Features · Prompts and answers

Prompt tracking: what each AI answers, question by question

GeoRush prompt tracking shows, for every question your buyers ask, how often each AI engine names you, in a question × engine heatmap. You can also read every original answer word for word, with your brand highlighted, brands in order of appearance, cited sources and the web searches the engine ran before answering.

Updated on

The problem it solves

An overall mention rate tells you whether things are going well, not where. ChatGPT might recommend you for “best CRM for agencies” and never for “cheapest CRM for freelancers”. Perplexity might name you while Gemini doesn't. Without the detail, you can't tell which page to write or which source to work on.

GeoRush splits this into two tabs. Prompts gives you the full picture: every question on every engine. Answers lets you read exactly what the AI said, so you understand why it picked someone else.

How it works

  1. 1

    Add questions the way a buyer would ask them

    Write them as someone would type them into ChatGPT, without your brand name (including it only measures the obvious). Give each one a topic and an intent. GeoRush also suggests question ideas.

  2. 2

    Tag the intent

    Informational (“what is…?”), comparison (“X vs Y”, “best…”), decision (“which one should I pick…?”) or local (“in Chicago…”). Decision questions are the closest to a purchase.

  3. 3

    GeoRush asks them on every sweep

    Each active question runs on all your engines, several times. Pause a question to stop spending credits on it without losing its history.

  4. 4

    Read the heatmap

    Dark cells mean you are named often on that engine; light cells are your gaps. Hover a cell to see how many answers sit behind it and the likely range.

  5. 5

    Drill into the answers

    From any row, open that question's answers and filter by engine or outcome: mentions you, doesn't mention you, cites you, or failed.

What you see in the Prompts tab

  • Question × engine heatmap with the share of answers that name you in each cell.
  • Per question: mention rate across engines, share of answers linking to your site, and your top competitor.
  • Your strongest and weakest question, with a prompt to create the missing content.
  • The searches each engine runs before answering: if your site ranks for them, you are more likely to be cited.
  • Filters by topic and intent, search, and sorting (most mentioned, least mentioned, most cited).

What you see in the Answers tab

  • The full answer, word for word, with your brand highlighted.
  • Brands in the order they appear, with tone and associated themes.
  • Cited sources, labeled as your site, a competitor or a third party.
  • The searches the engine ran before answering.
  • Factual claims about your brand, to verify in Brand perception.
  • How it was collected: official API for chat engines, the real Google interface (SERP) for AI Overviews and AI Mode.
  • A ready prompt to analyze why that answer leaves you out and what to change.

How to read intents

IntentExampleWho usually wins
Informational“How does sales tax work for an online store in Texas?”Whoever explains the topic best.
Comparison“HubSpot vs Pipedrive for a small sales team”Brands that appear in rankings, comparison sites and reviews.
Decision“Which accounting software should I use as a freelancer in the UK?”Highest commercial value: improve these first.
Local“Best family dentist in Austin that takes walk-ins”Local listings and reviews carry the most weight.

Start tracking your questions

Create a free account (100 credits a month) and measure your first questions across every engine. Or try the free check first.

Frequently asked questions

The way a real buyer would ask an AI assistant: natural language, without your brand name. Including your name would only measure the obvious.

Yes. The Answers tab stores every full answer word for word, with your brand highlighted, brands in order, cited sources and the searches the engine ran.

A grid with your questions as rows and AI engines as columns. Each cell shows the percentage of that engine's answers that mention you for that question.

They are marked as failed, excluded from your metrics and don't use credits.

Yes. A paused question is skipped in future sweeps and uses no credits, and its past answers are kept.

Keep reading