— AI Visibility Tracker —
Your SEO dashboard can’t see what ChatGPT is saying about you.
Keyword-rank reports are measuring pages almost nobody visits anymore. The AI Visibility Tracker measures what actually shows up in the answer — across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. Dual-detection with Playwright and Gemini Vision catches citations text scrapers miss.
The Problem
Ranking #3 is a Pyrrhic victory when an AI Overview above you gave the answer.
Traditional SEO tracks the ten blue links. But more and more of the answer is being rendered above the links — in carousels, summaries, and direct citations. Whether you appear there is what matters. And the tools built for tracking page rank simply can’t see it.
Legacy SEO Tools
- Rank tracking on 10 blue links — irrelevant above the fold
- No access to ChatGPT, Claude, Gemini, or Perplexity answer surfaces
- API scraping breaks the moment UI changes
- Can’t detect visual citations (logos, carousels, branded cards)
- Sentiment and competitor co-mentions: invisible
AI Visibility Tracker
- Tracks the answer surface itself — where decisions get made
- Five platforms monitored continuously on real schedules
- Real browsers — no API fragility, no rate-limit whiplash
- Vision layer catches logos, carousel cards, and visual-only mentions
- Sentiment, competitor co-mentions, and bounding-box position per hit
The Engine
Dual-detection: text layer and vision layer, reconciled.
A single detection method will always miss something. Text scrapers miss logos and carousels. Vision alone misses structured citations. We run both on every query and reconcile the results — which is why we catch citations competitors swear aren’t there.
LAYER 1
Playwright — the text layer.
A headless browser fleet runs your target queries on every AI surface as real users. We wait for the answer to fully render, capture the DOM, and extract every citation, link, and mention by parsing the rendered HTML.
- Real browser, not API — resilient to UI changes
- Waits for streaming answers to complete
- Extracts structured citations, source links, inline mentions
- Honors authentication where required
LAYER 2
Gemini Vision — the visual layer.
Every rendered screenshot goes through Gemini Vision. It reads the answer the way a human would — catching logos, branded carousel cards, image-only mentions, and anything where the brand appears visually but not in parsed text.
- Reads logos and branded images embedded in answers
- Detects carousel cards, map pins, and product tiles
- Returns precise bounding-box coordinates for every hit
- Scores sentiment and visual prominence per appearance
Reconciliation — pseudocode
const page = await browser.newPage();
await page.goto(`https://chat.openai.com/?q=${encodeURIComponent(query)}`);
await page.waitForAnswerStream();
// Layer 1 — text
const dom = await page.content();
const textHits = extractCitations(dom, targetBrand);
// Layer 2 — vision
const screenshot = await page.screenshot({ fullPage: true });
const visionHits = await gemini.vision.analyze(screenshot, {
prompt: `Find every mention of "${targetBrand}" — text, logo, card, or image.`,
return: [ "bounding_box", "visual_share", "sentiment", "competitors_visible" ]
});
// Reconcile: a citation must be confirmed by either layer, but we
// use the intersection to score confidence and the union for coverage.
return reconcile(textHits, visionHits);Coverage
Five answer surfaces. All tracked continuously.
These are the five places your customers actually go to ask. Each has its own ranking logic, citation behavior, and competitor mix — so we track each one separately and report them side by side.
AI Overviews
The panel above the ten blue links. If you’re not cited here, the user rarely scrolls.
OpenAI
ChatGPT
The largest active AI product. Tracks both Search mode and default conversational answers.
Anthropic
Claude
Heavy adoption among technical and enterprise buyers. Citations carry outsized influence.
Perplexity
Perplexity
The most citation-transparent surface — every answer lists sources. Ideal canary for entity health.
Gemini
Deep Google integration — strongly influenced by Knowledge Graph and Wikidata signals.
The Output
Citation-grade metrics per query, per surface.
- Bounding-box position — exact pixel coordinates of every brand appearance inside the answer.
- Visual share — what percentage of the rendered answer’s real estate you occupy.
- Competitor co-mentions — who appears alongside you, above you, and below you, per query.
- Sentiment — per-appearance positive/neutral/negative scoring on the text surrounding your mention.
- Citation type — inline link, source footnote, carousel card, logo tile, or image reference.
- Historical trend — citation share over time, with alerts on meaningful drops or rises.
Sample per-query output
{
"query": "best coffee roaster brooklyn",
"captured_at": "2026-04-17T08:32:14Z",
"surfaces": {
"perplexity": {
"cited": true,
"rank": 2,
"bounding_box": { "x": 312, "y": 180, "w": 420, "h": 94 },
"visual_share": 0.184,
"citation_type": "inline_link",
"competitors_above":["partners-coffee"],
"sentiment": 0.82
},
"chatgpt": {
"cited": true,
"rank": 1,
"visual_share": 0.241,
"citation_type": "source_footnote",
"competitors_above":[],
"sentiment": 0.91
},
"google_ai_overview": {
"cited": false,
"opportunity": "high — 3 competitors cited"
}
}
}The Loop
Measurement feeds optimization. Automatically.
The Visibility Tracker isn’t just a reporting tool — it’s a feedback loop. When a query starts citing a competitor instead of you, the tracker tags the gap. Pass 3 of the pipeline sees the signal on its next run and prioritizes the E-E-A-T, sameAs, and authority attributes that the winning citations have. Your graph gets stronger in the directions that actually matter.
Find out who’s winning your queries.
A free scan runs five of your target queries across all five AI surfaces and reports whether you appear, who appears instead, and what the gap looks like. Five minutes. No install.