Competitive AI Performance
What this is for, in one sentence: the head-to-head table — you and every tracked rival, side by side, on the same six measures, so a number like “41.7% coverage” finally has context.
When to come here: whenever you’re tempted to ask “is my number good?” — this screen is the answer. Also before any pitch, board meeting, or client review: nothing lands like “here’s exactly where we stand against the rivals you care about.”
We continue with Sarah at MenuPilot and her four tracked competitors.
Reading the table
Section titled “Reading the table”
Every tracked brand gets a row with the same measures you know from the dashboard — Mention Coverage, Share of Voice, Citations, Average Ranking, Average Sentiment — plus Top Descriptors (the actual words AI uses about each brand). Your own row carries the YOU highlight so you can always find yourself at a glance, alongside an Unranked badge — Zicy doesn’t hand your own brand a competitive rank number (you’re the reference point, not a ranked rival).
Cells are color-coded by performance (green/amber/red), which makes the strategic read fast:
- Scan your row for red — those are your weak measures.
- Scan each column for green — that’s who to study on each measure.
- Find the near-ties — Sarah trails PlateIQ Labs on coverage (41.7% vs 51.7%) but leads on sentiment (72 vs 70). Close races are where effort pays fastest.
One reading habit that prevents wrong conclusions: a rival can lead on one measure and trail on another — PlateIQ is mentioned more often, but when MenuPilot appears, AI speaks slightly more warmly of it. There’s no single “winner” column; the story is the pattern.
Why your Share of Voice percentages don’t sum to 100
Section titled “Why your Share of Voice percentages don’t sum to 100”Add up the SoV column and you’ll get less than 100%. Not a bug: AI answers also mention brands you aren’t tracking — small players, adjacent tools, one-off names. Their mentions sit in the denominator. If one untracked name keeps appearing in your per-prompt answers (M-03), that’s a rival worth watching — as it surfaces more often in AI answers, Zicy will fold it into the tracked set.
Top Descriptors — the narrative column
Section titled “Top Descriptors — the narrative column”The most underrated column. Each brand’s cell shows the handful of words AI most attaches to it across all answers — the compressed version of every framing you’d find reading answers one by one. (Want the full per-rival breakdown with exact counts, including negative associations? That lives in Competitor Sentiment Profiles (M-08).)
Read your own descriptors and ask: is this the story we’re trying to tell? Sarah’s “practical, easy to use, AI-driven” matches her positioning — good. Read your rivals’ descriptors and ask: what story is AI telling for them that we could contest? PlateIQ owns “established” and “enterprise-grade” — which cedes “easiest for independents” to whoever claims it.
Who exactly are you being compared against?
Section titled “Who exactly are you being compared against?”Zicy auto-selects your competitor set from the brands that actually surface alongside you in AI answers — you don’t pick or add them by hand. Read the list critically — it’s a smart guess, not a verdict, and every number on this screen (and Share of Voice everywhere) depends on who’s in it.
Sense-check the set before you quote the table:
- Are these the rivals you actually lose deals to? If a brand your sales team argues about isn’t here, that’s a signal in itself — it simply isn’t surfacing in AI answers for your prompts yet. Broadening your prompts (M-02) so answers naturally mention it is how it earns a row.
- Any names that look like duplicates or near-duplicates? Two rows for what’s really one company (e.g., a brand and its product line) split one rival’s numbers into two smaller ones and quietly flatter your position. If you spot a suspected duplicate, verify by domain before quoting the table.
- Anyone irrelevant? A brand from a neighbouring category inflates the denominator — note it when you read the numbers.
Common questions
Section titled “Common questions”A competitor’s numbers look strangely low — are they really that invisible? Check two things first: how recently they started appearing (a newly-surfaced rival has thin data that reads as weakness), and whether their mentions might be split across a name variant. Then believe the number.
Can I choose who I’m compared against? Not directly — Zicy builds the competitor set automatically from the brands that appear in your prompts’ AI answers; there’s no manual “add a rival” step. To influence who shows up, shape your prompts (M-02): the brands AI names alongside you are exactly the ones that land here.
My own row says “Unranked” but also shows an Average Ranking — which is right? Both — they measure different things. Unranked just means Zicy doesn’t hand your own brand a competitive rank number in this table (you’re the reference point, not a ranked rival). Average Ranking (e.g. #2.8) is a separate metric: your typical position inside AI answers. Quote the Average Ranking — “typically the 2nd–3rd brand AI names” — and read the table by comparing your columns against each rival’s.
Do competitors cost extra quota? No — they’re measured from the same AI answers your prompts already generate. Prompts cost quota; competitor tracking doesn’t.
What to do next
Section titled “What to do next”- See the words behind each rival’s sentiment score: Competitor Sentiment Profiles (M-08).
- Find the topics where each rival beats you: Key Topics Analysis (M-09).
- See whose websites AI trusts: Citation Analysis (M-10).