CitadexDocs

Competitors — Share of Voice, Brand Ranking and Prompt Gaps

The two tabs of the Competitors page (Competitive landscape, Prompt gaps), how weighted recommendation share is calculated, how to configure competitors, how the batch fix produces a brief, and where retests live.

Content Tools8 min read

Layout

The bar at the top, on the right: the last scan time in grey, Export report, 文A (Translate all shows the prompts in the language you pick) and the dark Configure competitors button. The page refreshes itself after a scan completes. Two text tabs: Competitive landscape and Prompt gaps.

Competitive landscape

One filter row at the top: market, engine and period (all time, last 7, 30, 90 days).

Competitors: your configured competitors, edited with Configure competitors. Under it, Suggested competitors lists brands AI names often that are neither on your list nor excluded, each with "N mentions": "Add" puts it on your list, "Not a competitor" excludes it everywhere.

Five metric cards: share of voice (its ⓘ shows your rank among all brands), appearance rate, rec. share, avg. position and trailing prompts (prompts where a competitor appears on more engines than you, or is named while you are not).

Brand ranking table, one row per brand; the ⓘ beside each column name explains it, and a column with no data yet is hidden:

ColumnMeaning
Share of voiceThe brand’s mentions ÷ the mentions of every brand in the table; adds up to 100%
Appearance rateShare of answers that mention the brand
Rec. shareShare of answers that recommend the brand; your cell may show "Conditional {p}%" under it
Weighted rec.Each recommendation is divided by the brands named in that answer: alone = 1, among 3 brands = 1/3; divided by answers
Average position, sentimentAverage place in the answer’s list (you and the rivals ranked together; mentions outside a list don’t count) and overall sentiment
Citation shareShare of AI answers that cite the brand’s own site
TrendSmall bars per period

Weighted recommendation share keeps a recommendation shared with ten brands from counting the same as a sole recommendation. The Dashboard "From mention to recommendation" card uses the same method; see From mention to recommendation. Two switches sit to the right of the title: Overview / By engine / By topic (engine and topic show an appearance-rate matrix) and My competitors / All brands (all brands named in answers). Hover a rival row and click ✕ "Not a competitor" to exclude it. The table shows 5 brands by default (your own brand is always among them); "Show n more" shows the rest and "Show less" folds them back, in the By engine / By topic matrices too. Each brand drawn in the trend chart has a small dot before its name in the same colour as its line. Below the table is the share-of-voice trend line chart: your brand is the thick amber line with dots; competitors (up to 4) are thinner lines in warm tones (brown, sand, taupe, bronze), also told apart by dash style.

Trailing prompts · n: under the title, one row of pills, one per competitor ahead of you, each with its count (e.g. "Profound 23"); the selected pill has a dark outline and the first is selected by default. Click a pill to see only that competitor’s prompts. On the right of the same row, for the selected competitor: "Replaced n" (when some of its prompts used to name you), a "Cited sources" link that opens Citation analysis filtered to those prompts, and a dark "View countermeasure · n" button that opens the batch-fix dialog. The list shows 10 prompts per page, with "‹ 1 / 3 ›" page buttons ("Previous page" / "Next page") at the bottom right. Each row is one prompt, with "Replaced · date" and "{n} published · MM-DD" tags when they apply. Only articles that went live count (a publish record, or the article marked published by hand); unpublished drafts are not on the web, so they cannot change AI answers. When an article has been live for a while and the prompt still trails a competitor, the tag reads "{n} published · No effect yet" and a "Re-test" link appears next to it: it opens Automations → Re-tests with a new re-test pre-filled with that prompt and the article URL. On the right: "Coverage c/e" (engines that name you / engines) and a › arrow. A row turns grey on hover; click anywhere on it (or press Enter) to open the evidence panel with the AI answers for that prompt. The "Re-test" link inside a row works on its own. Only configured competitors get their own pill (up to 8); brands not on your competitor list share one Other brands pill, and each row names the brand.

Batch fix: "View countermeasure · n" pre-selects the largest same-language, same-intent cluster among that competitor’s prompts (up to 8 prompts). The dialog shows "Covers n prompts"; if languages or intents are mixed it suggests splitting them and offers "keep only this group"; prompts that already have an article are called out, with "Keep ungenerated (m)"; then it lists the cited articles (third-party articles AI cites for these prompts). "Generate brief" produces a writing brief and deducts credits by the actual AI cost straight away, with no quote step; it appears in the credit history on the credit purchase page (/api-keys) under the reason assistant:brief. The brief can then be turned into an article that goes to Content for review.

Competitor citation activity: competitor website pages AI cited in the last 7 days whose citation status is New or Returned, with "n citations in 7 days"; when answers cite a page without naming you, "n prompts without you" shows and those pages come first. Click a row to open the evidence for those answers; the link icon on hover opens the competitor page. Every Monday, when a competitor page has just started being cited in 3 or more answers that leave you out, the AI assistant tells you in the chat (Market health follow-up).

Brand attributes: the words AI answers use most for each brand (e.g. "enterprise-grade", "expensive", "quick to set up"), one row for you and each configured competitor; rows still collecting show "Collecting data", and older answers only give broad categories (price, ease of use, features, case studies & reviews, awareness, support). Attribute gap: what rivals are often praised for and you never are.

Chat box at the bottom ("Questions? Ask AI"): it stays at the bottom of the screen while you scroll. Your question goes to the AI assistant, which knows what is on the page (filters, trailing prompts per competitor and so on) and answers with this page’s numbers.

Pages that cite a competitor and have never cited you are under Citation analysis → "Rival-only"; the "Cited sources" link next to the competitor pills gets you there too.

Prompt gaps

Real searches where a competitor is recommended, you are not mentioned and you do not track the question yet. Every question comes from the market's real searches (Search Console, Google autocomplete, People also ask, forums); each shows its source and the search itself, and keeps that evidence when you add it to tracking. Questions with no search evidence never appear here. When there is no new real search to check, that batch is skipped without using credits, and the page suggests connecting Search Console or finding more candidates in Prompt Library. Needs configured competitors (otherwise "No competitors configured"). They are found by the "Prompt gap scan" automation: when you turn it on, pick weekly, every 2 weeks or monthly, and the first batch runs within minutes. Each candidate question is verified on three AI engines, charged in credits for the calls that get an answer, even if no gap is found. A grey line on the page says when the next scan runs, with "Settings" or "Turn on". Results are grouped by market (flag and market name), each tagged with its kind of question (Comparison, Best-of, How-to), with "Add to tracking" or "Add all to tracking (n)".

Retests

Checking a change now happens under Automations → Re-tests: click + New re-test for one market, and Citadex picks target and control prompts and compares before and after. See Retests.

Configuring competitors

Configure competitors → type a name or domain and Add, or pick from the AI-detected suggestions → Save. This list overrides automatic detection everywhere on the page and is also used for weighted recommendation share, the brand-search checks and the competitor-ahead tasks in This week’s tasks.

Common questions

What does "{n} published · No effect yet" mean? An article for this prompt has been live for a while, but the prompt still trails a competitor. Click "Re-test" to measure again; if nothing changes, click the row to compare what the winning competitor’s cited pages cover and rewrite from another angle. Every Monday the AI assistant also tells you in the chat about new ones (Results of what you did in Automations; turn off assistant follow-ups to stop it).

Don’t understand a number? Ask in the chat box at the bottom of the page.

Plan notes

Competitor analysis runs have a monthly cap per plan. Engines not in your plan are locked in the platform filter.

Was this article helpful?