Docs menu

What Orvio Measures

Prompts, engines, answers, mentions and citations, and which report answers which question.

Orvio measures whether AI engines name and cite your brand when buyers ask them for advice. Every report stands on the same few ideas: a prompt, an answer, a mention, a citation.

What it shows

Orvio works the way a buyer does: it asks AI engines questions and reads the answers.

  • Brand. One company whose AI visibility is measured, added with its website, target market and language.
  • Topics. The demand areas the brand wants to be recommended for, such as "WhatsApp marketing automation". Reports are read topic by topic.
  • Prompts. The questions a buyer might ask an AI engine, grouped by topic. Orvio drafts them from your site and your topics, and you can edit them.
  • Engines. Up to three of ChatGPT, AI Overview, AI Mode, Gemini and Perplexity, in the same versions buyers use.
  • Runs. Once a day, Orvio asks every prompt on each of the brand's engines. Every answer is saved with the brands it named and the pages it cited.
  • Competitors. Up to ten tracked competitors are counted in the same answers, so you see who the engines name instead of you.

A run counts what the answers contain. No AI model scores them, so the same answer always gives the same count.

How it is counted

Mentions and citations

A mention means an answer names your brand. Your name, or one of your other names, appears in the answer as a whole word, in any letter case.

A citation means an answer links to a page on your site in its own text. Your site is every page on your domain, plus any other websites you list in Brand Config › General.

  • Mentions and citations are separate facts. An answer can name you without linking to you, or link to your page without naming you. Reports count each on its own.
  • Pages an engine only looked at while searching, without citing them in the answer, are not citations.
  • A page on another company's site that talks about you is not a citation of you. It counts as a citation of that site.

Discovery and branded prompts

A discovery prompt doesn't name your brand, such as "what is the best CRM for small agencies". It shows whether engines surface you unprompted, which is how most buyers meet a brand. The 40 prompts Orvio drafts are all discovery prompts.

A branded prompt names your brand. Orvio spots them by your brand name, so you never mark one by hand. They sit in their own group, About your brand, rather than in a topic. They show how engines describe you when a buyer asks about you by name.

The two are never blended into one score. Visibility counts, such as how many answers name you, use the discovery prompts in your topics.

💡

Counting branded prompts with discovery prompts would blend two questions into one number. Keeping them apart means a visibility count shows what a buyer who doesn't know you yet would see.

Beyond the daily run

  • Sentiment. About every fourth week, a sentiment check asks the engines how they describe your brand and its tracked competitors. It counts what they say by tone and category, with the pages behind each finding. Every week, ChatGPT also rates each brand from 1 to 10.
  • Facts. Every week, Orvio collects the statements engines made about your brand, such as a founding year or a plan's price. You can check them against your own site. This needs the Fact Checks feature.

What "not measured" means

A measured zero is a number: "0 of 12 answers" means the engines were asked and never named you. When something wasn't measured, reports say so in words instead of showing a zero. A prompt might not have run in the period, or an engine might have shown no AI answer at all. Hover the word to see why.

How to read it

Each report answers one question.

  • Overview: where do I stand right now? A summary of mentions, citations, competitors and what changed.
  • Topics: how am I doing topic by topic and prompt by prompt? Your rank among the brands the answers named, and the answers behind it.
  • Visibility Analysis: how does each engine treat me? Topics by engine, the sites the engines cite, and the tone of what they say.
  • Sentiment Analysis: how do the engines describe me next to my competitors? Findings by tone and category, and AI ratings.
  • Fact Checks: what do the engines state about me, and is it true?
  • Actions: what should I change, on my site and elsewhere?
  • Investigations: for one prompt, why does an engine cite a competitor instead of me?

Overview, Topics, Visibility Analysis and Sentiment Analysis read the period chosen in the top bar. See Engines and Runs.