> ## Documentation Index > Fetch the complete documentation index at: https://docs.trybluemoon.com/llms.txt > Use this file to discover all available pages before exploring further. # Actions, tools, agents and Ask AI > What Bluemoon recommends and how it knows whether it worked, the content and advertising tools built on your answers, the agent workflows, the AI chat — and what each of them costs in AI credits. Everything on this page is available to Admins and Editors. Viewers can read the measurement screens but do not see the Actions and Tools groups. ## Actions ### Earned An **earned action** is one source, on one theme, where the AI engines already cite that source and do not name you. Do the action, and you get named in those answers. What to do follows what the source is: out-rank a rival, be present on a community, pitch a publisher, get listed by a retailer, buy a placement on display inventory. A source nothing has classified yet is held and counted out loud, never handed a generic playbook. * The catalogue is rebuilt from your latest answers and citations every few hours; a market with nothing worth recommending says so. * Priority comes from rank, not from a threshold; start with **High**. * Status moves **Suggested → To do → Pitched → Landed or Stalled**. Marking an action *Pitched* freezes the before-picture; whether it landed is then **detected** from the source's later citations, not declared by you. * Each action opens on its evidence: the answers behind it, the pages the AI reads on that source, and the playbook, with a first draft to edit — never to publish as is. ### Owned **Owned actions** target your own site: content to publish or change for questions where you are absent. When you mark one **Done**, Bluemoon captures a before-picture of the questions it targets over the trailing 21 days, plus a control cohort of your other themes, and later reports the difference-in-differences with a confidence interval. See [Measure whether an action worked](/guides/action-impact). ### Impact The **Impact** page shows, per shipped action, the share of answers naming you on the targeted questions before and after, and a chart of the targeted questions over time. An action is measured 14 days after it is marked done; it reads "not enough data" below 8 answers on either side. ## Tools ### Create articles Start from a **gap** — a theme where a competitor wins the AI answers and you do not, or a Google search you already rank for whose matching question scores low — and Bluemoon writes a **brief** from the AI answers on that theme: what the draft has to cover to get cited, the keywords, AI terms and brands the answers contain. The draft is scored against the brief as you type; the relevance total needs one AI run. Publish on your own domain (the strongest signal for a citation) or on an external site the answers already cite. Published articles are then tracked for AI citations. ### Optimize content Two views over the answers of the last N days: your **pages**, biggest first, with what AI takes from each and which questions it answers; and your **questions**, competitors first, with who AI cites instead of you. ### Ads generation Pick a theme and Bluemoon writes ad copy from the real answers AI gives about your brand: ads for ChatGPT, or responsive search ads for Google. Kept ChatGPT ads go to your OpenAI Ads account as a **paused draft campaign** — nothing is published from Bluemoon (see [OpenAI Ads](/integrations/openai-ads)). Google Ads generation is not live yet: a saved Google set can be reviewed and copied, nothing is exported. ### Ad targeting Ranks the sources AI cites on your questions by how you would reach them: **buy an ad** (self-serve), **pitch a placement** (sponsored), or **out-rank with content** (earned), and names who sells each source's inventory, read from the source's public `ads.txt`. ### Technical SEO Audits which AI crawlers your domain lets in, reading your `robots.txt` (an absent rule is permission), and generates an `llms.txt` for you to upload to your site root. ### Agents Recipes that read your Bluemoon data and act on your gaps: an AEO FAQ generator, a topic gap filler, a competitor counter-page, a content brief, a schema audit. Pick a template and it becomes yours to edit on a canvas, step by step; or describe the goal and let the **architect** assemble the steps. An agent runs by hand or on a cron schedule (UTC). Each run costs AI credits; a finished run's files are downloaded from the editor. ### Ask AI A chat that answers questions about your AI visibility from your last scan, with a fast model and a deeper one. It starts from where the AIs find you and the questions where rivals win and you are absent. Credits scale with message length and model. ## AI credits AI-powered actions draw on a monthly allowance of credits included in the plan, plus any top-up you bought (top-ups never expire; the monthly allowance resets on the plan's cycle). The costs, per action: | Action | Credits | | ---------------------------------- | -------------------------------------------: | | Ask AI message | 1 | | Agent run | 2, or the cost the agent's own steps declare | | Generate questions for a theme | 2 | | Generate a theme | 3 | | Generate a persona | 5 | | Design an agent with the architect | 5 | | Generate `llms.txt` | 10 | | Generate ads | 15 | | Edit an article with AI | 15 | | AI insights | 25 | | Onboarding kit (setup generation) | 30 | | Generate an article | 50 | Actions above a threshold ask for confirmation first. The balance and every debit and refund are on **Settings → Billing**. A workspace that brings its own provider key (**Settings → API keys**) is not debited for the actions that run on that provider: the calls are billed by the provider directly.