For in-house marketing teams

Your customers increasingly ask an AI assistant before they ask a search box. When someone types "best [your category] for [their use case]" into ChatGPT or gets a Google AI Overview, either your brand is in the answer or a competitor is. An in-house team already owns the org's search visibility; MentionFlow gives you the same ownership over its AI answer visibility — the questions you win, the rivals taking ground, and a prioritized list of what to fix — measured from the answers engines actually give, not a model's guess about them.

This guide is the in-house playbook: what to set up, what to read each week, and how to turn a gap into a shipped fix.

Set up what you want to be found for

Start by deciding the questions that matter. A monitor is your sampling configuration — one brand, one market and language, a chosen set of engines, and a set of prompts — collected on your plan's cadence (daily on Growth and above). The prompt portfolio is where those questions live, and it is built to be organized like a real content operation:

• Topics, tags, and funnel stages group prompts so you can read visibility by campaign, by product line, or by top-/mid-/bottom-of-funnel.
• Category prompts — questions that don't name you or a named rival — are the honest test of whether you get recommended when nobody put your name in the question. The Overview has a "Category prompts only" toggle for exactly this view.
• As the list grows, batch actions and search let you retag, re-topic, or relocate hundreds of prompts at once.

Read the scorecard

The Overview is your weekly standup screen — one rolling window as five KPI scorecards plus a content-coverage tile:

• Visibility (0–100) — how prominently answers name you, weighted by position.
• Share of Voice — your slice of the answer versus your competitor set, as a zero-sum split.
• Citation Share, Sentiment, and Estimated impressions round out the headline.

Every headline number blends only the search-grounded engines — the ones that run live web searches — so the figure keeps one consistent meaning; the knowledge-only engines (DeepSeek, Mistral) are segmented out, never silently averaged in. The same scope feeds the API, so a dashboard and an export never disagree. A metric built from fewer than five runs is badged low-confidence, and a gap in collection draws a null day, never a fabricated dip — judge trends, not ticks.

One signal worth building a habit around: on the "You vs. competitors" trend, switch to the Visibility view and watch the dashed "your site cited" line against the solid name line. When the dashed line runs above the solid one, AI assistants are reading your site but not saying your name — a direct brief for your content and PR teams.

Watch the competitive set

Competitors answers the question a CMO actually asks: who is winning the answer, and where am I losing it?

• The leaderboard ranks you and your rivals by Share of Voice; a trend and an engine heatmap show how that split moves over time and by engine.
• The answer-gap matrix flags prompts where a rival clearly beats you — a gap fires when a prompt has at least three runs, your presence is below 0.2, and some rival's is at least 0.6 — and links each gap to an action or a content brief, so it becomes a task, not a chart.
• Discovery surfaces rivals you never entered but that keep appearing in answers, so your set stays current.

If engines call a competitor by a nickname, product name, or misspelling, teach MentionFlow with Matching rules so those mentions are counted correctly — and re-scan history to apply the rule to answers you already collected.

Work the fixes

Actions turns the deterministic recommendation engine into a persistent kanban board. Fourteen pure rules score each recommendation — priority = round(100 × impact × confidence), with confidence saturating at five runs so a thin sample can't out-rank a solid one — and cards move through Suggested → Planned → In progress → Done. The board auto-closes a card, marking it "verified by data", when the engine stops emitting it, and un-resolves one if the gap comes back. Each card offers Evidence, Source, and a one-click content brief grounded in your real answers and knowledge base. Actions is available on every plan.

Ground demand in your own search data

Connect Google Search Console and two things sharpen from modeled estimates into measured facts:

• Measured demand tiers. When a query behind a prompt appears in your own Search Console data, its volume is labelled measured rather than modeled — MentionFlow never sums your GSC impressions into a volume band, it labels which is which (see the demand confidence tiers).
• Striking distance. Research surfaces the queries you already rank 8–20 for with real impressions — the near-misses where a small push could break you into the answer engines' source set.

A connection is read-only, gated to an owner or admin, and a property is rejected unless it covers your brand's domain, so the join can only ever use pages you actually own. Disconnecting demotes any "measured" evidence back to an honest em-dash.

A weekly rhythm

1. Open the Overview, toggle Category prompts only, and read the five scorecards against last period.
2. Check Competitors for any rival that moved, and send fresh gaps to the board.
3. Work the top of the Actions board; spin a brief for the highest-priority content gap.
4. Skim Sources for a new domain or community thread shaping your answers — an "opportunity" flag is a placement to chase.
5. When a number surprises you, open the receipt behind it and read exactly what the engine said.

Related

• Dashboard overview — the weekly scorecard.
• Competitors — the Share-of-Voice leaderboard and gap matrix.
• Actions — the scored, self-closing action board.
• Google Search Console — measured demand and striking distance.
• Metrics overview — every headline number, defined.
• Content — where a gap becomes a drafted, tracked piece.