AI share of voice
Know what AI recommends when developers ask
BackTalk tracks your AI share of voice: it asks Claude, ChatGPT, Gemini, Perplexity and more the buying questions your customers ask, on a schedule and on your own API keys, then shows whether the recommendation is you or a competitor, with the sampled answers kept as receipts.
Flat plans from $19/mo · 7-day free trial · runs on your own API keys, no markup · View as Markdown
Recommendations you never see happen
When a developer asks an AI assistant what to use, the answer is a shortlist you are either on or not, and nobody posts about it afterwards. BackTalk makes that moment measurable: you define topics (the questions, your product's names, the competitors), and every probe asks each question several times per model, because one answer is noise and repeated samples are a signal.
Receipts, not vibes
Every percentage opens into the actual sampled answers, with each brand mention classified as recommended, mentioned, or advised against. Rank changes and newly recommended competitors land in the same inbox as your Hacker News and Reddit mentions, so the conversation that moved an AI answer is usually sitting right next to it.
Frequently asked questions
Which AI models can I track?
ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek, plus open models (Llama, Qwen, Mistral and more) through a Venice key. Several models from one provider can compete side by side, and the dashboard splits share of voice per model.
Do the probes run on my API keys?
Yes. You bring a key per provider, saved once and encrypted at rest. There is no markup and no metering: the scheduled samples cost pennies on your own accounts, and BackTalk stays a flat price.
How often does it probe?
Daily on Pro and Business, weekly on Solo, and you can dial the cadence down to save tokens. Pro and Business also get manual probes for an on-demand check.
Learn more
- AI visibilityTopics, prompts, probe cadence, and how to read the share-of-voice standings.
- Claude (Anthropic)Score mentions with Claude (3 models).
- OpenAIScore mentions with OpenAI (3 models).
More use cases: Question · Bug · Complaint · Competitors · Praise · Digest reports · Shareable recap · Agent-native