A self-service tool that measures where your brand is named in AI answers. This page covers what it is, why it exists, how we work and what we do not do.
Firemark asks a fixed set of buyer questions to AI engines on a schedule and records who each answer names and links. It tracks ChatGPT, Perplexity, Google AI Overviews, Google AI Mode and Gemini, in US English, and shows your share of answers, your rank against competitors and the sites the engines cite.
The MCP server lets Claude and ChatGPT read your results, so you can ask questions in plain language. Read about each engine on its own page, starting with ChatGPT, or see how MCP works.
A fire mark was a metal plaque, stamped with an insurer's emblem and a policy number, that was fixed to the front of an insured building. It dates from the 1700s, before towns ran their own fire services. Streets often had no names and houses no numbers, so the plaque told the insurer's fire brigade which buildings it had promised to save. It also told everyone else who stood behind the building.
An AI answer works the same way. When a buyer asks an assistant who to shortlist, the answer names some brands and passes over the rest, and the brands named are the ones that get the visit. Firemark reads the mark: it records whether each engine names your brand, whether it links your site, and who it names instead, so you can see where you are covered and where you are not.
Buyers now shortlist vendors by asking AI assistants, and the answers name a few brands and skip the rest. Rank trackers report a position for a keyword on a results page. AI answers have no fixed positions, and they differ from one ask to the next, so a single check tells you very little.
What is needed is a fixed set of prompts, asked repeatedly on several engines, scored by stated rules, and reported with the sample size beside every figure. That is the product. We wrote down the rules so that you can judge the numbers instead of taking them on trust.
Tenpoint Labs LLC built Firemark to do that one job and to show its working.
We publish the method and apply it the same way to every account. These are the rules you will see in the product.
Every number follows a published rule. A competitor named fewer than 10 times shows as "fewer than 10". Percentages need at least 20 prompts. A movement is reported only after three comparable runs and a change of 5 points. See the full method.
If an engine fails to answer, the cell is retried for up to 24 hours. An engine that answers under 95% of your prompts in a run is shown as short and left out of that run. An error is never recorded as your brand being absent.
Named means your brand is in the answer text. Linked means your site is among the sources. A name found only inside a list of sources does not count as named.
US English only. Counts are reported as measured, since single answers vary. Each result is a sample of the prompts you track, not a census of everything people ask.
Plans and prices are on the pricing page.
Who runs Firemark and how to reach them.
| Company name | Tenpoint Labs LLC |
|---|---|
| Contact | support@tenpointlabs.com How to reach us |
Start tracking your brand, or open the demo to see a full report first.