AI Engines and Scans — Engines per Market, Repeated Sampling, Data Sources and History

Engines

Engines outside your plan show a lock in every platform filter. Google AI Overviews and AI Mode read real Google results.

Which engines each market scans

Every market has a country; there is no Global market. AI engines always answer from somewhere, so every scan asks in the market's country and language. Each market has its own engine set:

• China markets scan only Doubao, DeepSeek, Qwen and Kimi, so Chinese prompts are never sent to ChatGPT and the like. Pro does not include those engines: China-market prompts show "Business only" and are not scanned until you upgrade.
• Other markets scan the standard 7 by default and never the China-only Doubao, Qwen and Kimi. Grok, DeepSeek and Mistral can be added per market (Business and above).

Change it under Prompt Library → My Prompts → "⚙ Markets & engines": By market lets you tick engines market by market, and "Also apply to other markets" changes several at once; By engine is a markets × engines grid that shows at a glance where each engine runs. Manual scans, single re-checks and auto-scan all follow the market's engine set, and each Prompt Tracking row shows only the engines its market scans.

How a check runs

For each prompt, Citadex builds the query in the market's language and country (discovery prompts get a "list the top 5 to 10 with names and key features" instruction in that language; brand prompts are sent as is) and asks every enabled engine in parallel. The answer is parsed for mention, position, recommendation level, citations, competitors and sentiment, and the raw answer is archived together with the searches the engine ran before answering (AI searches). New prompts are queued automatically; you can also re-run one with "⟳ Scan now" in the evidence drawer.

Data source and model

Every answer records its source: captured from the AI app itself (what users see) or the model API, plus the model name. The evidence drawer shows it above the answer, and the Workspace From mention to recommendation card states it too. Answers scanned before this was recorded show no source.

Repeated sampling

AI answers vary from one run to the next. Turn on "Ask 3× per scan" in the evidence drawer and each engine answers that prompt three times per scan; the result shows mentions out of 3 and the number of recommendations, so you can see how often the brand appears. Available on Business and agency plans, with a cap on how many prompts can use it. It raises that prompt's credit use in line with the number of runs, a dialog asks you to confirm before it turns on, the credit check before each scan includes the extra runs, and you can turn it off at any time.

Platform-wide changes and model updates

Every day Citadex aggregates anonymous metrics per AI engine across all projects (cited links per answer, brands named per answer). When an engine changes noticeably as a whole, or starts using a new model:

• the Workspace shows the Platform-wide changes card;
• the prompt's coverage trend marks the day;
• in This week’s tasks, drops during a platform-wide change count half and say so.

When your numbers move on the same day and the same engine as the platform, the cause is usually the platform, not your brand.

Auto-scan

Turn on Auto-scan in the Prompt Tracking header (paid plans). Each prompt then refreshes on its own frequency: daily, weekly (default) or monthly, set in the evidence drawer. The number of daily prompts is capped per plan, and daily runs refresh a subset of engines; the full engine set runs in the weekly and monthly passes. When credits cannot cover the next scheduled scan, auto-scan pauses and resumes after a top-up.

Snapshots and history

Automatic scans are grouped into one snapshot per project per day; the snapshot picker on Prompt Tracking shows earlier states. Every raw answer is kept and can be read by date under Earlier answers in the evidence drawer. History is kept for 30 days on Free and 365 days on paid plans. Deleted prompts stay in the recycle bin for 30 days.

Frequently asked questions

Why does a row show fewer engines than my plan? Each row shows only the engines its market scans; also, an engine that timed out or hit a quota is left out of that scan's coverage and filled in on the next scan.

Can it scan more often than daily? No. For a one-off re-check use Scan now; manual re-checks of the same prompt have a cool-down.

I see a "data gap" notice. At least half of one engine's latest checks returned no answer. That is a collection problem, not a visibility drop: those results are left out of the mention rate for now and filled in once collection recovers.