---
title: "Track your Claude share of voice"
description: "Track whether Claude recommends your product when developers ask what to use. BackTalk asks your buyers' questions on a schedule, on your own Anthropic key, and shows your share of voice against competitors with the sampled answers as receipts."
source: https://backtalk.sh/track/claude
---

# Track your Claude share of voice

Track whether Claude recommends your product when developers ask what to use. BackTalk asks your buyers' questions on a schedule, on your own Anthropic key, and shows your share of voice against competitors with the sampled answers as receipts.

Flat plans from $19/mo · 7-day free trial · runs on your own Anthropic key, no markup. Start a free trial: https://backtalk.sh/login · Anthropic key setup: https://backtalk.sh/docs/anthropic.md

## How it works

1. **Define your topics.** One topic is one buying decision: the questions buyers ask, the names your product goes by, and the competitors to watch.
2. **BackTalk asks on a schedule.** Each prompt is asked several times per probe on your own key, because answers vary run to run. Repeated samples are a signal; one answer is noise.
3. **Read the standings, then the receipts.** The dashboard shows your share of tracked-brand recommendations, the trend, and per-model splits. Every number opens into the actual sampled answers.

## The assistant developers actually ask

Claude has an outsized developer audience: it powers Claude Code and sits inside the editors and terminals where tooling decisions get made. When a developer asks Claude which database, host, or monitoring tool to pick, that answer competes directly with your marketing. Tracking your Claude share of voice shows whether the recommendation is you, and what Claude says about the alternatives.

## What you get

- Your Claude share of voice against the competitors you name, sampled on a schedule
- Per-model splits, so a model swap or a new release shows up as its own line
- The sampled answers as receipts, each brand mention classified: recommended, mentioned, or advised against
- Rank changes and newly recommended competitors filed into the same inbox as your Hacker News and Reddit mentions
- Trend charts with CSV export, and a public REST API plus MCP server over the same data

## Frequently asked questions

### How does BackTalk track Claude share of voice?

You define topics: the buying questions your customers ask, your product's names, and the competitors to watch. BackTalk asks Claude those questions on a schedule, several samples per prompt because answers vary run to run, then extracts which brands were recommended and in what order. Your share of voice is your slice of those recommendations, with the sampled answers kept as receipts.

### Do I need my own Anthropic API key?

Yes. Probes run on your own Anthropic key, saved once in Settings, encrypted at rest, and only decrypted server-side at the moment a probe runs. There is no markup and no metering: you pay Anthropic pennies for the scheduled samples, and BackTalk stays a flat price.

### Which Claude models are probed?

The models you choose. Pick a model when you save the key, and compare several models side by side under "What runs each task" in Settings; the dashboard splits share of voice per model, so a model swap or a new release shows up as its own line.

Claude is one voice. BackTalk compares every model you bring a key for in one scoreboard, next to what developers say about you on Hacker News, Reddit, GitHub and more. How AI visibility works: https://backtalk.sh/docs/ai-visibility.md

Also track: [ChatGPT](https://backtalk.sh/track/chatgpt.md), [Gemini](https://backtalk.sh/track/gemini.md), [Perplexity](https://backtalk.sh/track/perplexity.md), [Grok](https://backtalk.sh/track/grok.md), [DeepSeek](https://backtalk.sh/track/deepseek.md)