Track what ChatGPT says about your brand.
Your prompts, put to ChatGPT, Claude, Gemini and Perplexity on a schedule — with every answer kept word for word.
Whole answers, kept on your schedule.
What you put in front of a client is the answer itself, word for word — every run comes back as a ledger of full responses you can open, not a percentage you have to defend.
- Prompts
- The questions you track, each with its engines, cadence and last run.
- Responses
- Every answer collected in full, with per-engine mention rates across runs.
- Sentiment
- How each mention reads — traced back to the sentence it came from.
The engines name their sources. Read the list.
Share of voice tells you that you are losing. The citations tell you to whom, on which question, and from which page.
- Citations
Every URL an engine linked to, split into your pages, competitors and editorial sources.
- Citation Gap
The prompts where a competitor won the citation — exportable as an outreach list.
- Page Intel
The pages that keep winning, crawled and scored, so you can see what they have that you do not.
Where AI answers meet your search data.
See which of your search wins carry over into AI answers, and which do not — your prompts joined to your real Search Console queries on what they mean, not how they are worded.
- Matched by meaning
- Prompts joined to the queries you actually rank for, spelling differences ignored.
- Ranked, never cited
- The queries you already win in search and never get cited on — usually the cheapest work on the list, because the page exists.
- Needs Search Console
- Connect it to the same property; without it the bridge has nothing to join to.
- reports over the stored answers
- 18
- reports over the stored answers
- engines asked on your schedule
- 4
- engines asked on your schedule
- credits with every new account — one full scan
- 100
- credits with every new account — one full scan
- a month where answer tracking starts
- $19
- a month where answer tracking starts
Frequently asked questions
Which engines does Prism ask?
ChatGPT, Claude, Gemini and Perplexity. Three of the four are asked with live web search turned on, so the answers reflect what the engine can find today rather than what it memorised during training. Perplexity always searches the web.
What does a scan actually store?
The complete answer text, not a summary. From it Prism pulls whether your brand was named and how early in the answer, the sentiment around the mention, every other brand that appeared, every URL the engine cited, and any follow-up searches it ran before it replied.
How many prompts can I track?
Ten on Starter, 50 on Pro and 100 on Enterprise, per property. You choose how often they run, and the credit cost of a run is shown before you schedule it.
What happens if an engine fails to answer?
You get the credits back. Prism refunds the run when every answer fails, and refunds the failed portion when only some do.
The rest of the platform
- Search Performance Which queries have been sliding for six months, which of your pages are splitting a query between them, the dated moments Google changed the page it ranks for one. 16 reports over a Search Console warehouse that outlives Google's 16-month window.
- Site Health Severity says how alarming a crawl finding looks. Fix First classifies the same findings by what each one is doing to your traffic: broken while earning, recoverable losses, upside, housekeeping. 13 reports over one crawl.
Ten prompts is enough to find out where you stand.
Write the questions your customers ask, point them at all four engines, and read what comes back word for word.