All skills / Prompts

AI search readiness check prompt for any chatbot

Copy the prompt, paste it into any chatbot, then add your own material where it says. Free, no sign-up.

Also in: Deutsch, EspaƱol (translated, this English version is the tested one)

What you need

A website URL. In this prompt version you paste the text yourself. The steps are under "How to use it" below.

The prompt

Use this in any chatbot (ChatGPT, Claude, Gemini and others). Nothing to install. Tested in Claude only; not yet tested in ChatGPT or Gemini.

How to use it

  1. Open https://<site>/robots.txt, the home page's source (right-click, View page source) and the home page itself. Check whether /llms.txt and /sitemap.xml open.
  2. Copy everything from "COPY FROM HERE" to the end of this file and paste it into a new chat.
  3. Under it, paste the robots.txt text, the <head> part of the source (plus any application/ld+json blocks), whether llms.txt and the sitemap open, and the visible text of the home page.

A score out of 100 from 8 fixed checks, the evidence for each, the fixes in order and a client summary. The installed skill fetches and scores with a script; here the chatbot reads what you paste and estimates word counts, so check any score that decides a fix.

Prefer to do it by hand? The 8-check table in the instructions below works as a manual checklist.

COPY FROM HERE

You check whether AI search tools (ChatGPT search, Perplexity, Claude, Google and Bing) can read a website, for a marketer who will fix it. Work only from what I paste. Never guess a fact you weren't given. (This prompt was tested in Claude only.)

**What I will paste (ask me once for anything missing, then work with what you have):**
1. The full text of `https://<site>/robots.txt` (or "none" if it doesn't open).
2. From the home page source (right-click, View page source): the part from `<head>` to `</head>`, plus any further blocks that say `application/ld+json`. Long `<script>` and `<style>` blocks that don't say `application/ld+json` can be left out.
3. Whether `https://<site>/llms.txt` and `https://<site>/sitemap.xml` open (yes or no).
4. The visible text of the home page (select all on the page, copy).

**Score the site out of 100 with exactly these 8 checks:**

| # | Check | Points |
|---|---|---|
| 1 | robots.txt lets AI search and user agents read the home page: OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Claude-SearchBot, Claude-User, Googlebot, Bingbot. All allowed: 25. Some blocked: 10. All blocked: 0. A group for `User-agent: *` applies to every agent without its own group. A group applies only to the exact agent name it gives (`User-agent: Claude` is not Claude-User). If robots.txt gave a server error (5xx), score 0: crawlers treat the whole site as blocked until it loads. | 25 |
| 2 | Home page not marked noindex (meta robots in the head). | 10 |
| 3 | Readable text on the home page (the visible text I pasted): 150 words or more 15, 50 to 149 words 7, under 50 words 0. You are estimating this count: say "about N words". Text that only appears after JavaScript runs can't be told apart here; say so. | 15 |
| 4 | Title present (5) and meta description present (5). | 10 |
| 5 | JSON-LD structured data with an Organization or LocalBusiness type (or a specific business type such as Dentist or Plumber). | 15 |
| 6 | Question-style headings (a short line in the visible text ending in "?") or FAQPage structured data. | 10 |
| 7 | A sitemap is listed in robots.txt or /sitemap.xml opens. | 10 |
| 8 | /llms.txt opens. Optional: it is a proposal, and no AI search engine has confirmed it uses it. | 5 |

**Reply in exactly this format:**

## AI search readiness: N/100
One line: how many checks passed in full, and the first fix.

| # | Check | Points | Evidence |
For every check, give the points as "points/max" and quote the evidence from what I pasted (the robots.txt lines, the meta tag, the JSON-LD types, the headings).

## Fix first
Every check that lost points, the biggest loss first. Say what is wrong, quote the evidence, and give the concrete change. For robots.txt, quote the blocking group and say which line to change.

## Training crawlers (not scored)
Which of GPTBot, ClaudeBot, Google-Extended, CCBot, Applebot-Extended and Meta-ExternalAgent robots.txt blocks. Blocking them is a business choice and does not remove a site from AI search answers.

## Check these yourself
- Word counts and robots.txt matching here were done by reading, not by a program. Check any score that decides a fix.
- Not checked: firewalls and bot protection, other pages, content added by JavaScript, and whether any AI tool actually cites the site.

## Client summary
Exactly three plain sentences (no questions) for the client: the score, the biggest gap, the first fix. No counts you made yourself.

**Rules:**
- Any count you make yourself (words, questions, links) is written as "about N (counted by reading)" and is never repeated in the client summary.
- robots.txt is a request, not a block: ChatGPT-User and Perplexity-User fetch pages when a person asks, and their operators say robots.txt may not always apply to them. Say "robots.txt does not ask AI search agents to stay out", never "AI tools can read the site".
- Quote, don't paraphrase, when you give evidence. If I didn't paste something, write "not provided" and give that check 0 points with that reason.
- No promises that fixes will get the site cited, ranked or recommended, and no traffic numbers.
- Don't name the website platform unless the source shows it.
- This score is a checklist with fixed weights, not a score from any AI company.

A real example

Real public data, anonymised. Taken from the skill's README.

Source: a small local plumbing business site.

Before: "Can ChatGPT and other AI search tools actually read my client's site?"

After (skill output, unedited, excerpt): "## AI search readiness: 85/100 [...] 1. Add Organization or LocalBusiness structured data (15 points lost). The home page's JSON-LD only declares WebSite. LocalBusiness (or the more specific Plumber type) states who the business is in a form machines can read. It must use [the business's] real name, address and phone number."

Limits

  • It checks the home page and robots.txt only, and can't test firewalls or bot protection.
  • The score is our checklist with our own weights, not a score from any AI company, and promises no citations or traffic.

The same job as an installable skill

AI Search Readiness Check is also a skill you install once and call in plain words, so you do not paste the prompt each time. It can use its helper scripts and web access where your assistant allows. See the skill page, or read how to install skills.

More for marketers and agencies