Understanding Your AI Visibility Metrics

AI assistants like ChatGPT and Perplexity now answer the questions your customers used to type into Google. Ripenn's job is to tell you how your brand shows up in those answers. Every metric on the dashboard is answering one of two questions: when someone asks an AI a relevant question, does it mention your brand — and when the AI cites web pages to back up what it says, are those pages yours? This page explains each metric the dashboard shows: what it means, how it's computed, what to do when it moves, and whether watching it over time is worth your attention.

First, where the numbers come from. For each question you track (we call them prompts), Ripenn asks several AI engines the question, often several times, and reads every answer. Each individual answer is a run. We currently check four engines: ChatGPT, Perplexity, Claude, and Gemini. One caveat up front: not every engine tells us which web pages it used. Today we can read cited source URLs from ChatGPT and Perplexity, but not yet from Claude or Gemini. So the "are your pages cited?" metrics cover the engines that expose their sources; the "are you mentioned?" metrics cover all four.

One more definition that shapes the headline numbers: organic mentions. If a prompt literally contains your brand name — "Is [YourBrand] good for sensitive skin?" — the AI mentioning you is not a win you earned; the question handed it to you. We call these brand-aware prompts and leave them out of your headline visibility numbers, so the score reflects how often you come up when customers aren't already asking for you by name.

Are you in the answer?

Overall Visibility (Mention Rate)

What it means. Out of all the times we asked your prompts, the share of answers that mentioned your brand. This is your headline visibility number.

How we work it out. We count every answer that named your brand, divide by the total number of answers, and turn it into a percentage. Brand-name prompts are excluded (see "organic" above).

How to use it. This is your "am I on the shelf?" number. A low mention rate means AI assistants rarely bring you up for the topics you care about; the fix is usually content and authority work on those topics. A high rate means you're well established in the AI's answers.

Does the trend help? Yes — this is the single most useful metric to watch over time. A rising mention rate after you publish or earn coverage is direct evidence the work is landing; a falling one is an early warning that competitors are crowding you out.

Share of Voice

What it means. Of all the brand mentions in your space — yours plus your competitors' — the slice that is yours. Mention Rate asks "how often am I mentioned?"; Share of Voice asks "of everyone mentioned, how much of the conversation is me?"

How we work it out. For each prompt we note which brands showed up at all, then compare how many prompts you appear in against the field. It's counted per prompt — "did this brand show up for this question, yes or no" — which is a slightly different lens than Mention Rate.

How to use it. This is your competitive standing in one number. You can have a decent mention rate and still a small share of voice if a few competitors dominate. Use it to see whether you're the default answer in your category or one of many.

Does the trend help? Yes. Share of voice is a zero-sum pie. Watching your slice grow or shrink relative to named competitors tells you whether you're winning or losing ground, which a solo mention rate can hide.

Average Position

What it means. When the AI lists several options, where you fall in the order — first, second, fifth. Being recommended first is worth more than being mentioned last.

How we work it out. We read the order the brands appear in the answer and record your rank, then average it across the answers that mention you.

How to use it. Treat it as informational today. We surface it so you can see ordering, but we are still validating its accuracy, so it does not yet feed your headline score. Don't make big decisions on it alone yet.

Does the trend help? Eventually yes, but hold off until it's promoted from informational to scored. We'll announce when it's ready to trust.

Is the answer good for you?

Showing up isn't enough; how you show up matters.

Sentiment

What it means. The tone of the mention — is the AI speaking positively, neutrally, or negatively about you?

How we work it out. Each mention is scored on a scale from negative to positive (0 = negative, 50 = neutral, 100 = glowing), and we average those scores across your mentions.

How to use it. A strong mention rate with poor sentiment is a red flag: the AI is talking about you, but not kindly — often echoing a bad review, a comparison you lose, or outdated information. That's a content and reputation fix, not a visibility one.

Does the trend help? Yes. A sentiment dip often traces back to a specific new source the AI started leaning on. Catching it early lets you address the source.

Ownership Status

What it means. For each question, a plain verdict on how you did against direct competitors:

• Owned — you showed up and no direct competitor did. You win this question.
• Shared — you and at least one direct competitor both showed up. Contested ground.
• Lost — a direct competitor showed up and you didn't. They're winning this question.
• Uncovered — nobody relevant showed up. Open territory, nobody owns it yet.
• Expected — you showed up, but only because the question named you. Doesn't count as an earned win.

How we work it out. We look at whether you appeared, whether any direct competitor appeared (not adjacent or unrelated brands), and whether either of you was simply named in the question.

How to use it. This is your action list. "Lost" questions are where to invest first — a competitor is taking traffic you could have. "Shared" questions are where a push could flip you to "Owned." "Uncovered" questions are land grabs. "Owned" questions are what to defend.

Does the trend help? Yes, but the bigger value is the snapshot. Watching a question move from Lost to Shared to Owned is the cleanest proof that a specific piece of work paid off.

Are your pages behind the answer?

These metrics cover the engines that reveal their sources — today, ChatGPT and Perplexity.

Citation Rate

What it means. How often a page on your own website is cited as a source in the AI's answer. Being mentioned is good; being the page the AI learned it from is better, and more durable.

How we work it out. We look at the scans where your prompts ran and count how often one of your pages appears in the AI's cited sources, divided by the total.

How to use it. This is your content-authority signal. Low citation rate means the AI is answering about your space using other people's pages — a clear cue to publish stronger, more citable content on those topics.

Does the trend help? Yes. Citation rate climbing after you ship a new guide or resource is direct evidence the page is being picked up as a source.

Retrieval Rate

What it means. How often the AI fetched your page while researching the answer, whether or not it ended up citing it. Retrieval is "the AI found and read your page"; citation is "the AI trusted it enough to credit it."

How we work it out. We count the scans where your page was pulled in during the AI's web search, divided by the total.

How to use it. Read it together with Citation Rate. High retrieval but low citation is the classic "found but not trusted" pattern — your page is discoverable but isn't winning the citation.

Does the trend help? Modestly. It's most useful as the denominator for the Trust Ratio below.

Trust Ratio

What it means. Of the times the AI found your page, how often it actually cited it. A "do they trust you when they find you?" number.

How we work it out. Citations divided by retrievals.

How to use it. A low trust ratio — we flag it when it's under about a third, with enough data to be sure — is a specific, fixable diagnosis: the AI is reaching your page but choosing something else to credit. That points at authority, freshness, or clarity problems on the page itself, not a discoverability problem.

Does the trend help? Yes. It's the cleanest way to confirm that a page-quality fix worked: same page, more citations per fetch.

Consistency

What it means. When your page does surface for a question, how reliably the AI cites it across repeated asks. High means the engine dependably picks your page; low means it surfaces but the engine often grabs something else instead.

How we work it out. Among the repeats where your page came up, the share that actually cited it. We only show it once there's enough repetition to be meaningful — otherwise it's left blank rather than showing a misleading 100%.

How to use it. Low consistency means a citation you can't count on. It flags pages that are "sometimes the answer," worth strengthening so they become the dependable source.

Does the trend help? Somewhat, but it's mainly a per-page quality read at a point in time.

Trends

The little up / down / stable pill on a metric compares the current period against the previous period of equal length and tells you the direction. If the value moved by more than a small threshold, we call it up or down; within that threshold it's stable, so normal noise doesn't look like a real move.

One note you'll see in the UI: when your time filter is set to "all time," there is no earlier period to compare against, so the trend shows a dash (—) rather than "stable." A dash means "we can't compute a trend here," not "no change."

Trends turn a static number into a story — is this getting better or worse? — and that's usually what drives the decision to keep doing something or change course.

Per-engine breakdown

Every metric above can be split by engine: ChatGPT vs Perplexity vs Claude vs Gemini. Engines disagree more than you'd expect — you might own a question on ChatGPT but lose it on Perplexity. The per-engine view tells you where to focus. Remember that the source-based metrics (citations, trust, consistency) only cover the engines that expose their sources today.

Quick reference

Which metric answers which question:

Your question	Metric to look at
Am I showing up in AI answers at all?	Overall Visibility
How big is my slice vs competitors?	Share of Voice
Where do I fall when brands are listed?	Average Position
Are mentions positive?	Sentiment
Which questions am I winning/losing?	Ownership Status
Are my own pages the source?	Citation Rate
Found but not credited?	Retrieval Rate + Trust Ratio
Is a page a dependable source?	Consistency
Better or worse over time?	Trend (on any metric)
Which engine is the problem?	Per-engine breakdown

What action each metric drives:

Metric	If it's bad, you should…
Overall Visibility	Build topical content + authority where you're absent
Share of Voice	Find which competitors dominate and target their ground
Sentiment	Fix the narrative — bad reviews, lost comparisons, stale info
Ownership Status (Lost)	Prioritise these questions — competitors are taking them
Citation Rate	Publish stronger, more citable pages on those topics
Trust Ratio (low)	Improve the page itself — authority, freshness, clarity
Consistency (low)	Strengthen the page so it's the dependable answer