars: us, pl, de, fr, es, it, gb, br, jp, ...). Drives DataForSEO geo + locale routing for the AI Overview / AI Mode engines (53 countries supported). Unknown codes silently fall back to US/English. Default: us.
• competitors — optional. Up to 5 brands you want ranked alongside you (matrix + scoring covers them). Each gets the same enriched output: AIS, mention rate, share of voice, top‑3 rate, sentiment trio, dominant framing, plus aggregate strengths/weaknesses and citation_mix per competitor (since v0.5).
• platforms — optional override of the default 5‑engine mix (chatgpt, perplexity, gemini, copilot, ai_overview). Pick from the 7 supported (above + grok, ai_mode).

---

📦 Sample output (real scan, trimmed)

{
  "brand": "Ahrefs",
  "category": "SEO tools",
  "summary": {
    "ais": 74.6,
    "mention_rate": 89.6,
    "share_of_voice": 14.4,
    "consistency_score": 46.7,
    "avg_position": 2.0,
    "top3_rate_pct": 86.8,
    "sentiment_positive_pct": 39.7,
    "sentiment_neutral_pct": 57.2,
    "sentiment_negative_pct": 2.7,
    "dominant_framing": "leader",
    "owned_citation_pct": 9.3,
    "earned_citation_pct": 90.7,
    "total_citations_seen": 2735,
    "aggregate_strengths": ["backlink analysis", "keyword research", "Content Explorer", "Site Explorer"],
    "aggregate_weaknesses": ["no free trial", "expensive", "steep learning curve", "restrictive credit system"],
    "avg_prominence": 87.3
  },
  "competitors": [
    {
      "brand": "Semrush",
      "ais": 67.2, "mention_rate": 82.0, "share_of_voice": 13.6, "top3_rate_pct": 78.4,
      "sentiment_positive_pct": 46.9, "owned_citation_pct": 4.2, "earned_citation_pct": 95.8,
      "aggregate_strengths": ["all-in-one platform", "broader keyword database", "site audit"],
      "aggregate_weaknesses": ["UI clutter", "weaker backlink data than Ahrefs"],
      "citation_mix": {"blog": 38.5, "media": 12.1, "review_site": 8.2, "brand_owned": 4.2, "other": 37.0}
    }
  ],
  "per_platform_per_brand": [
    { "platform": "copilot",    "brand": "Ahrefs",  "ais": 84.7, "mention_rate": 98.0, "share_of_voice": 28.0, "top3_rate_pct": 87.0 },
    { "platform": "ai_mode",    "brand": "Ahrefs",  "ais": 80.9, "mention_rate": 96.0, "share_of_voice": 29.7, "top3_rate_pct": 82.7 },
    { "platform": "chatgpt",    "brand": "Ahrefs",  "ais": 77.7, "mention_rate": 94.0, "share_of_voice": 28.7, "top3_rate_pct": 85.4 },
    { "platform": "perplexity", "brand": "Ahrefs",  "ais": 73.6, "mention_rate": 90.0, "share_of_voice": 26.9, "top3_rate_pct": 93.0 },
    { "platform": "ai_mode",    "brand": "Semrush", "ais": 64.9, "mention_rate": 82.0, "share_of_voice": 25.3, "top3_rate_pct": 77.5 },
    { "platform": "copilot",    "brand": "Semrush", "ais": 62.4, "mention_rate": 78.0, "share_of_voice": 27.4, "top3_rate_pct": 82.1 }
  ]
}

Three insights this actor already hands you for free from the data above:

1. Ahrefs dominates Copilot (AIS 84.7, 98% mention rate) — that's your strongest engine.
2. Perplexity puts you top‑3 in 93% of responses — highest top‑3 rate anywhere. Strong editorial signal on that engine.
3. 11‑point AIS spread across engines (73.6 → 84.7) — your brand is not uniformly visible. Content ops have clear per‑engine targets.

Pay $0.30, get actionable board‑deck material like this.

---

🧭 Who this is for

• SEO & GEO agencies running monthly AI‑visibility audits for clients. Use this to export directly into Looker or a client dashboard.
• B2B SaaS brand managers who need to prove "we won the AI mention race" in a quarterly board deck. Full competitor comparison included.
• Content strategists looking to see exactly which query intents leave them absent from AI answers — the per_query section with gap class and opportunity type is the work order for your editorial calendar.
• RevOps and Sales Ops personalizing cold outreach at scale: 1 query per prospect via Clay or Apify API.
• Freelance SEO consultants who don't want a $99+/month subscription just to run one audit.

---

⏱️ How long does a scan take?

Queries	Approx. runtime	Approx. data points
3	3–5 min	45–75
10	8–15 min	150–250
30	20–40 min	450–750
100	60–120 min	1,500–2,500

Scans are parallel across engines; the bottleneck is slower AI engines (ChatGPT, AI Mode) on heavy load. Your Apify dataset populates the moment aggregation finishes, and PPE events only fire on successful completion — partial / failed scans aren't charged.

---

🧰 FAQ

What exactly counts as 1 query? One topic or question you care about. Behind the scenes we expand it to 3–5 semantic variants (from Google PAA + related) and run each across 5 engines. You pay once; you get 15–25 real AI data points.

Do I need to bring my own OpenAI / Anthropic / Google API keys? No. Everything is included in the $0.30 per query.

What happens if a scan fails mid‑way? Every task is idempotent (tracked by a unique correlation key). If one AI engine throws, we retry automatically. The PPE event fires only on successful completion, so a dead‑ended scan isn't charged.

Can I scan a brand in Polish, German, or another language? Yes. Ground‑truth validation uses Unicode NFC + casefold normalization, so "Żabka", "München", "İstanbul" all match reliably against any normalization form the AI uses.

Can I trust the numbers? Yes, and the system is designed around that question. Every metric except Sentiment is backed by a literal markdown substring match. Sentiment is LLM‑classified but gated by markdown evidence. Every response carries ground_truth_validated: true confirming the validation ran.

Where does the data come from? AI engines are scraped through real browser automation and public SERP APIs. Raw markdown is preserved for every response because hallucination validation needs it.

Is my input private? Brand and category are public information. Results live in your Apify dataset under your account. We don't share or sell scan content.

Can I use this for a client's brand? Yes. The output is a complete, self‑contained report you can deliver as-is to clients.

How does this compare to running 100 scans on a cheaper actor? A cheaper actor at $0.008 per item gives you 100 items. This actor at $0.30 per query gives you 1,500–2,500 validated AI data points across 5 engines with scoring, ground-truth validation, entity resolution, fan-out, and a full competitor matrix. On a per-validated-datapoint basis we are cheaper, not more expensive.

Refund policy? Apify's standard PPE model: you're only billed on successful task completion. Failed or partial scans are not charged.

Will there be an MCP server? Yes — it's the next major channel after the Apify actor. The same ground-truth validated brand-visibility analytics will be exposed via a Model Context Protocol server so Claude Desktop, Cursor, ChatGPT custom GPTs, n8n / Zapier MCP nodes and any other MCP-aware AI agent can call the dataset as a native tool. Pricing model carries over: agents pay per query, no monthly commitment. Roadmap below.

Who designed the metric set? The actor is co-developed with a senior SEO / GEO analyst who runs paid brand-visibility audits for agency clients. The AIS formula, the 5-class gap analysis, the per-platform × per-brand matrix layout, the tracked-only Share of Voice convention, and the citation-domain rankings are all built around what an analyst actually puts in a client deck — not what looks good in a tool screenshot.

---

🗺️ Roadmap

Features we're actively building. No ETAs — we ship when it's good, and changes land behind the per_* fields you already consume (backward compatible).

Coming soon

• MCP server (Model Context Protocol) — a hosted MCP endpoint so any MCP-aware client (Claude Desktop, Cursor, ChatGPT custom GPTs, n8n / Zapier MCP nodes, in-house AI agents) can ask "how does HubSpot perform across all major AI engines vs Salesforce on Q3 buyer-intent queries?" and receive ground-truth validated, structured analytics in a single tool call. Same pricing model — billed per query consumed by the agent.
• Executive summary (LLM narrative) — 2–4 sentence plain-English interpretation of your visibility profile. "Ahrefs dominates Copilot at AIS 84.7 but loses 11 points on Perplexity — invest in Perplexity-native citation sources to close the gap." Co-written with our SEO analyst. Available on every paid tier.

Planned

• Discovered brands — brands AI engines surface organically around your topic, outside the competitors[] list. Catches blind spots.
• Cited pages per owned domain — per-engine citation counts for every URL on your own domain. Tells you which pages to double down on and which to kill.
• Content opportunity map — full dump of People-Also-Ask + related candidates from the fan-out pass, including those the selector didn't pick. Idea-generation surface for content strategists.
• Clay / Zapier / Make.com native integrations — 1-click action nodes.
• Looker / Sheets / BI exports — ready-to-paste dataset templates for analyst-grade reporting.

Shipped recently

• Per‑platform × per‑brand matrix with tracked‑only Share of Voice.
• Owned vs. earned citation split.
• Top‑3 rate and position scoring on flowing-markdown engines (Copilot, AI Mode, AI Overview).
• Aggregate strengths and weaknesses (top-10 per brand).
• Full sentiment breakdown (positive / neutral / negative).
• Preview mode on demo tier with upgrade hints.

---

🏷️ Keywords

AI brand monitoring · ChatGPT brand visibility · Perplexity brand tracking · Copilot brand monitoring · Google AI Overview monitoring · Google AI Mode tracking · AEO (AI Engine Optimization) · GEO (Generative Engine Optimization) · LLM brand tracker · AI SEO · brand mentions in ChatGPT · share of voice AI · AI visibility score · ground truth validation · hallucination-free AI analysis · competitor AI monitoring · AI citation analysis · AI search optimization · SERP AI monitoring · platform-local share of voice · per-engine brand matrix · MCP server brand visibility · Model Context Protocol brand monitoring · MCP brand intelligence · Claude MCP brand visibility · Cursor MCP AI search · ChatGPT MCP integration · agentic brand intelligence · analyst-grade AI brand data · professional AI brand analytics · SEO analyst built tool · GEO analyst tooling · enterprise-grade AI brand monitoring · agency-grade brand visibility · board-deck brand visibility report

---

🤝 Built by

doesaiknow.com — the only brand-AI monitoring stack built around 3-layer Ground-Truth Validation to eliminate LLM hallucination from your data, co-designed with a senior SEO / GEO analyst, and built to ship the same dataset to Apify, MCP-aware AI agents, BI tools and APIs under a single pay-per-query economic model.