Prompt Tracking
Prompt tracking is the practice of running a fixed set of buying questions against AI engines on a schedule, recording which businesses each answer names and which sources it cites, so visibility becomes a per-engine trend.
Also known as: AI rank tracking, LLM monitoring, Answer engine monitoring
Prompt tracking is to AI visibility what rank tracking was to SEO. You define the questions that matter, the ones a customer would ask when they're ready to hire a roofer, book a dentist, or pick a restaurant for a birthday dinner, and you run them against each engine on a schedule. Every run records the answer: which businesses were named, in what order, with what language, and which sources were cited. Over days and weeks those records become a trend.
It exists because one-off checks don't work. Ask ChatGPT for a roofer in Omaha once and you get an answer; ask again and it may differ. Engines re-retrieve on every run and models aren't deterministic. A single screenshot proving you were named, or weren't, is an anecdote. Prompt tracking replaces the anecdote with a rate: across all runs of this question on this engine this week, you were named in some percentage of them.
It has to be per engine. Profound found ChatGPT and Perplexity share only 11% of cited domains, and Google AI Overviews and AI Mode cite the same URLs only 13.7% of the time. Your visibility on one engine says little about the others. A good tracking setup runs every question on every engine you care about and reports them separately, so you can see that you're strong in Perplexity and invisible in ChatGPT rather than averaging to a meaningless middle.
The prompts themselves need to be realistic. Nobody types roofer near me into ChatGPT. They type something like who's a reputable roofer in west Omaha for hail damage, and does insurance usually cover it. Prompt sets should reflect the trade, the specific services, the cities and neighborhoods served, and the way people actually phrase urgency, budget, and quality concerns. A tracking set of five generic prompts measures almost nothing.
Tibly generates your prompt set from your trade and service-area cities, runs it daily across ChatGPT, Perplexity, Gemini, Google AI Overviews, Google AI Mode, Microsoft Copilot, and Claude, and records mentions, position, sentiment, and cited sources for you and your competitors in every answer. You can add your own prompts, retire ones that stop mattering, and see per-question results so you know exactly which buying questions you own.
Keep exploring the glossary
- Prompt VolumePrompt volume is the number of times a buying question gets asked across an AI engine in a measurement, and it needs to be high enough, repeated over time and across engines, to produce a share-of-voice reading you can trust.
- Mention Rate (Visibility Rate)Mention rate, also called visibility rate, is the percentage of AI answers to a tracked set of buying questions that name your business, calculated per engine over repeated runs, so it measures how reliably an engine recommends you.
- Competitor Benchmark (AI answers)A competitor benchmark in AI answers is a side-by-side comparison of how often each business in a market gets named for the same buying questions on the same engines, revealing who holds the slots you don't and why.
- AI VisibilityAI visibility is how consistently a business gets named when AI systems like ChatGPT, Gemini, and Perplexity answer a customer's buying question, measured across repeated prompts and multiple engines rather than a single check.
Common questions about Prompt Tracking
- How many prompts do I need to track?
- Enough to cover your main services and cities, phrased the way customers ask. For most local businesses that's a few dozen, not five and not five hundred. What matters more is running each one repeatedly and per engine, since a single run of any prompt is noise.
- How often should prompts be run?
- Daily is the practical standard. Engines re-retrieve sources on every run, so daily runs build a rate you can trust within a week or two and catch shifts when a directory changes or a competitor publishes new pages. Weekly runs work but take longer to separate signal from noise.
- Can I just ask ChatGPT myself instead of tracking prompts?
- You can, and it's a fine first look. But one answer isn't a measurement, and your chat history and location can influence what you see. Systematic tracking runs the same questions repeatedly from a neutral context across every engine, which is what turns an impression into a number.