> ## 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.

# Tasks

> Your weekly queue of actions to improve AI visibility, grouped by work stream and ranked by priority.

**Tasks** is where the rest of the dashboard turns into work. Each week, Cited reads your latest [Source Intelligence](/dashboard/source-intelligence) analysis and generates a set of concrete actions, ranked so you know what to do first and grouped so the right person can pick up the right pile.

The queue regenerates weekly rather than accumulating. That is deliberate: a task list that only grows becomes a graveyard, and the point is to reflect the opportunities that are live *now*.

## Work streams

Tasks are grouped by the kind of work they are, because they usually belong to different people:

| Work stream   | What lives here                                                                |
| ------------- | ------------------------------------------------------------------------------ |
| **Onsite**    | Changes to your own pages — content, structure, depth, answer formatting       |
| **Offsite**   | Getting cited somewhere you do not control — PR, outreach, editorial placement |
| **Community** | Presence in forums, communities and social sources AI answers draw on          |
| **Technical** | Crawlability, structured data, and the mechanics of being readable to AI       |

A brand with a healthy Onsite pile and an empty Offsite one is optimising the half of the problem it controls. That imbalance is itself worth noticing.

## Reading a task

Each card carries:

* **Priority** — High, Medium or Low. See [how priority is decided](#how-priority-is-decided) below.
* **Effort** — a rough sense of what the task costs to do.
* **Where it came from** — *Recommended from your visibility data* for tasks Cited generated, or **Custom** for ones your team added.
* **Stale** — shown when the data that produced a task has moved on. You can keep a stale task if it is still worth doing; the badge is a prompt to look, not an instruction to delete.

## The summary strip

|                          | What it tells you                                             |
| ------------------------ | ------------------------------------------------------------- |
| **Active / WIP Limit**   | Tasks in progress or to-do, against your work-in-progress cap |
| **By Work Stream**       | Where the queue is concentrated                               |
| **Completed This Month** | Throughput                                                    |
| **High Priority**        | How much of the queue is urgent                               |

The **WIP limit** is the one worth respecting. AI visibility work has a long feedback loop — content published this week may not change an answer for weeks — so a large number of simultaneously-open tasks tends to mean nothing finishes rather than everything progresses.

## How priority is decided

Priority is computed from your own data, not from a generic template. Four things feed it:

* **How much of your prompt library the gap touches** — a gap across 30 prompts outranks one across 3
* **How big the competitive gap is** — and whether you are behind (offensive) or ahead but closing in on (defensive)
* **How actionable the fix is** — a technical change you can ship this week outranks a multi-quarter authority play
* **How valuable the prompts are** — commercial and transactional intent outranks purely informational

It is **brand-relative**: High means high priority on *your* brand's scale, not against every other Cited customer. If your positioning is declared in **Settings → Brand Intent**, tasks matching that direction are lifted, so the gaps that matter to your strategy surface first.

The full mechanism is in [Impact Scoring](/methodology/impact-scoring). You do not need to read those numbers — the priority on the card is their output.

## Working the queue

**Take the top of a work stream, not the top of the list.** The global ordering is honest, but if Offsite work needs a different person than Onsite, priority order across the whole queue is not a schedule.

**Respect the WIP limit before adding.** If the strip shows you at your cap, the useful action is finishing something, not starting something.

**Add your own tasks.** The generated queue reflects what Cited can measure. Things it cannot see — a campaign already in flight, a constraint from legal, a partnership in progress — belong in the queue too, as Custom tasks, so the board reflects reality.

**Import in bulk if you already have a plan.** A CSV import exists for teams that plan elsewhere and want one board.

## From a task to the work

Tasks that an agent can execute carry a **Run as Agent** action in the detail view. That takes a task like "this topic needs a page that answers the comparison question" and produces a writer-ready brief grounded in your own citation data — rather than making you restate the problem to a separate tool.

See [Agents](/dashboard/agents) for what each one produces.

## Export

The queue exports with your current filters applied, so what you export is what you are looking at.

## Related

* [Source Intelligence](/dashboard/source-intelligence) — the analysis tasks are generated from
* [Impact Scoring](/methodology/impact-scoring) — how priority is computed
* [Agents](/dashboard/agents) — executing a task rather than only tracking it
* [Configure positioning](/get-started/configure-positioning) — how to make priority reflect your strategy
* [Close competitor gaps](/playbooks/competitor-gap-workflow) — the workflow behind many Offsite tasks