# Citability Audit > Audit a page for citability — AI-crawler access, llms.txt, answer-shaped headings, structured data, classic on-page SEO — then apply the ordered fix list in the codebase and re-audit. $0.01 per audit. - Skill: `zaialamm/citability-audit` (Agent Skill) - Install (CLI): `npx skillmds@latest add zaialamm/citability-audit` - Raw SKILL.md: https://api.skillmd.com/api/skills/zaialamm/citability-audit/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Marketing & Growth - Author: zaialamm (https://skillmd.com/u/zaialamm) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/zaialamm/citability-audit --- # Citability audit ## Goal Find out why a page is or isn't citable by search and AI answer engines, fix it, and prove the fix with a second audit. The audit returns an ordered work list; your job is to work through it, not to summarise it. ## Required inputs - An absolute `http(s)` URL of one page. Ask if the user gave only a domain — audit the homepage first unless they name a page. - Whether the site's source is in the working directory. If it is, you apply fixes; if not, you hand the user a fix list they can give to whoever owns the site. ## Before spending - Call `citable_prices` (free) once per session. It returns the wallet, network, per-call prices and the spend cap (`CITABLE_MAX_PRICE`). - One audit is $0.01. Ask before auditing more than 5 pages in one go. - Failed calls are never charged. A paid result carries `_payment.transaction`, the settlement signature. ## Tools - `citable_onpage_audit(url)` — $0.01. Returns `score` (0–100), `checks[]` (each with `id`, `status`, `weight`, `current`, `fix`) and `topFixes[]`, an ordered list of what to change first. ## Workflow 1. Run `citable_onpage_audit` on the URL. Read `score` and `topFixes[]`. 2. Report the score in one line, then the top fixes as a numbered list in the order the audit gives them — that order is by weight, so it is the order to work in. 3. If the source is available, apply fixes in that order. Typical mappings: | Check | Where it lives | Fix | |---|---|---| | `ai-crawler-access` | `robots.txt` | Allow `GPTBot`, `ClaudeBot`, `PerplexityBot`, `Google-Extended`, `CCBot` for the audited path. Say so explicitly if the user blocked one on purpose. | | `llms-txt` | `/llms.txt` at the site root | Plain-text summary: what the site is, its main pages, contact. | | `answer-headings` | The page's H2/H3s | Rewrite headings as the question a buyer would ask; put the direct answer in the first sentence under each. | | `article-signals` | Page template | Visible publish date and author. | | `js-content` | Rendering setup | The audit saw an empty shell. Server-render or prerender the page; AI crawlers do not run JavaScript. | | `structured-data` / schema checks | `` JSON-LD | Match the page type: Article, Product, FAQPage, Organization. | | Title / meta length, H1, canonical, alt text | Page template | Follow the `fix` string on each check. | 4. After deploy (or against a preview URL), run `citable_onpage_audit` again. 5. Report before → after score, what you changed, and anything that needs a human: DNS, CMS settings, a crawler the user must decide about. ## Guardrails - Never quote a score you did not get from the tool. - Do not loosen `robots.txt` for a crawler the user deliberately blocked without saying so first. - If `js-content` fails, do not try to "fix" it with more client-side JavaScript. - One URL per audit. The audit is per page, not per site.