Core concepts

A handful of nouns carry the whole product. Learn these and every screen reads clearly.

The entity model

A workspace holds brands. A brand has one collection config and many prompts. A prompt produces runs. A run produces mentions, citations, and shopping placements.

• Workspace is the account. It holds your plan, billing, members, and referral code, and owns one or more brands. Quotas (prompts, seats, engine slots) are pooled here.
• Brand (also called a project) is the thing you track: a name plus a primary domain. Every metric is computed per brand.
• Collection config is how a brand is sampled: its markets (up to five, each a country plus a language), an optional persona, and the engines it collects with. Every brand has exactly one, created with the brand and edited in Collection. Projects inherit the workspace default engine set from Engines. A project can Customize its own set within the same slots, and Reset to inherit again. (The data model and the CSV export call it a monitor.)
• Prompt is a plain-language question you want to watch, for example "best CRM for startups". Prompts belong to a brand. You can group them into topics (one prompt can sit in several) and tag them.
• Run is the smallest unit of data: one prompt, one engine, one region, one UTC day. It stores the answer word for word, the raw payload, the model version, and the web searches the engine ran. Runs are idempotent, so the same prompt, engine, region and day is only ever collected once.

What a run produces

• Mentions: every brand or product named in the answer, with its order (1-based position of first mention) and a sentiment (positive / neutral / negative, or null when unscored). Sentiment is never guessed. If a run's language model pass failed, it is marked "unscored", not "neutral".
• Citations: the source links the engine attached, each sorted into a class (owned, competitor, editorial, UGC, reference, social, other).
• Shopping placements: for shopping prompts, each product shown in an AI shopping carousel, with its slot position.

Receipt

A receipt is a finished run shown in full: the answer as the engine wrote it, the ranked list of brand mentions, the citations, and any ad or shopping surface. Every metric traces back to a receipt, so you can audit any number down to the answer behind it. See Answers and receipts.

Fan-out

A fan-out is a real web-search query an AI assistant ran while answering your prompt. It is lifted word for word from the stored payload, never simulated. Each one can become a tracked prompt in one click. See Fan-outs.

Carousel

In shopping answers, a carousel is one product shelf the engine drew. A placement is one product in one slot of that carousel. Shopping metrics are computed over carousels, not raw placements. See Shopping overview.

Engines and engine classes

MentionFlow samples ten answer engines: ChatGPT, Perplexity, Gemini, Google AI Overviews (AIO), Google AI Mode, Claude, Copilot, DeepSeek, Grok, and Mistral. Each has its own production adapter, and they reach the answer differently:

They fall into two classes:

• Search-grounded: engines that run live web searches (ChatGPT, Perplexity, Gemini, AIO, AI Mode, Claude, Copilot, Grok). These carry citations.
• Knowledge-only: engines that answer from model memory with no web search (DeepSeek, Mistral). Their citations are honestly empty.

The split matters for the numbers. Headline blended metrics (Visibility and Share of Voice on the Overview) include only search-grounded engines, so the numbers keep one steady meaning. A few surfaces widen or narrow that set on purpose, and Data honesty names every one.

Windows

Metrics are computed over a rolling window of days. Two rules make windows trustworthy:

• A window anchors at your newest collected day, not at the wall-clock "today". A gap in collection never quietly stretches a window.
• Missing days are filled with null, never zero. A chart never draws a made-up dip through a day you did not collect.

The dashboard default window is 7 days. You can switch to 14, 28, 90, or 180 days, or set a custom range up to 200 days. One picker in the top bar governs every data page at once. See Date ranges and the top-bar picker.

Confidence

A metric built from fewer than five runs is flagged low confidence. Volume estimates are always shown as approximate bands, never exact counts, because AI-search demand cannot be measured that precisely. See Data honesty.