Measure the answer, not just the mention.

Generative Pulse is configured around your questions, contexts, facts, and messages. It collects the AI answers people encounter, then makes the evidence behind your AI portrayal inspectable.

Explore the framework See AI portrayal

Configured StudyNorthstar Airways

Business traveler perception

9 unbranded prompts4 audiences5 AI systemsDaily

One stored observationPrompt × model × audience × location × language × run

One portrayal profile, four distinct measures.

Share of Answers tells you whether AI includes a subject. Evaluation, Error Rate, and Narrative tell you what AI says when it does. Together, they define AI portrayal.

Share of Answers

How often does the subject appear?

Presence is the primary measure: the share of eligible unbranded AI answers that name a subject or one of its aliases.

Order of mention, recommendation, and share of answer are available as supporting measures.

Evaluation

How favorably is it evaluated?

AI evaluation applies a structured rubric to compare the target with configured subjects along each evaluative dimension.

Results aggregate relative favorability with equal answer weighting. A response may be positive, negative, neutral, or not assessed when it offers no evaluative signal.

Error Rate

How often are the claims incorrect?

Customer-supplied facts provide the known reference. AI claims are evaluated as correct or incorrect within the context of their response.

The result is calculated at the claim level, with claim-response pairs weighted into the aggregate.

Narrative

Does AI carry the intended narrative?

Customer-defined messages are evaluated against each answer to show whether their content is carried forward, contradicted, or not addressed.

Each answer-message pair is evaluated independently and weighted equally in the aggregate.

Your team defines the Study. Generative Pulse observes what AI says.

Every Study starts with the choices that make a result useful to your organization. The configured context remains attached to each observed answer.

How adaptive monitoring changes a Study

Subjects

The target, competitors, and names or aliases to recognize.

Questions

Unbranded prompts chosen by the team, with recommendations from Generative Pulse.

Contexts

Models, audiences, locations, languages, and collection cadence configured for the Study.

Reference

Customer facts, tracked messages, and evaluative dimensions that remain in customer control.

Compare subjects in the same answer conditions.

Every unbranded prompt is eligible for every configured subject. When Generative Pulse compares a target with its cohort, it holds the prompt, model, audience, location, language, and collection window constant.

Same configured contextOne answer at a time

PromptBest airline for business travel from Seattle?

ContextGemini · frequent business traveler · Seattle, US · English

SubjectsNorthstar Airways · United · Alaska

Different inputs. Clear roles.

The AI answers in a configured Study power measurement. Conversation intelligence and Muck Rack media data explain context and improve recommendations without entering the portrayal score.

AI answer monitoring

Generative Pulse stores recurring AI answers collected through a mix of APIs and consumer interfaces, depending on the model. One observation is a configured prompt, model, audience, location, language, and run.

Feeds SEEN

Conversation intelligence

Millions of conversations per month from an opt-in panel of AI users, across many countries and primarily the United States, help reveal questions and language worth monitoring. The panel contains conversations from 2026 onward.

Informs the Study · does not score it

Muck Rack media data

Generative Pulse is a Muck Rack company. Its direct connection to Muck Rack media data adds context around coverage, outlets, journalists, and links that matter to the team.

Explains sources · does not score portrayal

AI answers vary. The evidence should show it.

Sampling variability is part of the method.

AI systems are probabilistic. Repeated runs are treated as independent observations so a Study can capture response variability rather than relying on a single answer.

Scores stay tied to answer-level evidence.

Every result can be traced back to the answer, configured context, and relevant claim, message, or evaluation behind it.

A result only describes its configured Study.

Models, audiences, locations, languages, prompts, and cadence are chosen by the customer. Results should be read in that defined scope.

Make earned work visible in the story AI tells.

Teams can organize the links and campaigns that matter to them, then inspect how source material is cited, carried forward, and connected to a changing portrayal profile.

01
Cited
The link or source appears in an AI answer.
02
Carried forward
The answer carries the story, fact, or framing forward.
03
Influenced
The source context helps explain the portrayal in that answer.
04
Campaign impact
Tagged work can be viewed alongside changes in the results it was meant to affect.

Product measurement and published research are documented separately.

Each original research article includes its own dataset, calculations, and limitations. This page explains how Generative Pulse measures a customer's configured Study.

Read the research