The Claude prompt library

You are auditing my website for SEO/AEO/GEO readiness. Work in this order and report PASS/PARTIAL/FAIL with the specific fix for each: (1) are the AI retrieval crawlers allowed in robots.txt; (2) any noindex, orphan pages, or broken sitemap reference; (3) inventory every JSON-LD @type and flag missing/inaccurate ones; (4) does each cornerstone page front-load a self-contained answer with a number in the first 200 words. End with the 5 highest-leverage fixes ranked by impact.

Rewrite the first 150–200 words of the page I'm pasting as a standalone answer an AI engine could quote: sentence one answers the core question with a number or named entity, no preamble, self-contained, keep my facts, mark unsupported claims [needs cite]. Then one line on what changed.

Review this page for fact density and information gain: mark generic consensus sentences to cut or upgrade; per section, suggest one concrete addition (original data, a stat to source, a test to run, an expert to quote). Don't fabricate numbers — tell me what to add and where.

Restructure this page keeping all facts: convert headings to natural question form; make sections self-contained (~75–225 words) with the answer first; draft one comparison table to replace prose; add a 4–6 item FAQ from real questions, formatted for FAQPage schema.

Generate a robots.txt for DOMAIN that explicitly allows Googlebot, Bingbot, Google-Extended, GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot, Applebot-Extended, CCBot, Bytespider, Google-Agent (Crawl-delay: 2 on heavy bots), disallows only the non-content paths I list, and declares the Sitemap. Then tell me what my current robots.txt (below) is missing or blocking.

From this sitemap/page list, write a spec-compliant llms.txt: H1 site name; one-sentence blockquote; ## sections linking the most important pages as - [Title](URL): concrete one-line description. Skip utility/legal pages. No marketing adjectives.

Generate valid JSON-LD for this page: correct primary type (Article/HowTo) with headline, description, author (Organization), publisher, datePublished, dateModified, inLanguage, canonical; FAQPage only if the FAQ is visible, using the on-page Q&As; BreadcrumbList for its place in my site. One minified <script type="application/ld+json"> block; flag guessed fields.

Draft a reusable entity fact sheet from the details below: canonical name + one-sentence description (used identically everywhere); 3–5 category tags; 3 differentiators as checkable facts; 40-word and 100-word about; the same facts formatted for Organization schema, an llms.txt summary, and a directory listing. Keep every version factually identical.

For my topic below: list the subreddits/forums/Q&A sites where my buyers actually ask questions; for 5 real question types, draft genuinely useful answers that mention my site only where it's the best fit (no marketing voice, no over-linking); draft an honest note to be considered for relevant 'Best X' listicles. Flag anything that reads as shilling.

Generate 25 prompts real buyers would type into ChatGPT/Perplexity/Gemini/AI Mode for my topic (category, comparison, problem queries). I'll paste answers back; for each, score whether my brand is mentioned/linked/cited, which competitors appear, and which sources were cited. Summarize my visibility %, share of voice, and the 5 most-cited sources I'm missing from.

Act as a browsing agent trying to complete this task on the page I'm pasting: [task]. Navigate by the accessibility tree, not visuals. List every failure point — unlabeled inputs, div-buttons, facts only in images, ambiguous links, layout-shift risks — with the exact HTML fix for each, then give a pass/fail verdict for the task.

From this product page and feed row: generate spec-complete Product+Offer JSON-LD (GTIN/MPN, price, priceCurrency, availability, condition, shipping, returns); flag every field missing or inconsistent across page, schema, and feed — those gaps block AI shopping surfaces; then output a corrected feed row.