Visibility and reputation index
See how the TrustInsider index is calculated: five channels, explicit weights, data coverage and comparisons over time.
ReadThe dashboard
Eight sections. One brand to understand.
In the workspace, every question has a section. The dashboard gathers AI answers, search results, mentions, comparisons and the team’s work into eight views: here you find out what each one holds and which page explains it.
Preview with fictional dataInside the dashboard.
Northstar Audio · fictional brand created for this demonstration
These charts use the same components as the private dashboard. The values are fictional: they show how to read the product, not results achieved by a customer.
Three distinct signals on the same 0–100% scale. Mentions and recommendations are based on 10 answers; citations can be assessed in 8. The measures are not added together.
- Answers with a mention
- 30%3 / 10
- Answers with a recommendation
- 10%1 / 10
- Answers with a citation
- 25%2 / 8
Observed measure
Example: one model, one version of the questions and one market. In the workspace, each group retains its own context.
Example of the chart available for a connected Search Console property. Missing days and values interrupt the line. A zero value is a measurement, distinct from missing data.
Clicksn/a
Read the chart values
| Date | Clicks |
|---|---|
| 12 | |
| 18 | |
| n/a | |
| 10 | |
| 0 | |
| 16 |
From the number to the check
In the private area, open the sample and the original answers. You can ask the assistant to explain the selected measures; action proposals start from reviewed mentions.
Overview
One number to get your bearings. The components to decide.
The Overview opens with the visibility and reputation index: the current value on a 0 to 100 scale, data coverage and the trend over the chosen period. Below, the five channels show which component pushes or holds back the total.
A channel with no valid observations stays empty and the total is not recalculated with redistributed weights: the section says so, so you know what is missing before comparing two periods.
| What you see | Where it comes from | What it does not prove |
|---|---|---|
| Index 0–100 | Weighted average of the five valid components, with the weight version in force. | A market share or a comparison with brands not observed. |
| Coverage | Sum of the weights of the channels with valid observations in the period. | That missing channels are worth zero, or that the total is comparable with periods of different coverage. |
| Trend | Series of the values calculated on closed periods, with the same weight version. | A causal effect of the work done in the meantime. |
AI visibility
The answers customers read before they choose.
The section lists the questions in your panel, the models queried and, for each answer, whether the brand appears, is recommended or receives a linked citation. Every result keeps its date, model and collection mode.
From the same view you open the original answer and the cited sources: it is the starting point for understanding which page of your site convinced the model and which is still missing.
| What you see | Where it comes from | What it does not prove |
|---|---|---|
| Presence per question | Share of valid answers naming the brand, balanced per distinct question. | That the brand is recommended, or that the answer is the one every user sees. |
| Recommendations and citations | Distinct signals from the same reviewed answer, with the cited link. | A volume of users or traffic generated by the answer. |
| Model and mode | Qualified engine, model, API or app mode, language, market and collection date. | Equivalence between an API answer and the consumer interface of the same engine. |
Google and Bing
Search results, in the same market and on the same device.
For each search in the sample you see whether the brand is present in the first organic results on Google and Bing, with domain, position and date. If you have connected Search Console, the same section places the property’s impressions, clicks and average position alongside.
The search sample and the property data answer different questions and stay in two boxes: where the brand appears in the chosen searches, and how much traffic Google records on your pages.
| What you see | Where it comes from | What it does not prove |
|---|---|---|
| Organic presence | Share of sample searches with the brand in the top ten results, per engine. | Search volume, clicks or conversions. |
| Connected property | Impressions, clicks, CTR and average position read from the authorised Search Console. | An observed organic position, nor visibility in AI answers. |
| Site audit | HTML check of a bounded set of URLs, robots and llms.txt, without running JavaScript. | Core Web Vitals, full rendering or exclusion from AI engines. |
Reputation
Mentions, topics and sentiment, with the source alongside.
The section gathers the brand’s authorised mentions and sorts them by source, topic and tone. Every observation keeps its link to the original content, so a negative judgement is read in context before it becomes an action.
Topics and products form a hierarchy reviewed by the team: an opinion about one model is not confused with a judgement on the company, and an unclassified sentiment stays declared as such.
| What you see | Where it comes from | What it does not prove |
|---|---|---|
| Mentions | Authorised content naming the brand, deduplicated and attributed within the period. | Coverage of every existing source, nor the audience of the author. |
| Topics and products | Hierarchy reviewed by the team, with assisted suggestions marked as such. | That a topic equals a product just because it shares the name. |
| Sentiment | Classification per observation, with the “unclassified” state declared. | A judgement of financial reliability or a representative survey. |
Press and social
Articles and conversations, within the active sources.
Here you find the articles and social content collected from the sources enabled for your account, with outlet or platform, date and usage licence. A source that is not activated produces no rows: coverage is declared, never assumed.
In the same view, a news item cited by an AI answer is recognised by its observed URL: it is the link between press and AI visibility that the source graph makes readable.
| What you see | Where it comes from | What it does not prove |
|---|---|---|
| Articles | Outlets collected from the sources contracted and activated for the account, with a usage licence. | A complete press review or the reach of the articles. |
| Social content | Platforms with an authorised adapter, with a declared window and limits. | Listening across every network, or a measure of virality. |
| Source cited by AI | The same URL observed in an AI answer and in an article, with the two moments kept distinct. | That the article caused the citation. |
Competitors
The same criteria, for several brands.
Choose the brands to compare and the section applies the same questions, the same period and the same denominator: presence in AI answers, share of mentions and cited sources for each, with no estimates on data you have not collected.
Saved comparison groups remain reusable in later periods; a brand excluded from the sample is not counted as zero, it stays out of the count.
| What you see | Where it comes from | What it does not prove |
|---|---|---|
| AI presence compared | Same questions, same period and same denominator for every brand in the group. | A market share or customers’ actual preference. |
| Share of mentions | Mentions of each brand over the total of the authorised comparison group. | The weight of sources not collected or excluded from the sample. |
| Saved groups | Cohorts defined by the owner, reusable in later periods. | That a brand excluded from the sample is worth zero: it stays out of the count. |
Actions
From the observed problem to verifiable work.
The Actions section gathers plans, sheets, workflows, campaigns, landing pages and product catalogues: every intervention starts from a selected piece of evidence, goes through a review and keeps the approved version before any metered operation.
A plan declared complete does not change your site on its own: the follow-up check repeats the observation under the same conditions and shows whether the answer has changed.
| What you see | Where it comes from | What it does not prove |
|---|---|---|
| Plans | Proposal tied to selected evidence, review, declared completion and follow-up check. | That a completed plan has already changed the site or the answer. |
| Sheets and workflows | Versioned rows and steps, with dependencies, credit cost and explicit approval. | Automatic execution without confirmation, or an arbitrary connector. |
| Campaigns, landing pages and catalogues | Approved materials, hosted or authorised publication, versions and receipts. | Automatic advertising spend or publication on unconnected channels. |
Reports
A summary to share, with its sources.
In Reports you download the summary of mentions, topics and the work done in the period, in readable or JSON format, and schedule periodic deliveries to your account. Studies published with consent appear in the same section.
Every report states its period, sources and data version: consulting a report already produced costs no credits, regenerating it with new data does.
| What you see | Where it comes from | What it does not prove |
|---|---|---|
| Period report | Mentions, topics and work done, with the period, sources and data version declared. | A single document that automatically merges AI, search and the plan. |
| Scheduled deliveries | Periodic deliveries to your account, with a history you can reopen. | That a scheduled report regenerates data without the expected consumption. |
| Published studies | Aggregate editions published with the owner’s consent, revocable. | A representative sample of the market or a causal inference. |
Secondary settings
Sources, permissions and budget stay one step back.
Settings do not interrupt the analysis: active sources, team permissions, credit budget and language live in a secondary panel you open only when needed. The first screen shows results, or the real state of a collection in progress.
No reading of the dashboard starts a paid collection: every metered operation asks for confirmation with its cost in credits, and results already produced can be consulted without spending more.
- Results and real status
- Eight analytical views
- Sources, permissions, budget, language
Related paths
Choose your next step.
Check your brand’s presence, compare results or choose content improvements: continue with the feature you need.
From question to action
From the question to checking the outcome: context, sources, hypotheses and action plans in the TrustInsider analytical workflow.
ReadInteractive TrustInsider demo
Explore a brand check with fictional data: question, AI answer, sources, proposed improvement and follow-up check.
Read