CitadexDocs

Prompt Tracking: the brands AI recommends, your rank and the reasons

How to read the tracking table, the evidence drawer, scan frequency, auto-scan, resuming an interrupted check, bulk actions (including several articles at once from many selected prompts), and the Achievements tab.

Product Manuals13 min read

What this page is for

Every tracked question is sent to each AI engine as a real query. This page shows, prompt by prompt: coverage, whether each engine mentioned you, your rank, whether you were cited, how strongly you were recommended, and whether AI mistook your brand for a different business. You can generate a strategy article straight from a gap. What AI gets wrong about your brand and whether a change worked are checked on the Automations page (Brand fact check and Re-tests).

When to use it

  • After adding prompts, to see the first results.
  • Once a week, to spot prompts that need attention.
  • To give a prompt a higher scan frequency, or start a strategy article from here.
  • To manage your tracked prompts: this is the one list of them, where you change a prompt's question type, pause or resume tracking, or delete it.

The page

Top bar: a bar across the top of the page that stays in place as you scroll; in chat mode its left side shows "project / Prompt Tracking". On the right: the scan snapshot date (when there are several scans; pick an older snapshot, then Back to latest; the newest is marked in the list), Frozen comparison, the recycle bin icon (deleted prompts can be restored for 30 days), Generate strategy article (it suggests uncovered prompts on one topic, grouped from your prompts’ topics, and writes one article on them; one or the other with Auto-generate articles: while that is on, the button is muted, and clicking it asks whether to go to the Automations page to turn Auto-generate articles off) and, at the end, the dark Add Prompt (the same dialog as Add Prompt in the Prompt Library; its Market menu lists the market you have filtered to first, and + Add markets, its last item, adds a new market). On a phone, Frozen comparison, Generate strategy article and the recycle bin are under ⋯. The Auto-scan and Auto-generate articles switches are on the Automations page, under the library in the left sidebar. There is no big page title; the tabs come right under the bar. Opened in the AI assistant's side panel, the tabs (Tracked, Achievements) sit on the bar's left instead, with the ones that don't fit under More ▾; on the right are the date, ⋯ and Add Prompt, with the other buttons in ⋯.

Tabs: Tracked and Achievements. The separate AI searches tab was retired: when you tick prompts and click Generate strategy article, Include AI searches in the dialog folds a few searches AI ran for those prompts without citing you into the article (old ?tab=fanout links open Tracked). The former opportunities tab was retired; old links (/rankings?tab=opportunities and /citations?tab=opportunities) open Tracked filtered to Uncovered. Re-tests and brand searches used to be tabs here too; they are now on the Automations page (Re-tests and Brand fact check), and old links open them there. Outreach was a tab here as well; it is now in Citation Sources → Market health & repair, and old links open it there. The page address follows the open tab, so a copied or reloaded link opens the same tab; links in notifications, email and Slack open the project they refer to.

Tracked

Banners can appear above the table: on the free plan only part of the prompts is scanned (Upgrade to scan all); prompt language doesn’t match its market (Move to matching market); Last check was interrupted (Resume continues from where it stopped, Dismiss clears it); new prompts are syncing.

Filter row (above the table): on the left, which prompts to show — Coverage (All / Covered ≥ 50% / Uncovered < 50% / Context issue / No list, with counts; Paused n is added when any prompt is paused; which competitor replaced you on a prompt is in its Evidence), then Question type (Discovery / Comparisons / Products & services / Customer pain points / Use cases / Industry trends / Brand, the same kinds as Coverage and candidates in the Prompt Library, with the same counts), Group, Stage (Awareness / Consideration / Decision, counted the same way as Buying stage in the Prompt Library) and Market (with several markets). A filter that is set shows on a soft blue: Question type and Stage keep their name, e.g. "Question type: Comparisons"; Coverage, Group and Market show the value alone, e.g. "Uncovered 120", "Agency / white label", "United States" (a market by its country, with the language added when the project monitors that country in more than one, e.g. "Canada · English"), and hovering shows the filter's name with the value. Its ✕ clears it, and Clear filters appears while any is set. Arriving from n tracked on a kind in the Prompt Library's Coverage and candidates, Question type (or Stage) and Market are already set. On the right, how to look at them: search, the Order button (it shows the current order, e.g. Order: Latest checked; the first item in its menu is Priority, high to low) and the translate button (the 文A icon, "Translate all" on hover; once on it reads "文A Chinese ▾" and its menu picks the language, switches Show originals and turns translation off). Generate strategy article is in the bar at the top, next to Automation (greyed out while Auto-generate articles is on; it's one or the other). Selecting rows turns the toolbar into the bulk bar, led by the select-all box (this page): Delete, Pin to top, Generate strategy article, Set a goal, Cited sources, Move to group, Move to market and Fix {n} brand mismatches. Pinned prompts stay at the top of the list and show Pin to top; for one prompt use Pin to top / Unpin in the row's ⋯ menu.

Views: the Views button on the right of the filter row. Picking a built-in view (All prompts / Priority / Uncovered / Decision stage / No list / With ads / Newest added) or one under My views sets its market, coverage, question type, stage, ads, search, group, order and rows per page; Save current view saves the current settings under a name as a personal view (visible only to you, the same on every device, up to 20 per project), and each view's ⋯ menu has Set as default view or Restore default, Rename and Delete view. Prompt Tracking opens with your default view, except from a link that sets a filter, which opens as the link says; when the menu has no My views section, only the built-in views are available.

Columns: prompt (no badges under it: replacement by a competitor and context issues are in the evidence; alerts when you are no longer mentioned come by bell and email), then on its right Pin to top, Paused (when tracking is paused) and Generated n×, right against Coverage (a coloured % then engines that mention you / engines that answered), time checked, the Evidence button and the ⋯ menu (Cited sources / Rewrite this prompt for a prompt with no list / Phrasing test (on a checked prompt) / Question type: <current type> / Pause tracking or Resume tracking / Pin to top or Unpin / Move up / Move down (shown only in Custom order) / Delete).

Prompt Tracking · Tracked
PromptArticlesCoverageChecked
best CRM for small businessGenerated 2×33% 2/62 hours ago
Acme vs HubSpot67% 4/62 hours ago
CRM consultants in BerlinChecking…—
👆 Click Evidence to read each engine’s raw answer

Priority, high to low: the first item in the Order menu. Prompts are ordered by commercial opportunity: each prompt is graded High, Medium or Low in its own market, and higher grades come first. In this order each row shows its grade; hover or click it for Commercial opportunity: the grade, the scenario (Organic acquisition, Brand protection, Competitor displacement), the confidence, the main reasons, the basis of six factors (Demand, Buying intent, Recommendation gap, Competitors, Citation gap, Orders) and the conditions that limit the grade. With order data connected, orders count as Modeled evidence. The grade shows how much to prioritise a prompt; it is not a forecast of results. Prompts with no valid answer in the last 30 days show "—", go last and are graded once they are checked.

Phrasing test: in a row's ⋯ menu. The same question is asked 3 other ways plus the original, each once on ChatGPT, Gemini and Perplexity, and a table shows which engines name you for each phrasing. When a phrasing leaves you out on more engines, the result names it and you can track it with Track; when the results hardly differ, it says so. It is a one-off check: nothing is tracked unless you add it. It uses credits.

Question type: in a row's ⋯ menu, shown as Question type: <current type>. Click it to open the seven kinds in place (Discovery / Comparisons / Products & services / Customer pain points / Use cases / Industry trends / Brand) and pick one. Your choice is kept, and AI grouping no longer changes it; Coverage and candidates in the Prompt Library and the Question type filter here both count the prompt under the new kind. Choosing Brand marks it as a brand prompt; moving it out of Brand clears that.

Pause tracking / Resume tracking: in a row's ⋯ menu. A paused prompt is no longer checked, its past results stay, and the row carries a Paused tag. Resuming goes through the same credit check as adding a prompt; with too little balance it is not resumed.

The evidence drawer

Click Evidence to open it. At the top, set the prompt’s Scan frequency (Daily / Weekly / Monthly) or click Scan now. Then coverage, average rank, sentiment and the suggested content type, followed by each engine’s raw answer:

  • Whether you appear, your position and whether you were cited.
  • The recommendation level: Top pick, Recommended, Listed option or Mentioned only, with the reason.
  • The data source: captured from the AI app itself (what users see) or the model API, and the model.
  • Searches the AI ran before answering, when the engine searched the web.
  • Competitors mentioned and Sources.
  • Earlier answers: previous answers to the same prompt. If a judgment is wrong (the brand IS or is NOT mentioned, or the recommendation level should be different), use Submit correction; the correction is kept beside the answer and the answer itself is not changed.
  • Replaced by competitor · N: when an engine recommended a competitor instead of you, this tab next to the answer lists each such engine and the competitor it recommended.

When AI has placed your brand in the wrong industry, a card explains it, with options to write a clarification article or hand it to automatic article generation.

Achievements

Titled Optimization results, for the past N days: Coverage crossed 50%, Mention regained, New citations, Article worked, First coverage, Rank improved. Filter Improved, Declined or All. Each win shows why it moved (an engine started mentioning you, an article was published, a fix task was added), how many scans it has Held, a Stable badge after 3 or more scans in a row, and the AI visits that followed. View evidence compares before and after. Create report makes a shareable results report; Get these wins by email every week sends a Monday digest.

Outreach (moved to Citation Sources)

In Citation Sources → Citation analysis: open a site, and media articles and roundup / comparison articles AI cites there have Outreach, which opens the outreach drawer to draft the email and record Pending / Contacted / Replied / Included / Declined. See Route 3 in the Citation Sources manual.

Common tasks

Re-check one prompt now

Open its Evidence → Scan now. New prompts are queued automatically; you don’t need to start them.

Turn on regular re-scans

  1. Turn on Auto-scan on the Automations page, under the library in the left sidebar (paid plans).
  2. Weekly is enough for most prompts (new prompts start on weekly). Only for a few prompts whose page you just changed and want to watch closely, set Scan frequency to Daily in the evidence drawer: daily means about 30 checks a month, several times the credits, and has a per-plan limit.

Generate articles from gaps

  1. Set Coverage to Uncovered and select the prompts you want written up.
  2. Click Generate strategy article in the bulk bar: the article dialog opens directly and suggests a title, an angle and the expected coverage gain.
  3. Click Generate article in the dialog. The article goes to Content for review.

While Auto-generate articles is on, this button is muted: use one or the other, so the same prompts aren’t written twice. Clicking it asks whether to go to the Automations page to turn Auto-generate articles off.

Fix brand misidentification

Set Coverage to Context issue; the bulk bar shows Fix {n} brand mismatches, which takes you to Answer Assets to write one clarification article covering all of them.

Check whether a change worked

Automations → Re-tests → + New re-test, describe the change, keep the suggested prompts, Start re-test. Come back after the observation window for the result.

Common questions

Why is there no "check all" button? Checking is automatic: new prompts are queued, then Auto-scan or each prompt’s frequency takes over. To re-check one prompt now, use Scan now in the evidence drawer.

Coverage keeps showing "Checking…". The queue runs in order and can take a few minutes with many prompts. Closing the page doesn’t stop it.

An engine is locked. Your plan doesn’t include it. Free and Pro check the standard 7 engines; Grok and Mistral need Business or above; the China-market engines (Doubao, DeepSeek, Qwen and others) are included in Free and in Business and above, not in Pro.

What does "Last check was interrupted" mean? The page was closed during a check. Resume continues from the prompt where it stopped; completed prompts are not charged again.

Troubleshooting

  • A check timed out: the engine was busy. Wait a minute and click Scan now again.
  • Out of credits: top up when the dialog asks, then continue.
  • Daily scan limit reached: set other prompts to Weekly / Monthly, or upgrade.
  • Deleted a prompt by mistake: Recycle bin in the header restores it within 30 days.

Was this article helpful?