New: signed, verifiable recommendations

IsChatGPT

Recommending You?

AI assistants don't list links. They name a few companies. Xaudient measures how often they name yours, and what to do to be there.

Sample conversation · illustrative

ChatGPT

Answers like this decide who gets shortlisted. Xaudient measures which brands the AI names — and how often it names yours.

> how_it_works

Set up once. Measure every cycle. Act on what you measured.

Three phases, one loop: each Monitoring Cycle shows whether the last round of changes moved your share of AI voice.

next Monitoring Cycle

Click on blocks to know more

01

Who you are for AI

Before measuring, we fix what to measure: identity, markets, competitors and questions.

An approved identity that never changes quietly

Setup fixes your brand, competitors, sectors, intents, customer problems and regions. Every part has a version; after you approve it, every change is flagged in the reports.

Each language and market is its own measure

Italian · Italy is one series, with its own Setup, competitors and Monitoring Cycles. A new market opens a new series, and the numbers of different series are never added together.

Real alternatives, verified one by one

AI assistants and search propose the candidates. Only companies that make the same promise to the customer enter Setup, each with its site measured and who proposed it.

The questions buyers ask, not keywords

Five categories: market, competitor comparisons, pain-driven, buyer journey, geography. Checked before launch, language included. Questions that name no brand measure your organic visibility.

02

What we measure

How often you are named, in what position, with what Tone, and next to whom. Each Monitoring Cycle, with the same method.

One number for how much you count: SAV

SAV is your brand's mentions as a share of all mentions across your confirmed competitive set. WSAV (weighted share of AI voice) is the same share weighted by position in the answer and by tone.

Against competitors, and in what position

Your share of mentions against the leading competitors, and how often you come first, second or third.

Assistant by assistant, question by question

For each question category and each AI assistant: how often you are named, in what position, with what Tone. That shows where you win, and where you disappear.

The real answers, one by one

The most relevant answers from each assistant, with their mentions and the sources they cite. That is where you see why a competitor wins a question.

When AI gets the company wrong, or disagrees, we say so

Answers that attribute another company's facts to you, and assistants that disagree on what to recommend, are flagged. When an assessment comes from AI, it is labelled Provisional.

Who AI and the press associate you with

The companies named next to you in AI answers and in independent press articles. When the press coverage is missing, the report says so.

03

What we do

We make your brand legible to AI, and say who should do what, with signed recommendations.

Is your site legible to AI?

A deterministic audit of your site, with no AI model scoring it: structured data, technical readiness and findability for AI, each scored 0–25 and compared with your competitors' sites.

Then we make it legible

We prepare the AI files (llms.txt, llms-full.txt, a knowledge graph and an ontology) and tell you what to change on the site, such as robots.txt, the sitemap and key pages served without JavaScript, and who does it.

Files an agent can trust

Web content can be built to manipulate the AI agents that read it (Franklin et al., AI Agent Traps, Google DeepMind, 2026). So every AI file we prepare states its provenance and is checksummed against a public manifest, and every recommendation file is signed with a published key.

Wikidata: the sources AI uses to recognise you

We check whether you have a Wikidata entry. We never create or edit one: we show how to get there, starting with at least two independent references, and your team or agency creates it.

Every recommendation says what we measured, why it matters and who acts

Each one follows the same structure: what we measured, why it matters, the action, how the next Monitoring Cycle will measure it, and who owns it.

Signed recommendations

Recommendations are drafted with AI, every number is checked against the report, and the file is signed, so anyone can verify nothing changed after it was issued. Drafts stay drafts until they are signed, and where we don't act, we say why.

04

Over time

Every Monitoring Cycle repeats the same measure, so you see what changes.

The same questions, competitors and method

When questions, competitors or the method change, the report says so instead of comparing different things.

A report that reads without us

In your language, ready to forward, and every statement traces back to a measured value.

How SAV, WSAV and Organic SAV are computed: the definition. Where the method comes from: the manifesto. AI files are checksummed and recommendations are signed. Check them yourself: Verifiable by design.

Ready for Your
First Monitoring Cycle?

All you need is your website and the competitors you know.