Scrunch Metrics Reference: Explorer Names and API Fields

Definition: Share of responses whose text names your brand — regardless of tone, and regardless of whether you're also cited as a source.

Formula: brand_presence_percentage = (responses naming the brand) ÷ (total responses). Brand and competitor mentions use the same detection logic, so the two are directly comparable.

Why it matters: The core "are we part of the conversation at all" number — the widest, simplest AI-visibility signal, and usually the first one teams track.

When to use it: Default choice for overall AI visibility tracking. Reach for Citations Rate instead when the question is "is AI pulling from our content as a source," not "is AI talking about us."

Watch out for: Citations of your domain do not count here — a page of yours can be cited without your brand ever being named in the answer text. That's tracked separately, by Citations Rate.

Definition: The average tone of responses that mention your brand, expressed as a single 0–100 score.

Formula: Each qualifying response scores positive = 100, mixed = 50, negative = 0; brand_sentiment_score is the mean across responses that mention the brand. Responses scored "Unclear" (the none band) are excluded from the average, not counted as a middle value.

Why it matters: Tells you not just whether you're mentioned but how — an early-warning signal for competitive disadvantage (AI naming you while recommending someone else).

When to use it: Track this as your headline sentiment number. For the underlying distribution (positive vs. mixed vs. negative), break down by Sentiment band instead of inferring it from the average — 85 does not mean "85% positive."

Watch out for: AI responses are rarely outright negative about a brand — most negative-band responses reflect a competitor being recommended instead, not explicit criticism. A falling Sentiment Average usually signals competitive framing, not reputational attack. See Understanding Sentiment in Scrunch for the full walkthrough.

Definition: How prominently your brand appears within responses that mention it, as a single 0–100 score.

Formula: Among responses that mention the brand: top-of-response = 100, middle = 50, bottom = 0; brand_position_score is the mean. Calculated only over responses where the brand appears — an absent response doesn't pull the number down.

Why it matters: A rough analogue to SEO ranking position — how close to the top of the answer you land, not just whether you appear.

When to use it: Track placement trends over time or across platforms; break down by Position bucket for the Top/Middle/Bottom distribution instead of the averaged score. Don't over-index on the absolute score alone — as with SEO, attention drops off sharply after a response's first section, and it's unclear how much of a long answer users actually read.

Watch out for: Like Sentiment Average, this only covers responses where the brand is already mentioned — it says nothing about how often you're mentioned at all. Read Mentions Rate and Position Average side by side rather than combining them into one "dominance" number.

Definition: On average, how far into the response your brand's first mention falls, relative to other entities mentioned in the same response.

Formula: brand_avg_rank averages the first-occurrence mention rank (1 = mentioned first) over responses where the brand received a rank; unranked responses are skipped, not scored as bad. Lower is better — Explorer's delta indicator flips its good/bad coloring accordingly.

Why it matters: A finer-grained signal than Position's three tiers: are you the first brand AI names, or the fourth?

When to use it: Use alongside Position Average when list-style responses ("here are 5 options for X") are common in your category and simple top/middle/bottom buckets feel too coarse.

Watch out for: This is a raw number (e.g. 2.3), not a 0–100 score like Sentiment or Position — don't read it on the same scale.

Definition: Share of all responses (whether they cite anything or not) that cite at least one URL you own.

Formula: brand_citation_rate = (responses citing ≥1 brand-owned URL) ÷ (total responses). Domain-based — matched against your registered brand websites — and independent of whether your brand is mentioned by name in the text.

Why it matters: Tells you whether your content is a source AI draws from, as distinct from whether AI talks about your brand at all.

When to use it: Explorer's default, all-responses denominator for "how often are we cited." Use Citations Share of Voice instead when you want to know how you compare among responses that cite something — a smaller, more competitive denominator.

Watch out for: Being mentioned isn't the same as being cited, and being cited isn't the same as being cited prominently — track Mentions Rate, Citations Rate, and Position Average as separate signals rather than assuming they move together.

Definition: Of the responses that cite anything at all, the share that cite a source you own.

Formula: brand_citation_share_of_voice = (responses citing a brand-owned source) ÷ (responses citing any source at all). Each citing response counts once toward this share, no matter how many of your pages it cites.

Why it matters: Measures how reliably you show up when citations happen at all — a coverage/consistency signal, distinct from raw volume.

When to use it: Most teams find this more actionable than Citations Rate for strategic tracking, since it isolates your performance from swings in how often AI cites anything at all in your category. Use Citations Total instead when the question is raw competitive volume, not consistency.

Watch out for: A high Citations Total alongside a comparatively low Citations Share of Voice usually means AI cites you deeply (several pages) in a smaller set of responses — a sign of content depth, not underperformance.

Definition: The raw count of citation occurrences pointing at a brand-owned URL, across all responses.

Formula: brand_citation_count = total citation events where the source is brand-owned. If one response cites four of your pages, that's four toward this count.

Why it matters: A volume measure — useful for competitive positioning ("we account for X citations vs. competitor Y's Z") and for tracking overall citation volume in your category.

When to use it: Pair with a Citation Source Owner breakdown for the brand/competitor/other composition of your citation volume. Reach for Citations Unique Responses instead if repeat citations within one response are inflating this number beyond what you want to compare.

Watch out for: Multiple citations in one response all count separately here, so this number can look large relative to Citations Share of Voice — a count-based vs. response-based distinction, not a bug (see the side-by-side comparison in Understanding Citation Metrics in Scrunch).

Definition: Distinct responses that cite at least one brand-owned source.

Formula: brand_citation_unique_responses = COUNT(DISTINCT response) where a brand-owned citation is present.

Why it matters: The numerator you'd use to hand-verify Citations Rate or Citations Share of Voice against a specific response count.

Definition: Distinct brand-owned hosts that appear in at least one citation.

Formula: brand_citation_unique_domains = COUNT(DISTINCT domain) among brand-owned citations, www-stripped.

Why it matters: How many different pages or sections of your site are earning citations, versus concentration on one URL.

Definition: The number of distinct (response, domain, owner) combinations across all citations.

Formula: citation_unique_response_domains — entity-agnostic. A response citing one domain under four URLs counts once; a response citing six different domains counts six. Owner is part of the key, so subdomains with mixed owners split across owner slices.

Why it matters: The reconciliation field — use it, broken down by Citation Source Owner or Citation Domain, whenever a citation count needs to match a per-source-and-owner grain exactly, since Citations Total (a raw URL count) uses a different grain.

When to use it: Pair with a domain-grain breakdown. Don't pair it with a Citation URL or Domain Topics breakdown — those change the counting unit and the totals stop reconciling.

Definition: Of the URLs cited in AI responses, the share whose page content (not the AI's answer text) actually names your brand.

Formula: brand_citation_mention_rate = (distinct cited URLs whose page text mentions the brand) ÷ (distinct cited URLs). The denominator is cited URLs, not responses — a different grain from every other metric on this page.

Why it matters: A quick sanity check on why a third-party page is earning you a citation — is your brand actually named on the page AI pulled from?

When to use it: Use it to audit specific citing pages, not as a substitute for Mentions Rate — that's determined directly from what the AI said, not from what a cited page contains.

Watch out for: A "No" here doesn't mean your brand is missing from AI's answer — it only reflects whether Scrunch could detect your name in that page's plain-text HTML. Bot-blocking (Cloudflare, Datadome) and JavaScript-rendered content are common reasons for a false "No." See Understanding the "Brand Present" Column in Citations for the full explanation.

Definition: How influential a cited source is, combining how often it's cited with how many different questions it's cited for.

Formula: weighted_index = (citations of this source ÷ total citations in the group) × (distinct prompts that cited it).

Why it matters: Separates a source influential across many different topics from one that got a single high-volume hit on one popular prompt.

When to use it: Requires a Citation Domain or Citation URL breakdown to mean anything — without one, there's no group to compute a share against and the chart returns no data. Reach for Citations Total instead for a simple citation-volume ranking with no prompt-diversity weighting.

Watch out for: Without the required breakdown applied, this metric is meaningless by design, not broken — always confirm a domain or URL breakdown is in place before reading the number.

Worked example: A source cited 30 times out of 300 total group citations (a 10% share), appearing across 12 distinct prompts → Influence Score = 0.10 × 12 = 1.2.

Definition: Your brand's mentions as a share of total mentions across the active competitive set, normalized so every entity's slice sums to 1.

Formula: brand_share_of_voice — unlike Mentions Rate (each entity's own independent rate), Share of Voice is a normalized pie: every tracked entity's slice adds up to 100%.

Why it matters: Answers "of all the attention going to brands in our category, how much of it is ours" — a relative-share question, distinct from Mentions Rate's absolute-rate question.

When to use it: Use this for competitive-share reporting and pie/donut visuals. Reach for Mentions Rate instead for each entity's own independent visibility number — mentions overlap across entities, so individual Mentions Rates don't have to sum to 100%, and can exceed it. See Understanding Competitive Presence in Scrunch for why an aggregate presence number and individual competitor numbers don't add up.

Total Responses: All responses collected in the selected range, regardless of which entities are mentioned. Use it to sanity-check sample size before trusting a percentage — Explorer's tooltips already show this alongside every rate (e.g. "43% (37 of 86 responses)").

Unique Prompts: Distinct prompts with at least one response in the range. Coverage of your monitored prompt set, not a mention-based count — use Mentions Unique Prompts instead if you want "how many prompts is the brand showing up for."

Watch out for: Neither is directly comparable period-over-period the way a rate is — a rising Total Responses count usually just means you added prompts or platforms, not that visibility changed.

Up to 5 metrics in one view.

Up to 3 breakdowns at once.

Up to 8 filters at once.

persona_id, competitor_id — the numeric IDs behind Persona and Competitor. Explorer always uses the name-based dimension (persona_name, competitor_name) instead.

source_domain — a citation host dimension computed on a different query path than Explorer's Citation Domain breakdown (domain); it can't be combined with citation metrics and isn't filterable. Prefer domain for anything Explorer-facing.

cited_competitor_name — the source-grain sibling of the Competitor breakdown, used only for citation metrics. Explorer's Competitor breakdown auto-substitutes this when a citation metric is active, so it never got a separate picker entry — a duplicate "Competitor" option tested as more confusing than helpful.

has_sponsored_ads (filter-only) — whether a response displayed a sponsored ad. This is gated behind a rollout flag and isn't yet available as an Explorer filter.

Query API overview — the simple GET endpoint and the full dimensions/metrics reference.

Structured Query endpoint — the JSON POST endpoint for having thresholds, negated filters, and period-over-period comparison.

Responses API overview — raw response text, citations, and per-response detail.

Scrunch MCP tools reference — query_metrics and get_metric_glossary for pulling this same data through an AI assistant.