AI visibility monitoring
ChatGPT, Perplexity and Google AI Overviews answer your customers’ questions every day, with or without you, right or wrong. CitedBy turns that into numbers: brand visibility measured engine by engine, and an accuracy score checked against your own reference documents.
Same prompt, 10 runs, ChatGPT
Cited 6 / 10
60% ± 15%
One-shot tool: cited ✓
The problem
Your customers ask an AI and read one answer. There is no page two, no click to win. If you are not in the answer, you are not in the decision.
SEO tracks rankings. Brand tracks sentiment. Nobody reports what AI engines say about you, so nobody is accountable for it.
A wrong rate, an outdated condition, a competitor's terms attributed to you. In regulated industries a wrong AI answer is a compliance risk, not a marketing problem.
How it works
Same questions, same method, every cycle. That is what makes the numbers comparable over time.
Generated from your site and your market, validated with your team, then frozen and versioned. Every measurement is comparable to the last.
Each question is asked N times per cycle, on every engine. Critical questions rotate more often than the rest.
Brand and competitor mentions, cited sources, and deviations from your reference documents, extracted from every answer.
Dashboard, critical alerts and a monthly report. After every fix, we re-measure, so you can show the change.
Metrics
| # | Metric | What it measures | Why it matters |
|---|---|---|---|
| M1 | Share of Model | The share of answers to your question set that mention your brand, next to your competitors. | Your presence in AI answers, as one comparable number. |
| M2 | AI Citation Rate | How often engines cite your own domain as a source when they answer. | Being named is not the same as being the source. |
| M3 | Influential Sources | The third-party pages engines rely on most for your questions. | Tells you which pages to fix, update or earn. |
| M4 | Answer Variance | How much answers change across repeated runs of the same question. | Separates a real shift from sampling noise. |
| M5 | Accuracy Score Regulated-industry option | Agreement between AI answers and your reference documents. Automated triage, human validation. | Evidence for compliance, not a guess. |
What you receive
Real structure, sample data. Your brands, your competitors, your questions.
Monthly AI Visibility Report
March 2026
Presence across 240 sampled answers, question set v1
Cited as a source in 12% of answers where the brand appears
“…with Your brand, early repayment is always free, whatever the loan type…”
Reference document says: fees may apply when the loan is repaid during its fixed-rate period. Detected in 3 of 5 runs on Perplexity.
Full report: 8 pages. Share of Model by engine and by theme, influential sources map, variance, open deviations log.
Who it’s for
Accuracy is an obligation, not a preference. Know when an AI engine misstates your products, and keep a record of what you measured and fixed.
White-label measurement for your clients. Replace screenshots of single answers with a method and numbers you can put your name on.
Presence trackers tell you if you appear. CitedBy also tells you if what AI says about you is true — and which sources to fix when it isn’t.
Answer engines generate every response from scratch. Sampling settings, model updates, retrieved sources and small changes in phrasing all shift the output. Ask the same question twice and you can get different brands and different citations. Variance is how these systems work, not a malfunction.
A single answer is one draw from a distribution. It can show your brand when most answers leave it out, or miss it when most answers include it. CitedBy asks each question several times per cycle, on every engine, and reports rates and their variance. A change you see is a change that happened.
Four: ChatGPT, Perplexity, Gemini and Google AI Overviews. Every question in your set is sampled on each of them, on a schedule.
It measures how closely AI answers match your reference documents: product terms, conditions, disclosures, official positions. Deviations are triaged automatically, then validated by a human before they count. It is the option built for regulated industries, where a wrong answer is a compliance risk.
Early access
