MentionX Dashboard Metrics Explained

If you open a MentionX scan and stare at Visibility, Mention rate, Recommendation rate, and Citation share, you are not alone — they sound like the same idea (“does AI like us?”) but they measure different moments in a buyer answer.

This guide matches what the product actually computes: stored engine responses, structured extraction, and versioned formulas in code — not a model guessing a score.

Who this is for

Most teams already have rank trackers, Search Console, and analytics. Those tools answer: Did our URL win a SERP slot and get a click?

MentionX answers: When a buyer asks an AI engine for advice in our category, does our brand appear in the generated answer — and in what role?

What SEO still does for AI visibility

• Crawlable, cite-worthy pages (comparisons, docs, pricing clarity) increase the odds your owned URLs appear when engines retrieve.
• Entity clarity (who you are, who you’re for) helps both classic snippets and AI summaries.
• Third-party proof (reviews, analyst, community) shows up in citation gaps — the same domains SEO/AEO teams already pursue, but prioritized by what AI cited in your scan, not a generic keyword list.

What SEO metrics cannot replace

• A #1 blog rank does not guarantee ChatGPT names you on “best {category} for {use case}.”
• High organic traffic does not prove you make the shortlist when Gemini outputs five vendors.
• Domain authority alone does not explain recommendation rate — engines may name incumbents from training while citing fresher third-party roundups.

Deep comparison: AI visibility vs SEO.

Most dashboard percentages ask:

In how many engine runs did this happen?

An engine run = one buyer-style prompt on one engine (ChatGPT, Gemini, Perplexity, etc.) in that scan.

total_engine_runs ≈ prompts × engines (for the scope you are viewing)

Filter Overview by one engine and the same definitions apply — only for that engine’s runs.

Every metric links to evidence: the stored answers that produced it. Open Answers or export citations when you need to show the room why a number moved.

Recommendation rate is the lead USP metric on the dashboard for a reason: buyers care when engines shortlist, not only when they name-drop.

These two are related but not interchangeable.

Mention rate

mention_rate = runs_where_your_brand_is_mentioned / total_engine_runs

Plain English: “What share of answers referenced us at all?”

Examples count: listed in a comparison, passing reference, product name, footnote — if extraction finds your entity in that run, it counts once for that run (not once per sentence).

Visibility score

visibility_score =
  0.40 × mention_rate
+ 0.35 × normalized_position
+ 0.25 × citation_share

Displayed as 0–100 on Overview (the ring). Internally it is a 0–1 composite multiplied by 100.

Plain English: “Overall, how strong is our AI presence on this scan?”

Typical story: Mention rate 39%, visibility 52. You show up often, but rivals outrank you or take more credited citations — visibility reflects that drag.

Typical story (reverse): Mention rate moderate, visibility strong. You appear less often but win lists and earn links when you do.

Worked example (one scan, fictional numbers)

Assume 100 engine runs (e.g. 25 prompts × 4 engines). Your Overview might show:

Outbound-safe headline: “We show up in 42% of buyer-style answers, but only get recommended in 28% — a 14-point gap between awareness and shortlist.”

That sentence is defensible because each number traces to stored runs — not a LLM vibe check.

Modern buyer answers are not binary (“mentioned / not”). Engines often output ordered recommendations.

Recommendation rate

recommendation_rate = runs_where_you_have_a_list_position / total_engine_runs

A recommendation means extraction found you in a ranked shortlist (position 1, 2, 3…). A paragraph that names you without a list position counts toward mention rate, not recommendation rate.

First-choice rate

first_choice_rate = runs_where_your_position_is_1 / total_engine_runs

At most one brand wins #1 per run.

Mention → Recommend gap

mention_recommend_gap = max(0, mention_rate - recommendation_rate)

Plain English: “Engines know our name but don’t put us on the shortlist.”

That gap is where proof content, comparison pages, review presence, and third-party citations usually matter — positioning without evidence.

Four stories the gap tells (and what to do)

Pattern B is the most common in outbound: incumbents get named from training; challengers win recommendations with fresher proof.

Share of voice (mention-based)

SOV = your_mentions / sum(mentions_for_you_and_tracked_competitors)

Relative metric on the same prompt set. A healthy mention rate means little if a competitor dominates the same buyer questions.

Share of voice is about conversation share, not preference. Pair it with recommendation rate to see who gets chosen, not only discussed.

Average position

When you appear in ranked lists, MentionX averages your list positions (lower is better). If you are often mentioned but rarely ranked, average position may be weak or empty while mention rate looks fine.

Position also feeds visibility through normalized position (best rank in the scan sets the scale).

People often assume:

“Citation share = what percent of links point to our website.”

That is not what Overview shows.

What a citation is

A citation is a URL the engine attached or surfaced in the answer — your site, a competitor’s, G2, Wikipedia, Reddit, a news article, docs, etc. MentionX merges engine-provided sources with URLs found in the response text and stores them as evidence.

What citation share measures

Citation share is a competitive split among your brand + tracked competitors:

citation_share(you) =
  citations_credited_to_you
  / sum(citations_credited_to_all_tracked_brands)

Credit means the citation is associated with that entity in extraction (often because the brand name appears near the link or the URL clearly supports that vendor). It is not limited to links on your own domain.

Easy example (four links)

Pool for share: 3 credited links (Acme 2, Beta 1).

• Acme citation share = 2 ÷ 3 ≈ 67%
• Beta citation share = 1 ÷ 3 ≈ 33%

Wikipedia does not turn Acme’s share into “2 out of 4 = 50%.” Neutral sources are still valuable evidence — they just are not a rival “win” in the share fraction.

Your ten-link thought experiment (revisited)

You asked: 10 citations, only 3 touch MentionX, 7 have no trace — what’s citation share?

Walk it twice:

Version A — 7 links are neutral (industry wiki, generic “best tools” list with no vendor credit):

• Credited pool might be only 3 (all to you) → 100% citation share vs rivals, even though 7 URLs “ignored” your brand in prose.

Version B — 7 links support competitors (each credited to a tracked rival):

• You 3, rivals 7 → 30% citation share.

Same raw link count, opposite board story. That’s why MentionX separates Citations tab (full evidence diet) from citation share (competitive credit).

Owned site vs credited proof

A strong owned count with weak recommendation rate often means your pages exist but engines still prefer other vendors in lists — fix positioning and third-party corroboration, not only CMS tweaks.

When citation share is 0%

For “did they link our domain?” use the Citations tab breakdown: Owned vs competitor vs third-party, plus citation gap tables for domains rivals earn and you do not.

See also: Brand mentions vs citations.

Engines cite because their product uses retrieval and grounding — to justify claims, show freshness, and reduce hallucination. They are not scoring your GEO program.

MentionX records all extracted citations because:

1. Transparency — every KPI traces to stored answers.
2. Category intelligence — neutral domains reveal which sources shape the topic even when your name is absent.
3. Actions — citation-gap detection finds third-party domains that appear for competitors’ runs but not yours.
4. Content strategy — if AI lives on review roundups and comparison publishers, you need presence there — not only on your blog.

Neutral citations answer: “What proof diet is this engine using for our category?” Citation share answers: “When proof is credited to vendors in our set, who wins?”

Both questions matter; only the second is the Overview KPI.

Intent coverage

Mention, recommendation, and first-choice rates split by prompt family — category discovery, use cases, buying criteria, competitive, alternatives, brand evaluation.

Use it when aggregate scores hide a weakness: strong on “vs Competitor X,” weak on organic “best tools for …” prompts.

Visibility over time

Trend of visibility (and related history) across scans — same definitions, compared longitudinally.

Engine and competitor breakdowns

Per-engine tables repeat mention rate, visibility, average position, and citation share because ChatGPT ≠ Gemini ≠ Perplexity on the same prompts.

Outbound email / LinkedIn (honest, evidence-backed)

Structure that converts without hype:

1. One category prompt you ran (paraphrase — no need to paste full prompt in cold email).
2. Two numbers: mention rate + recommendation rate (or the gap in points).
3. One rival contrast on the same scan (share of voice or their recommendation rate).
4. Offer proof: “Happy to share the redacted answer screenshots from the scan.”

Template (fill from your dashboard):

We ran buyer-style prompts on {engines} for {category}. {Brand} appears in {mention}% of answers but is recommended in only {recommend}% — a {gap}-point gap between name recognition and shortlist. {Rival} leads on recommendation rate for the same prompt set. The underlying engine answers are stored if your team wants to see the evidence.

Do not claim query volume, “ChatGPT search share,” or scores without a scan — MentionX does not fabricate those.

Discovery call talk track

Board / exec one-slide story

Lead with Recommendation Rate, not the ring alone:

• Headline: Shortlist rate + gap vs mention rate
• Context: Visibility score {v} (mention {m}, position, citation share)
• Competitive: SOV {sov} vs named rivals on same prompts
• Ask: 2–3 Actions tied to citation gaps or weak intent families
• Proof: Boardroom export or Answers appendix (plan-dependent)

Boardroom decks in MentionX follow the same hierarchy — recommendation up front, evidence-backed.

1. Recommendation rate — Are we on the shortlist buyers see?
2. Mention → Recommend gap — Are we famous but not chosen?
3. Share of voice — Who owns the narrative on our prompt set?
4. Citation share + Citations tab — Do we bring proof, and is it our site or trusted third parties?
5. Visibility score — Executive rollup; drill into the three ingredients before overreacting to the ring alone.

Then open Answers for the runs that hurt and attach them to Actions — that is the closed loop MentionX is built for.

Weekly rhythm (SEO + outbound + PMM)

Weights, extraction rules, and reproducibility notes live in MentionX Methodology. When we change scoring, we version it — historical scans keep the scorer version they ran under.

Category framing without the dashboard: What is AI visibility?

• Run a scan on your domain →
• MentionX Methodology — full formula catalog
• Mentions vs citations — four wild patterns and fixes
• GEO complete guide — turn gaps into a content program