What do I do to improve?
The Action Center lines up the work for each of the four questions, page by page and source by source, and checks on the same questions what has changed.
Trust is the new competitive advantage
Does AI trust your brand?
Find out whether AI assistants recommend your company or your competitors. TrustInsider analyses answers, sources and search results to help you identify the content to improve and measure what changes over time.
Start with Starter: analyse AI answers and compare your brand’s presence.
Search and AI trust
SEO, AEO, GEO and Trust: four questions.
Do they find you, use your information, recommend you, trust you: four different questions, with different sources. Follow a single search through the four stages and see what TrustInsider measures at each one.
A sample search about a fictional brand followed in four stages: Google’s results, the direct answer, an assistant’s recommendation and the sources that discuss it, then the comparison with a competitor.
ExampleThe search “Which noise-cancelling headphones for long flights?”
SEO
Do they find you?
- 1Headphones for travel: the in-flight testhifi-magazine.example
- 2Varolde Voyager, travel headphonesvarolde-audio.example
- 3Quessio 2, active noise cancellingilmarde-audio.example · Your website
- What the customer sees
- Anyone searching on Google or Bing sees ten results: your product page is third, behind a magazine and a competitor.
- Why it matters
- A page that does not appear among the top results rarely becomes the source of an answer.
- What TrustInsider measures
- Average position and first page for each question; clicks and impressions from Search Console.
AEO
Do they use your information to answer?
AI Overview
For long flights, adjustable noise cancelling, battery life and a hard case are what matter.
- hifi-magazine.example
- ilmarde-audio.example
- What the customer sees
- Above the links, Google answers on its own with an AI summary and cites the pages it takes the facts from.
- Why it matters
- If the summary cites a magazine rather than your product page, the customer reads your facts in someone else’s words.
- What TrustInsider measures
- How many questions show a direct answer, and how many of those cite your website.
GEO
Do they recommend you?
- ChatGPT
- Gemini
- Perplexity
For long flights I recommend the Ilmarde Quessio 2, comfortable and with a hard case; alternatively, the Varolde Voyager.
- What the customer sees
- Anyone who asks an assistant gets a written answer with two or three brands and, often, a specific recommendation.
- Why it matters
- Being named is not enough: people who have to choose follow the recommendation, which may go to a competitor.
- What TrustInsider measures
- How often AI assistants name you, recommend you and cite your website.
Trust
Do they trust you?
- ReviewsPositive toneComfortable even after ten hours in the air.
- PressMixed toneExcellent isolation, bulky case.
- What the customer sees
- Before buying, the customer reads reviews, articles and discussions, the same sources assistants cite.
- Why it matters
- A negative tone in independent sources makes its way into answers and weighs on the choice more than your own page.
- What TrustInsider measures
- The tone of reviews, press and conversations, aspect by aspect, with the source of every mention.
| Lens | Indicator | Ilmarde Audio | Varolde Audio |
|---|---|---|---|
| SEO | Questions on the first page | 55% | 70% |
| AEO | Direct answers citing the website | 30% | 45% |
| GEO | Answers that recommend you | 40% | 58% |
| Trust | Mentions with a positive tone | 68% | 61% |
Example values for fictional brands, not real data. Each lens has its own base and the four are never added up.
Example with fictional brands: Ilmarde Audio is your brand, Varolde Audio the competitor.
How search is changing
One answer. Everything else, left out.
People can ask ChatGPT, Gemini, Perplexity, Copilot or Claude which product to choose. They get a written answer, with a few names and some sources: anyone missing from that answer is left out of the choice.
This work is called GEO, Generative Engine Optimization: making your brand clear and verifiable for whoever generates the answer. It complements SEO, it does not replace it: for the same questions TrustInsider also measures Google and Bing.
The answer picks only a few.
Image generated with artificial intelligence · Illustrative sceneThe question is longer.
The customer describes needs, budget and context, not a keyword.
The answer selects.
Few brands make it into the recommendation. The others are not named.
Sources matter.
An incomplete product sheet or an outdated page describes your brand poorly.
AI Trust Score
One number. Never on its own.
The AI Trust Score summarises from 0 to 100 how AI assistants’ answers present your brand: whether they name you, whether they recommend you, whether they cite your sources, with what tone, whom you beat and how stable it is.
Visibility · 25%
How often answers name your brand.
Recommendations · 20%
How often they recommend it to people who need to choose.
Authority · 15%
How often your sources are among those cited.
Reputation · 15%
What tone answers use when they talk about you.
Preference · 15%
Who is named when you are compared with a competitor.
Consistency · 10%
How stable the answer remains across repeated runs and, when you measure more than one assistant, from one to the other.
The code counts. The model does not.
The score comes only from counts of answers that are classified, checked by a deterministic judge and reviewed. No number comes from a model.
Declared uncertainty.
Every value comes with the 95% interval, the coverage and the excluded dimensions, each with its reason.
Never an invented zero.
Below 30 observations a dimension stays “insufficient sample”. Below 70% coverage the score stays without a number.
Honest comparisons over time.
Two scores are compared only with the same method and the same weights. A change of method breaks the series.
AI visibility
See who gets recommended.
A customer asks AI for advice. Does your brand appear? Read the answers to questions about your products, by assistant, language and market, and compare the brands named, the recommendations and the pages cited.
Every answer keeps question, assistant, model, date and exposed sources. Mention, recommendation and citation remain three distinct signals, each with its own base.
Competitor analysis
The fairest comparison there is.
Same questions, same period, same denominator for every brand in the group. See where a competitor is named instead of you, with which sources, and open the answers that show it.
From measures to actions
Every page, its intervention.
Measures become a list of tasks, page by page: what to change on every URL of your site, which pages to create and the proposed text, written only from the brand’s approved facts.
Priority
First the interventions that affect the most questions and the most score dimensions.
Page by page
Every URL with its points: structured data, FAQ, specifications, missing content.
Outside your website
External sources to correct, to win over and to respond to.
Execution
You approve. The change reaches where it is needed.
An approved intervention reaches its destination: the CMS, the e-commerce platform, the catalogue feed or the connected ticketing system. There is one cycle: proposed, approved, applied, re-read, under observation, outcome.
For each connection you choose the mode: manual, with approval or automatic for low-risk changes. Manual mode is included in every plan, approval mode in Monitor, Advanced and Enterprise, automatic mode only in Enterprise. Nothing is published without an approved version.
Verifying results
You changed the page. Now compare.
After every intervention the same questions go back to the same assistants. See whether the published page contains the approved text and whether the measure has changed, before and after, under the same conditions.
Done and working are two different pieces of evidence: re-reading the page says the change is there, the next collection says whether the measure has moved. A comparison shows a change, it does not prove its cause: separating the effect requires experiments with a control group.
Ask TrustInsider
Ask. The answer brings the evidence.
Ask why a competitor is recommended or which source is missing. The assistant answers with charts, tables and sources from your workspace, and turns the answer into a proposal to approve.
A question never starts a paid collection: new collections go through the credit quote, and you decide.
Reputation, press and social
What is said about you, with the source.
Reviews, articles and conversations feed assistants’ answers. Collect mentions from enabled sources, read topics and tone by aspect, and always trace back to the original content.
Every voice has its source.
Image generated with artificial intelligence · Illustrative sceneBrand monitoring
Relevant mentions from enabled sources, deduplicated and attributed.
Press and social
Publications collected by an external data provider and connected platforms, with declared coverage.
Perceived trust
Warranty, support and repairs as customers describe them.
Who it is for
Same evidence. Different decisions.
Marketing, communications, e-commerce, product and agencies start from the same answers and the same sources to make different decisions, each with the permissions of their role in the workspace.
Brand and marketing
How AI describes the brand and which content strengthens its presence.
PR and communication
Which articles and sources feed the answers that talk about you.
E-commerce
How your products appear in purchase questions.
Product teams
The specifications that answers describe poorly or cannot find.
Agencies
Separate clients and workspaces, with reports to share.
International groups Enterprise
Several companies, markets and languages, with roles and access.
Plans
Start with Starter. Grow when you need to.
Starter for your first analysis, Monitor to follow your brand, Advanced for more questions and more brands, Enterprise for groups and teams. Every new collection shows the credit cost before confirmation.
Choose a plan for your brand.
A plan to get started, monitor a brand or coordinate several companies. Included credits renew every month, including on an annual subscription.
Save up to 20%
Starter
For a first check of your brand.
€0 / month
€0 / month
Monthly credits: 10 Companies: 1 · Brands: 1 · Users: 1 View previously collected results without using credits.
Start with StarterStart with StarterMonitor
For repeat checks of your brand.
€95 / month
€950 / year
Monthly credits: 100 Companies: 1 · Brands: 1 · Users: 2 View previously collected results without using credits.
Choose MonitorChoose MonitorRecommended
Advanced
For more questions, analyses and brands.
€295 / month
€2,950 / year
Monthly credits: 350 Companies: 1 · Brands: 3 · Users: 5 View previously collected results without using credits.
Choose AdvancedChoose AdvancedEnterprise
For groups with several companies and teams.
€995 / month
€9,550 / year
Monthly credits: 1,200 Companies: 5 · Brands: 20 · Users: 25 View previously collected results without using credits.
Choose EnterpriseChoose EnterprisePrices excluding VAT. Each new operation shows its credit cost before confirmation. No automatic overages; included credits do not roll over.
Before you begin
First check. Clear answers.
What TrustInsider measures, how the AI Trust Score is calculated, how you start and which plan to choose for your work.
What is AI visibility?
It is your brand’s presence in AI assistants’ answers to customer questions: whether it is named, whether it is recommended and which sources are cited. TrustInsider measures it on a question panel, by assistant, language, market and date.
What is GEO?
Generative Engine Optimization: the work on content and sources that makes a brand clear and verifiable for the systems that generate answers. It complements SEO, it does not replace it.
How is the AI Trust Score calculated?
From the answers collected on your panel, classified, checked by a deterministic judge and reviewed. The code counts six measures and combines them with published weights. The value comes with an interval and coverage; if the sample is not enough, it is not shown.
Can I decide what AI says about my brand?
No, and nobody can. You can identify the sources that describe the brand poorly, improve your pages and measure whether the answers change over time.
Do I need to connect my website straight away?
No. Start from your brand name. Search Console, Analytics, the CMS and other services are connected when you need their data, and stay revocable.
How much does it cost?
With Starter you run your first analysis. Monitor, Advanced and Enterprise add credits, brands and features: prices and credits are those of the current price list on the Plans page.