Understanding the business
EntityRise researches the public presence and builds a Business Profile to confirm.
Measurement methodology
How EntityRise measures AI visibility
EntityRise measures how a brand appears across a confirmed set of buyer questions tested against multiple AI providers, then compares the public evidence environment around the brand with its competitors. Every audit is a point-in-time measurement.
Key facts
The numbers that define the audit
- Buyer questions
- up to 40: 15 High Priority on all 5 providers, up to 25 more on one each
- AI providers
- 5 in every audit
- Observations
- up to 100 per audit
- Benchmark dimensions
- 6, identical plan for everyone
- Comparison states
- 5, set by fixed rules
- Technical readiness
- 4 areas, 30 signals
- Report sections
- 7, plus a methodology note
- Reproducibility
- the set hash in every report
- Price
- $49, one-off
How an audit runs
Every audit has seven stages and two confirmation steps: nothing is measured until you confirm the business profile, and then the buyer topics and questions together. That is deliberate. If the questions do not match what buyers actually ask, the result will be precise and still tell you nothing about your sales.
Buyer topics
From the confirmed profile we derive relevant topics, which you review and edit.
Buyer question set
The set has up to 40 buyer questions. You review them together with the topics, mark the important ones as High Priority and confirm topics and questions in one step before anything runs. The 15 High Priority questions run on all five providers; up to 25 more are spread evenly across the topics and each runs on one provider.
Provider execution
Every High Priority question is tested independently against every configured AI provider, and every additional question against one provider. We do not base the measurement on repeat runs of the same question.
Competitor and evidence research
Commercially substitutable competitors are selected from the confirmed profile and researched across the same public evidence dimensions as the brand.
Deterministic measurement
Scores are computed from recorded observations. The benchmark comparison direction follows fixed scoring rules, not discretionary judgement.
Priorities and report
EntityRise identifies the highest-priority conditions to improve and publishes the audit report with its reproducibility identifiers.
The confirmed question set is sealed with a test-set hash before the first test runs. The report records the Audit ID, the test-set hash, the provider model versions and the generation timestamp, so the same measurement can be repeated and compared later.
Buyer questions
Every audit rests on buyer questions generated from the confirmed profile and topics, which you can review and edit before the run. A set has up to 40 unique questions. The 15 High Priority questions are tested on all five providers, 75 observations, and the report presents them as before. Up to 25 additional questions are spread evenly across your topics and each is asked on one provider, so the audit covers up to twice as many buyer questions within the same maximum of 100 initial observations. When fewer additional questions are available, only those run. The numbers for a specific Audit are recorded in its Audit Integrity section.
Discovery questions
Do not name the brand. They measure organic discovery: whether the brand appears when a buyer asks about the category, a comparison or a purchase without naming any company. Mention rate, active recommendation rate, provider visibility, topic visibility and the AI Visibility Index all come from these observations.
Brand-aware questions
Name the brand explicitly. They measure recognition when the brand is asked about directly, and are shown as recognition evidence. Recognition in a brand-aware question is never counted as organic discovery.
This split is the single biggest difference between a measurement you can trust and a number you can inflate. Add your own brand name to a tracked prompt in a monitoring tool and the visibility chart goes up without anything changing in the market.
Measuring AI visibility
The current standard configuration covers five providers: OpenAI, Claude, Gemini, Grok and Perplexity. The exact model versions used in a given audit are recorded in its Audit Integrity section rather than on this page.
In the standard structure: 15 High Priority questions across 5 providers give 75 observations, and up to 25 additional questions, each on one provider, give up to 25 more, for up to 100 recorded question-provider observations in total. The provider count depends on the configuration; Audit Integrity records what actually ran.
What is recorded for every observation
Mention
The brand appears in the response.
Active Recommendation
The brand is explicitly presented as a recommended solution or choice.
Alternative
The brand appears as an alternative but does not meet the active recommendation class.
Recommendation Position
The observed ordinal position where an ordered recommendation exists.
Citation-Supported
Qualifying citation support exists under the production methodology.
Description Accuracy
How accurately the brand or product is described, where that component is determinate.
AI Visibility Index
The AI Visibility Index is a composite score from 0 to 100, computed as a weighted average of scoreable unbranded discovery observations. The weights are published in every report, so anyone can check the arithmetic.
Index component weights
Observable rates such as mention rate and active recommendation rate are reported separately from the composite index, because they answer different questions. Unavailable observations are excluded from rate denominators when their component is not determinate; they are never silently treated as zero.
Competitive benchmark
Up to three commercially substitutable competitors are selected from the confirmed Business Profile through dedicated market research and substitutability validation. Companies mentioned in AI answers are not automatically treated as benchmark competitors, because an AI naming a company does not make it a commercial substitute for your business.
The client and every selected competitor receive the identical research plan across six intelligence dimensions: category association, third-party authority, recommendation footprint, attribute association, proof and verifiability, and reputation. Research coverage is reported per company, so a comparison built on partial coverage is labelled as such.
Comparative patterns are deterministic and do not establish causality. Where the public record does not settle a comparison, the report says indeterminate instead of guessing a direction.
Five possible comparison statesClient strongerComparableBenchmark strongerMixedIndeterminate
Technical readiness and diagnostic evidence
Alongside the live measurement, EntityRise evaluates the conditions around the brand across four areas and thirty signals: technical accessibility, brand and product understanding, buyer information and proof, and external trust and authority. Each signal carries its own evidence basis, meaning the number of retained evidence records behind it.
Supported
The evaluated scope supports this signal.
Partially supported
Part of the evaluated scope supports it; the share is reported.
Limited evidence
The evaluated scope does not support it, and that is stated plainly.
Indeterminate
Not enough evidence was retained to determine the signal in this audit.
Diagnostic readiness and measured AI visibility are separate concepts. Strong foundations can make evidence easier to discover and verify, but they do not guarantee provider recommendations, and the report never presents them as proven causes.
Strategic priorities
Priority AI visibility fixes describe what needs to improve and why it matters. They do not prescribe tactics: how to carry them out is a decision for your team or agency, not part of the measurement. Priorities are ordered by materiality to measured outcomes, breadth across high-priority questions, benchmark significance and confidence.
Confidence and validation
Every finding carries a published confidence level of high, medium or low, and that level is constrained by backend validation using evidence breadth, benchmark determinacy and counter-evidence. Relevant counter-evidence is retained in the report whenever it materially affects confidence, instead of being dropped to make a finding look stronger. Every piece of research evidence is validated before it is trusted and can be marked validated, partial, unverified or conflicting; weak evidence is kept for review but excluded from automatic conclusions. Published Snapshots are immutable.
Limits of the measurement
These limits are printed in every report, not hidden in a footnote. Without stated boundaries, a measurement is not a measurement.
Audit, not universal ranking
Results apply to the configured test: these questions, providers, market and language.
AI outputs are non-deterministic
Equivalent queries may produce different responses over time or across environments.
Observation is not causation
Competitive or diagnostic differences do not establish why a provider produced a result.
Unavailable is not zero
Failed provider calls are retried after 60 seconds, up to three times. If it still fails, the observation is recorded as unavailable, excluded from rate denominators and disclosed in Audit Integrity. The audit completes with the observations that succeeded.
Insufficient evidence is not absence
Failure to retain public evidence does not prove that the evidence does not exist.
Owned claims are not independent validation
First-party statements are treated differently from independent public evidence.
Provider environments differ
EntityRise uses standardized API-based AI-search tests. Results represent the documented provider environment, model, date, market and language of the audit, and may not reproduce every consumer chat session.
How the data is handled
Audits analyze public page content and public AI answers. Private client information is processed separately, is never shared with competitors and is never used in another company's audit. The details are in the privacy policy.
This methodology exists so it can be checked.
Run the audit and compare the result with what you just read. Every audit carries a hash, model versions and a timestamp.
Frequently asked questions
How many questions does an audit cover?
The standard confirmed set is up to 40 buyer questions: 15 high priority on all five providers and up to 25 more on one provider each. Every question-provider pair is tested independently, which gives up to 100 question-provider observations in the standard five-provider configuration.
Do you run each question multiple times?
No. The measurement is built on breadth across questions and providers rather than repeat runs of the same prompt. Repeating one prompt mostly measures the model's own sampling variance, not your market position. The only repeat is a retry after an error.
Can I add my own buyer questions?
Yes. You review and edit the generated set before the run, and you confirm the final set. After confirmation the set is sealed with a hash for that Audit.
Who approves the buyer questions?
The company being measured, or the agency together with its client. Up to 40 questions for the market, 15 of them marked High Priority. The set is locked before the first test, so the next measurement asks exactly the same questions.
Why does a report say “insufficient evidence” in places?
Because that is the honest result when the public record does not settle a question. Missing signals are not proof that something is absent, and a report that hides this is easy to read and impossible to act on.
Which model versions do you use?
The exact versions used in your audit are recorded in its Audit Integrity section, so a later comparison knows what it is comparing against.
What exactly do I get in the fix plan?
Strategic priorities: what is weak, what the gap against alternatives is, which condition has to improve, why it matters and how strongly the evidence supports it. It is not a task list of the add-this-tag, get-that-many-links kind, because how to carry the priorities out depends on your market and your resources, and your team or agency decides that. The $499 Expert Audit adds human verification and an ordered plan of concrete steps.
Why pay when there are free AI visibility checkers?
Free checkers generate prompts automatically and return a number on a dashboard. You do not know what they asked, you cannot repeat it and you have nothing to hand to anyone. With us you approve the question set yourself and flag up to fifteen as the most important, the set is sealed with a hash, and the result is a report with the measurement, the benchmark and the priorities.
How does the $49 audit differ from Expert Audit?
The $49 audit runs the full research, measures visibility across five providers and says which condition has to improve. Expert Audit adds human verification of every finding, written expert commentary in the report and an ordered plan of concrete steps for your team: what, where and in which order.