> ## Documentation Index > Fetch the complete documentation index at: https://docs.getcited.in/llms.txt > Use this file to discover all available pages before exploring further. # Agents > Run agents over your own AI-visibility data to produce briefs, reports, FAQs, schema and outreach targets — grounded in what Cited already measured. The rest of the dashboard tells you what is wrong. **Agents** does some of the work. An agent takes your own Cited data — your citations, your gaps, your competitors, your sentiment — runs a fixed multi-step pipeline over it, and produces a finished **artifact**: a document, a report, a spreadsheet, or a block of code you can paste. The distinction that matters is the grounding. A general-purpose AI writing tool can produce a content brief, but it is guessing at your category. A Cited agent building the same brief already knows which prompts you lose, which sources AI actually cites in your category, and which competitor holds the position you want. The output is specific to your measured reality rather than to a plausible-sounding average. ## The library Agents are grouped by where they sit in the marketing lifecycle. ### Research | Agent | What it produces | | ------------------------- | ------------------------------------------------------------------------------------------------------------------------------------ | | **Top Cited Sources** | The sources AI answers pull from most in your category, ranked, with how often you versus your competitors appear on each | | **Citation Gap Analysis** | The topics where competitors are cited and you are absent — your share versus theirs, per platform | | **Competitor Teardown** | Where your biggest competitor wins AI citations: their top sources, their content mix, and the sources they win where you are absent | ### Strategy | Agent | What it produces | | ----------------- | ----------------------------------------------------------------------------------------------------------------------- | | **Content Brief** | A writer-ready, AEO-optimized brief from a task or any topic — outline, target prompts, FAQs, and your top AI citations | ### Create | Agent | What it produces | | -------------------- | ------------------------------------------------------------------------------------------------------------- | | **Generate Article** | A full AEO-optimized article on a topic — direct-answer intro, structured sections, FAQ — in your brand voice | | **FAQ Generator** | An AEO-optimized FAQ section of 6–10 Q\&As for a page, ready to drop in with FAQPage schema | ### Optimize | Agent | What it produces | | --------------------------- | ------------------------------------------------------------------------------------ | | **Page AEO Analysis** | A 0–100 answer-engine readiness score for any page, with a prioritized list of fixes | | **Schema Markup Generator** | Ready-to-paste schema.org JSON-LD structured data for a page | ### Distribute | Agent | What it produces | | ------------------------ | ----------------------------------------------------------------------------------------------------------------------- | | **Source Target Finder** | The off-site sources citing your competitors but not you — your PR and outreach targets, ranked, with the play for each | ### Measure | Agent | What it produces | | ----------------------- | ---------------------------------------------------------------------------------------------------------------------------------------- | | **Brand Health Report** | Your weekly AI-visibility pulse — citation share and how it moved, who is ahead, per-platform strengths, sentiment, and what to focus on | ## How a run works An agent is a **deterministic pipeline**, not a single prompt. Each one executes a fixed sequence of steps, and you watch it run as a workflow graph — every step visible, so you can see which inputs produced which part of the output. That visibility is the point. When an agent tells you a source is worth pursuing, you can trace the step that decided it. An artifact you cannot interrogate is one you have to take on faith. Every run is saved. Artifacts land in a versioned repository, so re-running an agent next month gives you a new version alongside the old one rather than overwriting it — and you can see what changed. ## Two ways in * **From the Agents tab** — browse or search the library and run an agent directly. The search covers what an agent does and what it outputs, not just its name. * **From a task** — tasks that an agent can execute carry a **Run agent** action, so you can go from "this is the gap" to a finished brief without restating the problem. The second is the loop the product is built around: measurement finds the gap, [Tasks](/dashboard/tasks) ranks it, an agent does the first draft of the work. ## Credits Agent runs are metered in **credits**. Different agents cost different amounts, reflecting how much work the pipeline does — a schema generator is cheap, a full article is not. Your balance is shown in the Agents tab. Credits meter runs, not artifacts: an artifact you generated stays available in the repository regardless of your current balance. ## How to get a good result **Point agents at real gaps, not hypotheses.** The agents that read your data are only as good as the question. Running Content Brief on a topic you already dominate produces a competent brief for work you did not need. **Start with a Research agent.** Top Cited Sources and Citation Gap Analysis are cheap ways to find out where the actual opportunity is, before spending a larger run on producing something. **Treat the output as a first draft with evidence.** These artifacts are grounded in your data, which makes them a much better starting point than a blank page — but they are a starting point. The judgment about whether a topic is worth your quarter is still yours. ## Related * [Tasks](/dashboard/tasks) — where an agent run usually starts * [Win editorial coverage that LLMs cite](/playbooks/win-editorial-coverage) — the manual version of Source Target Finder * [Write content that gets cited](/playbooks/write-citable-content) — what Content Brief and Generate Article are optimizing for * [Cited MCP](/mcp/overview) — the other programmatic surface, for querying rather than producing