Feature · Prompt Tracking

Your buyers are asking AI about your category right now.

You just can't see the answers. Add the questions your buyers actually type, and Truffle runs every one of them across every LLM you care about. It records whether you're mentioned, in what position, and how that changed since yesterday.

Start 7-day trial → See pricing

7-day trial · no credit card · cancel anytime

Illustrative data · demo brand "Flowmetrics"

Why the prompt is the new keyword

A buyer used to open Google and type three words. Now the first move of a buying journey is a full sentence typed into a chatbot: "best CRM analytics tool for a mid-size sales team, under $500 a month." What comes back is a shortlist of three brands. That sentence is the new keyword. Google at least handed you a results page to audit and a rank to check. A chatbot answer leaves no trace at all, and no log file anywhere records that the question was asked. If you're not in the answer you don't find out you lost the deal, because you never learn there was one. You cannot optimize for a question you have never seen.

51%

51% of B2B software buyers now start their research with an AI chatbot more often than with Google, up from 29% in April 2025. The same survey puts reliance on AI chatbots for research at 71%, up from 60% seven months earlier. The entry point to the buying journey has moved into the chat window, and the questions typed there are your new keyword set.G2, "Half of B2B Software Buyers Now Start Their Research With AI Chatbots" (survey of 1,076 B2B decision-makers, fielded March 2026), April 2026

What prompt tracking is

Prompt tracking is the practice of running the questions your buyers actually ask against every AI engine on a fixed schedule, and recording the answer each time. The question it answers is "when someone asks this exact question, does the AI say our name, and where in the answer?" Keyword rank never told you that. Each run is a data point. Enough runs make a trend you can act on.

The vocabulary

Prompt: one real buyer question, in the buyer's own words, attached to one persona. It carries intent and constraints inside it, which a keyword never did.

Run: one execution of one prompt against one AI surface at one point in time. Six surfaces on a daily cadence means six runs per prompt per day.

Mention rate: the share of runs for a prompt in which the AI named your brand. Your hit rate on that question.

Position: where your brand lands in the answer's list. Being named third in a shortlist is not the same as being named first.

Gap: a prompt with zero mentions across every surface. A question your buyers ask where a competitor gets the whole answer.

Drift: the change in an answer over time. Models update, sources change, and today's answer is quietly not last week's.

Keyword rank tracking

Ten blue links and one number per keyword, on a results page that looks the same for everyone who searches it. You check position 4 for "crm analytics" and call that visibility. It tells you nothing about the sentence a buyer actually typed, or what a machine said back about you.

Prompt tracking

A full buyer question, run against six AI surfaces, on a cadence you set. You get mention rate, position and the complete answer text per surface per day. You also get the gaps, where the AI answers the question without you in it at all.

How it works in three steps

From a list of buyer questions to a daily record of what every LLM says about you.

Add your prompts

Write the questions your buyers ask, or let Truffle's prompt generator draft them from a persona and your domain, then change whatever doesn't fit. Every prompt belongs to a persona, so buyer-A's questions stay separate from buyer-B's. Plans carry 10, 50 or 200+ prompts per brand.

Truffle runs them across every surface, on your cadence

Each prompt goes to all six tracked surfaces: ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews and Google AI Mode. Daily refresh is the default; dial any prompt set down to every 2 days, every 3 days or weekly to spread spend across the month. Runs are deduplicated and anonymous, so you're benchmarking a clean answer, not your own session history.

Read the matrix, work the gaps

Every prompt gets a row: position per surface, mention rate, and the full stored answer. Sort by mention rate to find the questions you own. Filter to zero-mention prompts to find the ones you don't. Open any run's history to see when an answer changed and what it changed to.

Every prompt × every surface, on one screen

One prompt is an anecdote. The matrix is the argument: the same question asked of every AI surface, every day, with your position in each answer.

The row is the unit of work. Each line is one real buyer question, tagged with the persona who asks it, then scored across all six surfaces. A green ✓ #1 means the AI named Flowmetrics first. A dash means it answered the question without them. The mention-rate bar is simply that row's hit count over six. "Best CRM analytics tool for mid-size sales teams" lands on five of six and reads 83%, with ChatGPT, Gemini and AI Mode all putting Flowmetrics first. "Best dashboards for agency client reporting" reads 17%: one #5 from Gemini, five dashes. Read across a row and you see which surface doesn't know you. Read down a column and you see whether one engine is cold on you generally. Perplexity names Flowmetrics on two of these five questions. The two Google surfaces are not interchangeable either: AI Overviews and AI Mode disagree on three of the five rows, which is why both are tracked instead of collapsed into one Google number. Across the top sit the four stat cards: 212 prompts tracked daily, 58% mention rate, #2.4 average position when mentioned, and 17 prompts with zero mentions. That red 17 is the shortest to-do list you'll get. Seventeen questions your buyers ask where the answer belongs to someone else. Illustrative data for demo brand "Flowmetrics."

What you get with Truffle prompt tracking

Every capability below ships today. Nothing here is a roadmap promise.

10 / 50 / 200+

Prompts per brand, depending on your plan. A focused core set, or full category coverage across every buyer question.

Daily → weekly

Daily refresh by default. Dial down to every 2 days, every 3 days or weekly. Cadence is your spend lever: fewer runs per prompt buys you more prompts on the same budget.

Per-persona

Each persona carries its own prompt set, so buyer-A's questions are tracked and scored separately from buyer-B's instead of averaged into one number.

Full history

Every answer is stored verbatim. Compare today's response to last week's and see drift instantly, including the run where a model quietly dropped you.

6-surface matrix

Position and mention per prompt across ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews and Google AI Mode, so you can tell a brand problem from a single-engine problem.

Gap list

Zero-mention prompts surface as their own count. Filter to them and you have a content brief written by your buyers' actual questions.

Frequently asked questions

How many prompts can I track?

10, 50 or 200+ prompts per brand, depending on the plan. A focused set of 10 covers your core category questions; 200+ gives you full coverage across several personas, markets and the long-tail questions that only a few buyers ask. See the pricing page for the current per-plan limits.

Where do the prompts come from? Do I have to write them all?

You don't. Truffle's prompt generator drafts a realistic set of buyer questions from a persona description and your domain, phrased the way that role would actually ask an AI. You can then change any of them, or write your own from scratch at any time. Most teams start with a generated set, cut the ones that don't match how their buyers talk, and add the handful of questions they hear on sales calls every week.

Which AI models do you run the prompts against?

Six AI surfaces are included on every plan, the trial included: ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews and Google AI Mode. That spread matters because the surfaces disagree. It's common for one engine to name you first on a question while another doesn't mention you at all, and the two Google surfaces can differ from each other. A per-surface matrix tells you whether you have a brand problem across the board or a gap with one specific engine. Grok is available as a paid add-on rather than part of the included set; see the pricing page.

How often does it refresh, and why not real-time?

Daily refresh is the default, and you can dial any prompt set down to every 2 days, every 3 days or weekly. It's a scheduled refresh rather than a live stream, which is deliberate on two counts. First, AI answers are noisy turn to turn. A real-time number would jitter without telling you anything, while a daily run gives you a trend you can act on. Second, every run costs money. Cadence is a spend lever, so you can trade refresh frequency for breadth and track more prompts on the same budget.

Can I see what the AI actually said, or just the score?

The full answer. Every run stores the complete response text, so you can open any prompt on any date and read the exact wording, including which brands it named and which sources it cited. Because the history is kept, you can compare today's answer to last week's and see drift: the day a model stopped mentioning you, or started recommending a competitor ahead of you, along with the answer text that shows why.

How fast do I get the first data?

Your first run starts as soon as you've added prompts, so you'll see initial positions and mention rates the same day. One day gives you a snapshot rather than a trend. Mention rates stabilise as daily runs accumulate, and directional signal you can plan around typically lands within the first week. That's comfortably inside the 7-day trial.

Stop guessing what AI says about you. Ask it every day.

A 7-day trial with the full feature set. Add your buyers' real questions, run them across every AI engine, and read the answer you've never been able to see.

Start 7-day trial → See pricing

No credit card · cancel anytime · every plan tracks ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews and Google AI Mode.