Famelark Docs

What is Famelark

When people ask ChatGPT, Google AI Overviews, Perplexity or Gemini "what's the best X", those answers name specific products. Famelark tracks how often the products named are you — and then runs a weekly loop of concrete work to make it happen more often.

It is built for founders and marketers of B2B/SaaS products who see organic search shifting into AI answers and want a system, not a one-off audit.

Quick start

1. Run the free report — on /report, enter your domain (we can draft the description from your site). In about a minute you get your current citation share, no signup required.
2. Sign up— Google, or an emailed one-time code. Your report carries over; you don't retype anything.
3. Onboarding is two fields — domain and one paragraph. We generate your prompt set, suggest competitors and start the first tracking run automatically. First results appear on your Overview within ~20 minutes.
4. Come back weekly — the loop below refreshes every week on its own; your job is the Action Queue.

Citation share — the metric

Citation share is the percentage of tracked buying prompts where an AI engine cites your brand. It is the one number the whole product optimizes.

Methodology: every week we ask each engine every active prompt, several times per prompt (LLM answers are stochastic — a single ask lies). A brand counts as cited when it appears across repeated checks; results are averaged, and the same procedure runs for each competitor so the comparison is honest. Every number in the product discloses its N and time window.

Engines covered: ChatGPT, Google AI Overviews, Perplexity, Gemini and Claude.

Prompts

A prompt is a buying question we track, e.g. "best CRM for a two-person startup". There are two kinds:

• Branded — questions that name you ("is Acme any good?"). They power the recognition stat.
• Category — questions where you compete to be the answer. This is where citation share is won.

Prompts auto-generate at onboarding from your description; you can add, edit, pause or regenerate them on the Prompts page. Your plan sets how many are actively tracked — extras stay paused and unlock instantly on upgrade.

Each prompt is also classified by intent — commercial, informational, transactional or branded — so you can see whether the questions you win are the ones closest to a purchase decision. On the Prompts page, a Position column shows your best and average list position (#1 best · #2.5 avg) across the answers that name you — lower is better.

Prompt Explorer (/prompts/explore) finds new category prompts to track from three sources: a seed query expanded into buying-intent variants, a page on your own site summarized into the questions a buyer of it would ask, and Google's People Also Asked results for a seed query. A Google Search Console import tab is visible but disabled — "coming." Suggestions you add are inserted active while you're under your plan's prompt limit; past the limit they're added paused and unlock on upgrade, same as any other prompt.

Responses

The Responsespage shows the raw answer each engine gave for each tracked prompt — not just whether you were cited, but the actual text. Cards are grouped by prompt, one per engine sample, each with an engine chip, a cited badge ("Cited · #N" when you appear, using list position, or "Not cited"), a sentiment chip when the answer names you, the answer excerpt (first 200 characters, expand for the full text) and its source count.

Filter by engine or by cited/missed, and use the run picker to look at any of the last 8 runs. This is the page to open when a citation-share number moves and you want to read exactly what the AI said.

Sentiment

Every answer that mentions you is also scored Recommended, Neutral or Negative — three stat cards on the Sentimentpage show counts and each bucket's share of the answers that mention you (not of all answers — being cited neutrally is very different from being cited well, and the denominator makes that explicit).

A stacked bar per run shows the trend across your last 12 complete runs, and a per-prompt table lists every mention for the latest run with its engine and sentiment. Sentiment only appears once AI answers actually mention your brand — early on, or on a category prompt you don't yet win, the page is empty by design.

The weekly loop

Track → Diagnose → Act → Prove. Each page owns one step:

• Overview — citation share, trend, competitor bars, this week's movement.
• Sources — the gap map: every page AI engines pull answers from in your category. Red rows feed competitor citations where you're absent — that's exactly where the work is.
• Action Queue — the gaps turned into prioritized moves across three rails: outreach emails (drafted, one click to send), Reddit replies (drafted in your voice), and pages/listings to create or fix. Every draft has a copy button.
• Replies — verification that your posted Reddit replies are live, plus removal watch.
• Impact — before/after attribution: completed actions mapped to the citation-share shift on the exact prompts they touch. Exportable.
• Weekly digest — one email: how your share moved, what you completed, where a competitor gained. Opt-out anytime.

Reddit replies — native, not rented

Reddit is a top source AI engines cite. Famelark finds live threads worth answering (relevance-scored, freshness-checked) and drafts a reply in your voice. You copy, open the thread, paste and post from your own account — about 30 seconds. We never post via API: API-posted content gets removed roughly 8× more often, and a bot farm speaking for you is exactly what gets brands banned.

Guardrails are built in: weekly reply caps by plan, subreddit blacklist, and a promo-density limit so your account history never reads like an ad board. Add your Reddit username (just the name — no password, no OAuth) in Settings and we verify each reply went live and attribute its effect.

Outreach emails

For gaps that live on listicles, review sites and directories, the queue drafts a personalized pitch to the page owner. Sending happens from Famelark with one click (or copy the text and send from your own inbox). Follow-ups are drafted after a week of silence. Every email carries an unsubscribe link, and suppressed addresses are never emailed again — by any rail.

Pages & Content

Some gaps need a page on your site: comparison pages, alternatives pages, integration guides — the formats AI engines cite most. Famelark suggests the exact pages worth creating and generates a full draft on request. You edit and publish on your own site.

Reports & sharing

• Free report — the public snapshot anyone can run on /report; each report gets a shareable link with an OG image.
• Scorecard — one click on your Overview turns the latest weekly run into the same shareable format — for your team, investors or clients.
• Case Study Export & white-label (Scale) — agency-grade before/after case pages under your client's brand, plus 3 team seats.

MCP — "Your AI" mode

On Growth and Scale, your own AI (Claude, ChatGPT or any MCP client) can operate your Famelark account: read the queue, refine drafts, mark work done, trigger outreach, pull the citation report.

Getting your API key:

1. Open Settings and scroll to the "Your AI" mode card (marked with an MCP badge).
2. Click Generate API key. The card shows your key and the MCP endpoint, each with a copy button.
3. If the card says "Available on Growth and Scale", upgrade first — the key is gated by plan.
4. Leaked a key? Rotate key in the same card invalidates the old one instantly.

Connecting (Claude Code example — any MCP client with HTTP transport works):

claude mcp add famelark \
  --transport http https://famelark.com/api/mcp \
  --header "Authorization: Bearer <your-api-key>"

Tools exposed: list_actions, get_action, refine_draft, mark_done, run_outreach, get_citation_report.

Hard rule: no MCP tool posts to Reddit or any social platform. Your AI prepares and manages the work; publishing stays native, from your hands.

AI Traffic

Citation share tells you when an AI engine names you. AI Traffic tells you when that turns into a visit. One script tag counts sessions that land on your site with a referrer from ChatGPT, Perplexity, Gemini, Copilot or Claude — the act→prove loop closed end to end.

Install: grab your tracking key from Settingsand paste this in your site's <head>— it's the same key shown on the AI Traffic page:

<script src="https://famelark.com/fl.js" data-key="YOUR_KEY" defer></script>

To attribute a conversion (signup, trial, purchase) to that same AI-referred session, call window.famelark.convert('signup') from the action that completes it — e.g. an onclick handler or a post-signup redirect page.

Privacy: no cookies, no PII. The snippet only records a visit when the referrer matches a known AI assistant, stores a source tag in sessionStoragefor that tab's session, and never collects anything else about the visitor. Every insert is flood-capped server-side per key, so a misbehaving page can't run up your numbers or anyone else's.

Two notes on the tag itself: we update fl.jsin place — there is deliberately no SRI hash pinning the tag, because that would break tracking on every installed site the moment we ship an update; it's served first-party from famelark.com, the same trust model GA and Plausible use. And the tracking key in the tag is public by nature — like a GA measurement ID, it's meant to sit in your page source; it can only send events, never read your data, and abuse is rate-limited server-side.

REST API

Everything the dashboard shows is also available as JSON — for your own dashboards, scripts, Zapier/Make scenarios or a client report pipeline. Available on Growth and Scale; it uses the same API keyas MCP (Settings → "Your AI" mode), passed as Authorization: Bearer <key> or X-API-Key: <key>.

Example — this week's citation share:

curl https://famelark.com/api/v1/citation-share \
  -H "Authorization: Bearer <your-api-key>"

{
  "brand": "acme",
  "current": { "citationSharePct": 40, "promptsCited": 4, ... },
  "byEngine": [ { "engine": "chatgpt", "citationSharePct": 50, ... } ],
  "brands":   [ { "name": "acme", "isSelf": true, ... } ],
  "history":  [ { "runAt": "...", "citationSharePct": 40, ... } ]
}

Errors come as { "error": { "code", "message" } } with proper HTTP statuses (401 bad key, 403 plan, 429 rate limit — 120 requests/min per key). Unauthenticated GET /api/v1 returns a machine-readable index of all endpoints. Writes are deliberately minimal: like MCP, the API never posts to Reddit or any social platform.

What's new: GET /api/v1/prompts now returns intent, bestPosition and avgPosition per prompt; GET /api/v1/sources returns engines per row; GET /api/v1/citation-share adds deltaPp (week-over-week change in percentage points) alongside the overall share, each engine and each brand.

CSV exports:Prompts, Sources, Action Queue and Impact history can each be downloaded as CSV directly from the app (an "Export CSV" button on each page) — those routes use your logged-in session, not the API key, so they're not part of this REST surface.

Account & login

• Login — emailed one-time code (default), Google, or a password if you set one in Settings. Forgot the password? Log in with a code and set a new one.
• Multiple brands — Scale tracks up to 3 brands; switch the active project in Settings.
• Team seats — Scale includes 3 seats; invite by email in Settings, teammates see the same workspace.
• Weekly digest — toggle in Settings; unsubscribe links work instantly.

Plans & limits

Every plan starts with a 7-day free trial (card required, cancel anytime). The free citation report needs no signup at all. Pricing is on the home page.

FAQ