---
title: "Prompt tracking"
url: https://marketingsignals.com/glossary/prompt-tracking
type: Page
---

# Prompt tracking

Prompt tracking explained: why AI answers can't be tracked like rankings, what a reliable tracking setup includes and how to choose prompts that matter.

GEO glossary

## Prompt tracking

Prompt tracking is the regular monitoring of how AI assistants answer a fixed set of prompts relevant to a business, recording whether its brand is mentioned, cited and described correctly. It is the AI search equivalent of rank tracking, but measures frequency across repeated runs rather than one position.

## Why can't AI answers be tracked like rankings?

AI answers change between runs. In research reported in January 2026, Rand Fishkin of SparkToro and Patrick O'Donnell of Gumshoe.ai ran 12 recommendation prompts nearly 3,000 times on ChatGPT, Claude and Google's AI. ChatGPT and Google's AI returned the same list of brands less than 1% of the time. The researchers concluded that a single AI ranking position is misleading, while the share of runs in which a brand appears is more stable.

Prompts also differ from keywords. Google says early AI Mode users in the UK asked questions two or three times the length of traditional queries, so a tracked prompt should read like a real question.

## What does a prompt tracking setup include?

A useful setup holds everything constant except time.

- A prompt set grouped by topic and buying stage, in natural language
- The platforms the audience uses, such as ChatGPT, Perplexity, Gemini, AI Overviews and AI Mode
- Fixed location, language and account settings so runs are comparable
- Several runs per prompt each period, reported as percentages
- Fields for mention, citation, cited URL, competitors named and accuracy

## How to choose prompts

Start from real questions: sales call notes, site search logs, Search Console queries phrased as questions and People Also Ask boxes. Include unbranded discovery prompts, comparisons against named competitors and branded prompts that test what assistants say about the business.

## Example

A payroll software provider might track 'best payroll software for a UK business with 20 staff', 'is payroll software cheaper than an accountant' and a branded prompt asking what it charges. Running each prompt 10 times a month on four platforms gives 120 data points per prompt each quarter.

## Related terms

- [AI share of voice](https://marketingsignals.com/glossary/ai-share-of-voice)
- [LLM visibility](https://marketingsignals.com/glossary/llm-visibility)
- [Google AI Mode](https://marketingsignals.com/glossary/google-ai-mode)
- [Query fan-out](https://marketingsignals.com/glossary/query-fan-out)

## Further reading

- [How to Choose Which Prompts to Track in AI Search](https://marketingsignals.com/insights/how-to-choose-which-prompts-to-track-in-ai-search)
- [How to Measure Performance in AI Search](https://marketingsignals.com/insights/how-to-measure-performance-in-ai-search)

Our [AI SEO & GEO](https://marketingsignals.com/service/ai-search-optimisation) service puts this into practice for brands in the UK and US.

## Sources

- [AI recommendations change with nearly every query: SparkToro, Search Engine Journal (30 January 2026)](https://www.searchenginejournal.com/ai-recommendations-change-with-nearly-every-query-sparktoro/566242)
- [Google Search: Introducing AI Mode in the UK, Google (28 July 2025)](https://blog.google/around-the-globe/google-europe/united-kingdom/ai-mode-search-uk/)

## Ready To Get Found In AI Search?

Get in touch for a friendly, informal chat with our team about how we can help your business thrive online.