Complete summary
If a channel or provider with a positive weight is missing, the overall index is not calculated. Available results and weighted coverage remain visible.
Visibility and reputation index
Your brand. An index you can understand.
A number helps you find your bearings. Its components show where to focus. Read AI presence, search, social, press and reputation together, with sources, coverage and explicit criteria.
Demonstration with entirely invented data
Read the number. Open the detail.
Northstar Audio is an invented brand. This view uses the same component as the private dashboard, with three periods and one simulated source per channel: it shows no real provider shares, customer results or published weight version. In the private area, calculation requires authorized and reviewed collections, a compatible comparison group and current documented weights.
66/ 100
95% CrI: 61.5–70.2
Measured
Coverage: 100%
The credible interval measures how uncertain the observed part is; the range allowed by missing data lists the values still possible for what was not observed. Two intervals, two questions: they do not add up.
- AI visibility80Measured · Coverage: 100%
- Search visibility60Measured · Coverage: 100%
- Press visibility50Measured · Coverage: 100%
- Brand reputation70Measured · Coverage: 100%
Channel colours identify the type of data, not a positive or negative judgement.
Platforms and evidence
| Platforms and evidence | Index | Weight | Observed cases / eligible cases | Samples |
|---|---|---|---|---|
| demo | 80 | 100% | 80 / 100 | 100 |
| Platforms and evidence | Index | Weight | Observed cases / eligible cases | Samples |
|---|---|---|---|---|
| demo | 60 | 100% | 60 / 100 | 100 |
| Platforms and evidence | Index | Weight | Observed cases / eligible cases | Samples |
|---|---|---|---|---|
| Observed corpus | 40 | 100% | 40 / 100 | 100 |
| Platforms and evidence | Index | Weight | Observed cases / eligible cases | Samples |
|---|---|---|---|---|
| Observed corpus | 50 | 100% | 50 / 100 | 100 |
| Platforms and evidence | Index | Weight | Observed cases / eligible cases | Samples |
|---|---|---|---|---|
| Observed corpus | 70 | 100% | 70 / 100 | 100 |
Each point is an observed snapshot. Missing observations and methodology changes are not connected. Series recalculated using current criteria.
10/18/2610/20/26
Visibility over time · Measurement details
| Observed period | Index | Coverage | P(increase) · threshold 0.95 | Definition |
|---|---|---|---|---|
| 10/17/26 – 10/18/26 | 51.5 | 100% | visibility-index-v2 · synthetic-demo-only | |
| 10/18/26 – 10/19/26 | 58.8 | 100% | 0.98 | visibility-index-v2 · synthetic-demo-only |
| 10/19/26 – 10/20/26 | 66 | 100% | 0.99 | visibility-index-v2 · synthetic-demo-only |
Measurement details
Proprietary index
One summary. Its reasons.
The index brings together five perspectives on brand visibility and reputation. The method assigns 45% to AI, 20% to search, 15% to social, 10% to press and 10% to reputation. These are algorithmic weights that prioritise AI: they are not market shares or percentages of audience reached.
Read the result alongside channel values, sources, coverage and method version. A value on a 0–100 scale is not a probability of success, a certification or a financial assessment. Financial statements, solvency and creditworthiness assessments do not enter this index.
| Channel | Algorithmic weight | Scope to disclose |
|---|---|---|
| AI answers | 45% | Panel, provider, model and collection mode. |
| Google and Bing search | 20% | Queries, engine, market and device. |
| Social | 15% | Authorised sources and comparison cohort. |
| Press | 10% | Available publications and comparison cohort. |
| Reputation | 10% | Reviewed observations and evaluation criteria. |
What each component measures
Five channels. Five precise questions.
Open a channel to see the observations behind its value. AI presence does not measure sentiment; press mentions do not measure traffic. The summary combines different measures through an editorial choice, not an empirical calibration of commercial success.
| Channel | How it is calculated | What it does not prove |
|---|---|---|
| AI | Average brand presence per distinct question; repetitions are equally weighted within each question. | It is not a recommendation, citation or share of AI users. |
| Search | Share of sampled queries with the brand in the first 10 organic results, balancing repetitions per query. | It is not search volume, clicks or conversions. |
| Social | Observed social mentions of the brand divided by mentions of the authorized comparison group. | It does not automatically cover every platform or measure virality. |
| Press | Observed press mentions of the brand divided by those of the same group, within the press channel only. | It is not a share of all press coverage or an audience measure. |
| Reputation | Favorable observations divided by all classified observations, including neutral ones. The observed collection must be fully reviewed. | It is not financial reliability, solvency or a representative survey. |
How the total is built
A calculation you can follow. Every weight included.
With all valid data available, the index is AI × 0.45 + Search × 0.20 + Social × 0.15 + Press × 0.10 + Reputation × 0.10. Within AI and search, provider weights from the current documented version are applied first.
Invented example: values 80, 60, 40, 50 and 70 produce 66/100. If Social is missing, its 15% weight is not redistributed: the total remains unavailable, coverage is 85% and the possible range is 60–75. An absent or expired weight version prevents calculation; equal shares are not substituted.
- Valid observations
- Providers and channels
- Total and coverage
Coverage and availability
Missing data stays visible.
A channel without valid observations cannot be measured. It does not become zero or mean that a brand is absent. Coverage shows how much of the method is supported by available data; a partial result must remain clearly identified as partial.
Authorised access, licences, markets, period, review and provider availability determine what can be measured. A configured connection does not guarantee complete collections. The summary does not replace the details: before comparing two brands, check that scope and measurement bases are compatible.
Sample base
With complete coverage, a base below the method’s threshold makes the result provisional. AI and search require at least 10 distinct questions per provider: repeating the same prompt does not increase that base.
Missing-data range
Any range shown describes values still possible because of missing data. It is not a statistical confidence interval and does not replace the complete result.
Uncertainty
Uncertainty: two intervals, two meanings.
With method v2 the result carries two intervals that answer two different questions. The 95% credible interval says how uncertain the observed part is: it comes from the counts of each cell, k answers with the brand out of n complete reviewed answers, with the Jeffreys estimate, and propagates up to the index by deterministic Monte Carlo, resampling the questions. The range allowed by missing data says instead which values remain possible for what was not observed.
The two intervals do not add up and do not replace each other: with complete coverage only the first remains; with partial coverage both are shown and the total is not calculated. The seed of the calculation is declared, so the same panel gives the same interval on every machine. In the comparison between periods, the probability that the index has increased is a probability on the observed difference, read against the declared threshold: it is not a causal attribution.
| What you see | How it is calculated | What it does not prove |
|---|---|---|
| 95% CrI next to the value | 2.5% and 97.5% quantiles of 10,000 draws: for each question p ~ Beta(k+½, n−k+½), mean over the resampled questions, then platform and channel weights. | It does not cover unobserved channels and does not measure extractor or review errors. |
| Range allowed by missing data | Observed contribution as the lower bound; upper bound with the maximum of the unobserved weights. | It is not a statistical interval: it does not narrow as answers increase. |
| P(increase) in the trend | Share of draws in which the index of the later period exceeds that of the earlier period, from the cells of both sides with a single seeded stream. | It proves neither a cause nor the effect of an intervention; below the declared threshold it remains an observed change. |
| Answers and engines in the details | Sum of the complete reviewed runs of the AI cells; number of AI platforms with a measurement. | Refusals, errors and incomplete answers are counted separately, never as brand absences. |
Provider weights
The source matters. So does its weight.
Shares used to weight providers must have an identifiable source, date, market, metric and version. A share of visits, users or API usage describes different phenomena: one cannot be substituted for another without disclosure.
Without a verifiable source, no market share is invented. Provider weights remain separate from the weights of the five channels; errors, refusals or missing data do not become negative observations. Actual source availability determines which results can be shown.
- Verifiable source
- Consistent scope
- Published version
Documented shares
Usage-based shares require sources with comparable metrics and scope, recording version, date and expiry. An unavailable or expired source is not replaced with an invented estimate.
Current sources
This method version requires a source retrieved within the past 45 days, covering a period that ended no more than 90 days ago. Future dates, incompatible scopes and expired sources prevent those shares from being used in the calculation.
Comparisons over time
Measure change. Keep the same yardstick.
The historical series is recalculated using the currently published criteria and weights, applied to all comparable periods. The dashboard identifies this mode and the method version: the points do not necessarily represent the scores shown in the past.
Comparisons require compatible periods, questions and scopes. Changing weights can also change values for earlier periods; this variation alone does not demonstrate an improvement or decline in reputation. Separate attribution of the effect of new weights is not available in the historical view.
- Compatible periods
- Current weights
- Recalculated series
Related paths
Choose your next step.
Check your brand’s presence, compare results or choose content improvements: continue with the feature you need.
A transparent method, not a black box
TrustInsider’s criteria for interpreting reputation and trust: provenance, samples, reviews and the limits of conclusions.
ReadFrom question to action
From the question to checking the outcome: context, sources, hypotheses and action plans in the TrustInsider analytical workflow.
ReadFind the right answer
Find guides, FAQs and methodology to clarify indicators, coverage and access, with dedicated routes for data requests.
Read