ChatFeatured

View saved version →

---
name: aeo-visibility-report
description: Build an AI search visibility report for a brand — how often it appears in ChatGPT, Perplexity, Gemini, and other AI answers, how it compares to competitors, which sources cite it, and where it's losing ground. Use when the user asks about AI visibility, brand mentions in AI answers, share of voice, competitor comparison, or citation analysis. Not for website technical health (use the site audit) or for creating content.
---

# AEO Visibility Report

Produce a decision-ready report on a brand's AI search visibility using ChatFeatured analytics.

## Steps

1. **Resolve the brand.** Call `list_organizations`, then `list_brands`. If the user has exactly one brand, use it; otherwise ask which brand to analyze. Pass its `brandId` to every call below.
2. **Pull the core analytics** with `query_analytics`, defaulting to a 30-day window (`filters: { dateRange: "30d" }`) unless the user asks for another period:
   - `brand_overview` — headline visibility, mention counts, and platform mix.
   - `time_comparison` — current period vs the previous one; this is where trends come from.
   - `competitor_comparison` — share of voice against tracked competitors.
   - `source_analysis` and `cited_urls` — which domains and URLs AI answers cite in this space, and whether the brand's own pages are among them.
3. **Go deeper only where the data warrants it.** If a preset surfaces something notable (a competitor spike, a platform drop), use `query_analytics` with `source_changes` or `source_mentions`, or `query_analytics_sql` for a custom cut (call it with `explain: true` first to get the schema). Skip this step for a routine report.
4. **Synthesize.** Structure the report as:
   - Headline: overall visibility and the single biggest change since the prior period.
   - Platform breakdown: where the brand is strong/weak (ChatGPT vs Perplexity vs Gemini, etc.).
   - Competitive position: who leads share of voice and by how much.
   - Citation gaps: domains AI answers trust in this space where the brand is absent — these are the actionable opportunities.
   - 2–4 concrete recommendations tied to the gaps found.

## Boundaries

- Report numbers exactly as returned; never extrapolate missing periods.
- If a preset returns zero rows, say the data window is empty rather than inventing a trend, and suggest the user check that prompt tracking is running.
- Recommendations that involve creating content should point at the content-gap workflow rather than starting article generation from here.

View saved version →

---
name: content-gap-to-article
description: Find content gaps hurting a brand's AI search visibility, then create, generate, and publish SEO/AEO-optimized articles to fill them — end to end, from gap analysis through CMS publishing. Use when the user wants to improve AI visibility with content, write articles, fill citation gaps, or publish to WordPress, Ghost, Webflow, Framer, or the ChatFeatured CMS. Not for analytics-only questions (use the visibility report).
---

# Content Gap → Published Article

Run ChatFeatured's content pipeline: identify what content is missing, create targeted article ideas, generate them, and publish.

## Steps

1. **Resolve the brand** via `list_organizations` → `list_brands`; confirm with the user if they have several brands.
2. **Find the gaps.** Use `query_analytics` (`source_mentions`, `cited_urls`) to see which topics and sources AI answers cite where the brand is absent, and `get_existing_opportunities` for gaps ChatFeatured has already identified. `list_articles` with `status: "published"` shows what the brand has already covered — don't propose duplicates.
3. **Create ideas** with `create_article_ideas`, targeting the specific gaps found. Show the user the ideas before going further.
4. **Approve with consent.** `approve_articles` consumes the plan's monthly article quota — tell the user how many articles this will use, and check the `articleLimits` (used/limit/remaining) in the response. If the quota is exhausted, stop and say so.
5. **Generate** with `generate_articles`. Generation is asynchronous: poll `list_articles` with `status: "review"` to see when drafts are ready (typically minutes). Tell the user they can come back later rather than blocking on it.
6. **Review pass.** When a draft is ready, fetch it with `get_article` and summarize it for the user. Apply requested edits with `update_article_content`.
7. **Optional featured image.** `search_pexels_images` for a relevant photo, then `set_article_featured_image` with the photo's attribution fields.
8. **Publish — only with explicit user confirmation.** `list_integrations` shows connected CMS platforms; if there are several, ask which one and pass its `integrationId`. Use `publish_article` for immediate publishing or `schedule_article` for a future date, then confirm success with `get_publish_status`.

## Boundaries

- Never call `approve_articles`, `publish_article`, or `schedule_article` without the user explicitly agreeing in this conversation — approval spends quota and publishing writes to their public website.
- If no CMS integration is connected, complete the draft and tell the user how to connect one instead of failing.
- One pipeline run per request; don't loop generating more articles than the user asked for.

View saved version →

---
name: site-audit-triage
description: Triage a brand website's AEO health — technical, content, and AI-optimization issues that stop AI engines from crawling, understanding, and citing the site — and produce a prioritized fix plan. Use when the user asks to audit their site, asks why AI answers don't cite their pages, or wants technical/content/AEO issues found and ranked. Not for visibility analytics (use the visibility report) or for writing articles.
---

# Site Audit Triage

Turn ChatFeatured's site audit data into a prioritized, actionable fix plan.

## Steps

1. **Resolve the brand** via `list_organizations` → `list_brands`.
2. **Get the lay of the land** with `query_site_audit` `query: "overview"` — audit status, page counts, average scores, and open-issue counts by category. If no audit has run yet (check `run_status` when the overview looks empty), tell the user to start one from the ChatFeatured dashboard and stop.
3. **Enumerate the problems:**
   - `open_issues` — open issues grouped by category and check; run it per category (`technical`, `content`, `aeo`) if the user cares about one area.
   - `pages_needing_improvement` — the lowest-scoring pages with open issues.
4. **Drill into the worst pages** with `page_detail` for the 3–5 lowest scorers (pass each `pageUrl`) to see exactly which checks fail on them.
5. **Also check crawler access** with `query_crawler_analytics` (`stats`, then `top_pages`): if AI bots barely crawl the site, access problems outrank on-page fixes.
6. **Deliver a fix plan** ordered by impact:
   - Blockers first: anything stopping AI crawlers from reaching or reading pages (robots rules, timeouts, thin/duplicate content on key pages).
   - Then high-traffic pages with low scores — fixes there pay off fastest.
   - Then category-wide patterns (e.g. missing structured data across the blog) as a single batch item, not per-page noise.
   - For each item: what's wrong, which pages, and the concrete fix.

## Boundaries

- This skill reads audit data; it cannot start audits or change the website. Say so when the user asks you to "fix" something — hand them the exact change to make instead.
- Score numbers and issue counts come from the tool responses verbatim; don't estimate scores for unaudited pages.
- If the user wants new content to fix content gaps, route them to the content-gap workflow.