Python CLI reference

ai-visibility audit <url> fetches a live page and prints a full AI-visibility report; crawlers lists the registry; generate llms-txt/generate robots-txt write the respective files to stdout. This is a separate binary from the npm package's CLI, but the command names and shape are intentionally identical. Requires the cli extra.

Install

pip install ai-visibility[cli]

Adds click, httpx, and rich as dependencies and registers the ai-visibility console script (ai_visibility.cli:main). It's a Click command group:

Usage: ai-visibility [OPTIONS] COMMAND [ARGS]...

  ai-visibility: make your site optimally visible to AI crawlers.

Options:
  --version  Show the version and exit.
  --help     Show this message and exit.

Commands:
  audit     Fetch URL, score it for AI-visibility, and print a report.
  crawlers  List all known AI crawlers in the registry.
  generate  Generate llms.txt / robots.txt content to stdout.

audit — the headline command

Fetches a live URL over HTTP, best-effort checks for /llms.txt and /ai.txt at the same origin, and scores the response with analyze_content() — the same function documented on the API reference:

Usage: ai-visibility audit [OPTIONS] URL

  Fetch URL, score it for AI-visibility, and print a report.

Options:
  --timeout FLOAT  HTTP request timeout in seconds.  [default: 10.0]
  --help           Show this message and exit.

Real output, run against a bare-bones fixture page (a single H1 and one short paragraph, no schema, no llms.txt):

ai-visibility audit http://localhost:8931/

AI-Visibility Report for http://localhost:8931/

AI-visibility is poor (24/100). 1 critical issue likely block AI crawlers from
understanding this page well. 3 warnings should be addressed. Weakest area: eeat
signals (0/100).

Overall score: 24/100
  Answer Front Loading 20/100
  Eeat Signals         0/100
  Heading Structure    100/100
  Schema Coverage      0/100
  Fact Density         10/100
  Snippability         0/100
  Crawler Accessibility 40/100

Findings:
  [CRITICAL] No JSON-LD structured data found on the page.
      Fix: Add JSON-LD schema markup, e.g. via ai_visibility.schema.article_schema().
  [WARNING] No llms.txt found.
      Fix: Generate one with ai_visibility.generators.generate_llms_txt().
  [WARNING] Very little content appears before the reader has to dig further into the page.
      Fix: Lead with at least 40 words directly answering the page's main question.
  [WARNING] Missing authority/trust signals: an author meta tag, author/publisher in JSON-LD, a contact link (mailto:/tel:).
      Fix: Add an author meta tag, author/publisher JSON-LD, and a contact link to establish E-E-A-T signals.
  [INFO] No ai.txt found.
      Fix: Generate one with ai_visibility.generators.generate_ai_txt().
  [INFO] Low factual density (0.0 concrete numbers/stats per 100 words).
      Fix: Include more concrete numbers, dates, and statistics to make content easier to cite.
  [INFO] Some sections have too little content under their heading to stand alone as an excerpt.
      Fix: Ensure each section has at least 20 words of self-contained content.

exit: 1

audit exits 1 if any finding is CRITICAL, and 0 otherwise — enough to drop straight into a CI step without a separate --fail-under flag:

.github/workflows/ci.yml

- name: GEO audit
  run: |
    pip install ai-visibility[cli]
    ai-visibility audit https://staging.example.com

Unlike the npm CLI's audit/lint, the Python CLI has one audit command, no --dir local-directory mode, no --fail-under <n> numeric threshold, and no --json output in 0.5.0 — the exit code is a fixed any-critical-finding gate. For programmatic access to the full ScoreResult/AnalysisResult objects (JSON-serializable via dataclasses.asdict()), call analyze_content() directly instead of shelling out to the CLI.

crawlers — list the registry

Usage: ai-visibility crawlers [OPTIONS]

  List all known AI crawlers in the registry.

Options:
  --help  Show this message and exit.

ai-visibility crawlers

GPTBot               OpenAI          training   (verified)
ChatGPT-User         OpenAI          search     (verified)
OAI-SearchBot        OpenAI          search     (verified)
ClaudeBot            Anthropic       training   (verified)
Claude-User          Anthropic       search     (verified)
Claude-SearchBot     Anthropic       search     (verified)
PerplexityBot        Perplexity AI   search     (verified)
...
Bytespider           ByteDance       training   (unverified)
YouBot               You.com         search     (unverified)
cohere-ai            Cohere          training   (unverified)
Diffbot              Diffbot         indexing   (unverified)

Reads straight from get_all_crawlers() — see the crawler registry referencefor the full, annotated list with source URLs and what "verified" means.

generate llms-txt

Usage: ai-visibility generate llms-txt [OPTIONS]

  Generate llms.txt content.

Options:
  --title TEXT        Site or project title (used as the H1).  [required]
  --description TEXT  One-line summary (used as the blockquote).
  --url TEXT          Homepage URL, added as a link entry.
  --help              Show this message and exit.

ai-visibility generate llms-txt --title Acme --description 'Acme makes widgets.' --url https://acme.com > llms.txt

# Acme

> Acme makes widgets.

## Links

- [Acme](https://acme.com)

For anything beyond a single title/summary/link (multiple sections, per-link descriptions, an llms-full.txt with inlined page content), build a LlmsTxtConfig/LlmsFullTxtConfig and call generate_llms_txt()/generate_llms_full_txt() programmatically — see the generators API reference.

generate robots-txt

Usage: ai-visibility generate robots-txt [OPTIONS]

  Generate an AI-aware robots.txt.

Options:
  --help  Show this message and exit.

ai-visibility generate robots-txt > robots.txt

User-agent: *
Allow: /

In 0.5.0 this subcommand takes no flags — it always generates the default, fully-permissive RobotsTxtConfig(). For crawler-category blocking, per-crawler rules, or sitemap URLs, call generate_robots_txt(RobotsTxtConfig(...)) directly instead of the CLI — see the generators API reference for the full RobotsTxtConfig shape.

Same interface as the npm CLI, on purpose

audit, crawlers, and generatemirror the npm package's command names deliberately — a team running both a Next.js frontend and a Django/Flask/FastAPI backend can reach for the same mental model (and largely the same muscle memory) in either language, even though the two binaries are separate implementations with some flags that don't (yet) match one-for-one — see the note under audit above for the current gaps. See the npm CLI referencefor the Node.js equivalent's fuller flag set (--dir, --fail-under, --json, lint).

See the scoring guide for what the seven breakdown dimensions in an audit report mean, and the API reference for the programmatic equivalent of every command above.