FAQ

Short answers to the questions we hear most. For the longer, symptom-by-symptom version, see the Troubleshooting guide.

Collection and engines

How often does MentionFlow collect?

On a schedule, not on demand. Trial and Starter collect 3×/week — Monday, Wednesday, and Friday (UTC); Growth, Agency, and Scale collect daily. A new monitor waits for the next scheduled cycle rather than running immediately. See Plans and quotas.

Why can't I add Claude or Grok as an engine?

Claude isn't a selectable slot — it runs weekly (Mondays, UTC) on the Scale plan only, because it costs far more per answer and has a small consumer-search share. Grok isn't selectable yet; it's planned as a paid add-on. Your selectable engines are ChatGPT, Perplexity, Gemini, AI Overviews, AI Mode, Copilot, DeepSeek, and Mistral. See Monitors.

Why do DeepSeek and Mistral have no citations?

They're knowledge-only engines — they answer from model memory with no live web search, so there's nothing to cite. Their empty citations are honest, not a bug. The other engines are search-grounded and carry citations. See Core concepts.

Why do my headline numbers differ from a single engine's?

Headline metrics blend only the search-grounded engines (the ones that run live web searches — never the knowledge-only DeepSeek or Mistral) so the number keeps one meaning; filtering to a single engine shows that engine on its own, with a different denominator. Monitor cards and prompt-detail pages use their own scopes too. Every deliberate difference is named on the data-honesty page.

Can I track the same question from another country?

Yes — set a per-prompt location on any plan. But only location-capable engines honor it: ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, and Copilot. Claude, DeepSeek, Grok, and Mistral answer via APIs with no location control and keep the project's region — and MentionFlow records them under the region they truly ran in.

Metrics and data

Why is a metric an em-dash (—) instead of a number?

An em-dash means "not enough to say yet"; a 0 means a measured zero. Sentiment is an em-dash until a scored mention exists (and unscored is not neutral), citation share until a run carried citations, clicks until your tracking snippet fires. This is the core data-honesty rule.

Why does a change say "new" instead of a percentage?

There's no prior window to compare against (or the previous value was zero). MentionFlow shows "new" rather than fabricate an infinite or made-up percentage. Once a real prior window exists, the delta appears.

Why do some answers show as failed or queued?

Runs are queued, completed, or failed, and only completed runs feed metrics. Incomplete runs sit in a separate strip in the Answers feed and never count — a failed run is held out rather than miscounted as "brand absent". A few failures per cycle are normal and self-correct next cycle.

How far back does my data go?

The data horizon is 56 days — windows can't reach further, and a receipt older than 56 days returns a 404.

Prompts and quota

Why does a paused prompt still count against my quota?

By design. Active and paused prompts both hold a slot; only archiving frees one. If pausing freed slots, you could pause 50, add 50, and resume to run 100 on a 50-prompt plan. Archive to make room — archived prompts keep their history and are re-importable. See Plans and quotas.

My CSV import was rejected — why?

The prompts importer is all-or-nothing against your pooled quota (up to 500 rows / 2 MB): if the whole batch won't fit, nothing is written. Duplicates are matched on normalized text against active and paused prompts. Free slots by archiving, then re-import. Imports need editor rights and are blocked in demo. See CSV imports.

Reports and sharing

Why does a scheduled report say "Generated" instead of "Sent"?

Because it was built but the email was deliberately held: outbound delivery activates only once your workspace's sending domain is verified with the email provider. Until then every run still generates the report and surfaces the link (and posts to a configured Slack webhook). Runs are marked Generated, Delivered, or Failed — never a fake "sent". See Reports.

Can a client viewer change anything?

No. Client viewer is read-only, enforced on the server — hiding a button is never the only fence. It's how you share a live dashboard without granting edit rights. See Workspace and members.

Why is email verification blocking my API key or schedule?

For accounts created on or after 2026-07-11, the three actions that send workspace data out of MentionFlow — creating an API key, enabling a scheduled report, and setting a notification webhook — require a verified email. Everything else works unverified. Actions that only reduce egress (pausing, revoking) are never gated. Older accounts are grandfathered. See Password and email verification.

API and access

How do I authenticate the API?

Send your workspace key as a bearer token on every request: Authorization: Bearer mf_.... There's no X-API-Key header and no query-string key. A full scope reads the v1 API and MCP; an ingest scope can only post crawler events. See Authentication.

What's the API rate limit?

120 requests per minute per key (a token bucket; the REST reads and the MCP endpoint share one bucket per key). A refusal returns 429 with Retry-After and X-RateLimit-* headers. Reads are cached with a one-hour TTL, so polling faster than hourly returns the same window. See Limits.

Which plan do I need for agent analytics, Claude, or the portfolio?

Agent (crawler) analytics is a Growth-plan feature; Claude runs weekly on Scale only; the Portfolio rollup appears once a workspace has two or more brands (Growth allows 3, Agency 15, Scale 1,000). See Plans and quotas.

Demo and availability

Is the demo data mine?

No — it's a deterministic sample dataset, badged "Demo data" on every page. MentionFlow shows it when there's no database configured, the database is unreachable, or no brand is visible to your session (a signed-out visitor, or a brand-new workspace). Sign in to a workspace that owns a brand to see your real data. See What is MentionFlow.

Is there a video library?

Not yet as a standalone library. For guided, interactive walkthroughs of each surface, the in-app Learn academy is the closest thing today, and these docs are the written reference above it. A video library is on the roadmap.

Didn't find it?

The Troubleshooting guide has the longer, symptom-by-symptom version, and most "why does it say that?" moments come back to one principle: MentionFlow shows an honest gap before a fabricated number.