Free AI Visibility Checker
Find out whether ChatGPT, Claude, Perplexity and Google’s AI can reach your site, read it, and recommend you. We test 24 AI crawler rules, fetch your page as 19 of those crawlers, and show you what they actually see.
Open source · AGPL-3.0 licence
The checker above tests any public URL from outside, on a sample of pages. The free, open-source app runs on your machine for a site you control: it crawls every page and checks technical SEO, AI crawler access, AEO/GEO and agentic-web readiness. Add your own API keys and it tracks what ChatGPT, Claude, Gemini and Perplexity say about you.
npx github:Devs-Core-Company/ai-visibility-checker audit https://yoursite.comnpm install -g github:Devs-Core-Company/ai-visibility-checker
avc init
avc ui
See it run
One command audits every page. The dashboard explains each issue, shows the exact robots.txt line that blocks each AI crawler, checks what AI agents can find, and writes the fixes for your stack. Recorded on the benchmark’s demo site, which has planted issues; the AI answers shown use the app’s sample mode.
What it is
The app is a single command, avc. Use it from a terminal, open its dashboard, let an AI coding assistant drive it, or run it in CI. Everything it finds is stored on your machine.
Run audits, AI visibility checks, fixes and reports from a terminal or a CI job.
A local web dashboard at http://127.0.0.1:4317. Only your own machine can open it.
Lets Claude Code, Cursor or Claude Desktop run audits and apply fixes for you.
Runs the audit in CI on a schedule and reports issues to GitHub code scanning.
Crawls every page, up to your limit, and runs 400+ technical and content checks on each.
Whether 40 AI and search crawlers are allowed in, and whether the ones that fetch pages actually get in.
How easy each page is for an answer engine to quote, with a GEO score per page.
Whether AI agents, not only crawlers, can find and use your site.
Real answers from the engines, on your keys and your prompts.
Each failing check comes with the change to make, for the stack you run.
Numbers only you can see, read with your permission.
The same audit on a schedule, so you see what changed.
The web checker at the top of this page is still the quickest way to test any public site, including a competitor’s. The app is for a site you control.
Web checkerAnyone checking any public website, their own or someone else’s.
AppSite owners, developers and agencies working on a site they control.
Web checkerThe page you enter, plus a sample of 10 to 20 pages from your sitemap.
AppEvery page it can reach, up to the limit you set.
Web checkerAI access, readability, rendering gap, schema and llms.txt.
App640+ technical SEO, AEO/GEO and agentic-web checks, run across every page.
Web checkerrobots.txt rules for 24 crawlers and a live fetch as 19 of them.
App40 AI and search crawlers, with the deciding robots.txt line, plus a live fetch as each of the 25 that fetch pages and challenge detection on key pages.
Web checkerChatGPT, Claude, Gemini and Perplexity, asked eight questions each that we choose, once.
AppThe same engines, with your own prompts, run as often as you like and tracked over time.
Web checkerNone. Everything is measured from outside your site.
AppSearch Console, GA4, PageSpeed Insights, Bing Webmaster and your own server logs.
Web checkerWhat to change for each failing check.
AppSteps for your stack naming the file, copy-paste code, generators, and auto-apply to WordPress or a local repo.
Web checkerOne report at a time, with a link to share.
AppScheduled audits, history, diffs, alerts and a GitHub Action.
Web checkerOn devs-core.com.
AppOn your machine, in CI or on your own server. Data stays in a local SQLite file.
Web checkerFree. The AI scan needs a verified work email.
AppFree and open source (AGPL-3.0). AI scans use your own API keys, with spend caps.
Open benchmark · 13 October 2026
Five tools, one test site with 121 planted defects, the same machine and rules written before any tool ran. Pick an area to see how much each tool caught.
| Tool | Found · All areas | Precision | False alarms |
|---|---|---|---|
| AI Visibility Checker1.0 | 121/121 | 88% | 0 |
| Scrawly1.0 | 58/121 | 20% | 45 |
| Google Lighthouse13.5 | 31/121 | 51% | 1 |
| Unlighthouse0.19 | 27/121 | 46% | 2 |
| python-seo-analyzer2025.4.3 | 7/121 | 13% | 5 |
Found: share of the planted defects a tool reported. Precision: share of its findings that point at a real defect. False alarms: issues reported on 12 clean control pages. We built both the app and the test site; commercial tools were not tested. Method, raw outputs and how to rerun it
Before you start
Check with node --version. If it is older, install the current LTS from nodejs.org.
npm uses it to fetch the app from GitHub. Most developer machines already have it.
Optional. With it, the app renders each page the way a browser does and measures what exists only after JavaScript runs.
Optional. An alternative to installing Node.js, if you would rather keep the app in a container.
Install
Pick your system. Each route ends with the same avc command. To try it once without installing anything, run npx github:Devs-Core-Company/ai-visibility-checker audit https://yoursite.com.
Open Terminal and run the installer. It installs the avc command and connects it to the Claude app, Claude Code and Cursor if you have them.
curl -fsSL https://raw.githubusercontent.com/Devs-Core-Company/ai-visibility-checker/main/install.sh | bashOr, if you already have Node.js 22.5 or newer, install it with npm.
npm install -g github:Devs-Core-Company/ai-visibility-checkerSet up your site.
avc initOpen PowerShell and run the installer. It installs the avc command and connects it to the Claude app, Claude Code and Cursor if you have them.
irm https://raw.githubusercontent.com/Devs-Core-Company/ai-visibility-checker/main/install.ps1 | iexOr install Node.js 22.5 or newer from nodejs.org, then install the app with npm.
npm install -g github:Devs-Core-Company/ai-visibility-checkerOpen a new PowerShell window so the avc command is on your path, then set up your site.
avc initRun the installer in a terminal. It installs the avc command and connects it to the Claude app, Claude Code and Cursor if you have them.
curl -fsSL https://raw.githubusercontent.com/Devs-Core-Company/ai-visibility-checker/main/install.sh | bashOr install with npm, using Node.js 22.5 or newer from your package manager, nvm or nodejs.org.
npm install -g github:Devs-Core-Company/ai-visibility-checkerSet up your site.
avc initBuild the image straight from the GitHub repository.
docker build -t avc https://github.com/Devs-Core-Company/ai-visibility-checker.gitRun an audit. The avc volume keeps your history between runs.
docker run --rm -v avc:/data avc audit https://yoursite.comFirst run
avc audit crawls your site, runs every check on every page and saves the result. Without any keys or connections you get the full technical, AI crawler, GEO and agentic-web audit, with fixes.
avc ui opens the same results in a dashboard at http://127.0.0.1:4317, where you can browse issues by page, see fixes, connect data sources and schedule runs.
Run node --version in a terminal. If it is older, install the current LTS from nodejs.org.
Run npm install -g github:Devs-Core-Company/ai-visibility-checker, or the one-line installer for macOS, Linux or Windows.
Run avc init and follow the prompts.
Run avc audit https://yoursite.com. The app crawls the site, runs every check and saves the result locally.
Run avc ui and open http://127.0.0.1:4317 in your browser.
Run avc setup-claude (the one-line installer already does this). It connects the app to the Claude desktop app, Claude Code and Cursor. Fully quit the Claude app and open it again, then ask: audit https://yoursite.com with avc.
Claude app, Claude Code and Cursor
The app includes an MCP server. avc setup-claude connects it to the Claude desktop app, Claude Code in every folder, and Cursor, in one step. The one-line installer runs it for you.
Then fully quit the Claude app (closing the window is not enough) and open it again. In a new chat, ask in plain words, for example: “Audit https://yoursite.com with avc and tell me what to fix first.”
npm install -g github:Devs-Core-Company/ai-visibility-checker
avc setup-claudeclaude mcp add --scope user avc -- npx -y github:Devs-Core-Company/ai-visibility-checker mcpclaude mcp add alone only reaches Claude Code, not the Claude app’s chat. Manual setup for each app: MCP guide.
Connect your site
Connect what you have, from the dashboard or with avc connect. Keys can also be set as environment variables. Nothing is connected until you set it up.
Unlocks Clicks, impressions and queries per page, index coverage, and URL Inspection for the pages the audit flags.
Unlocks Visits referred by AI assistants such as ChatGPT, Perplexity, Gemini and Copilot, per landing page.
Unlocks Lab scores and real-user Core Web Vitals for your key pages.
Unlocks Bing crawl and index data, and sending changed URLs to search engines through IndexNow.
Unlocks Real AI crawler visits: which bots came, how often and to which pages, verified against each operator’s published IP ranges so faked user agents are not counted.
Unlocks Applying fixes to your WordPress site directly, with one-click rollback.
Unlocks Fixes written as a patch against your own code, for example a Next.js or Astro site. You review it; nothing is committed for you.
Unlocks AI visibility tracking: your prompts, sent to the engines you choose, with mentions, share of voice, citations, sentiment and fact-checks.
Unlocks The full AI visibility report with made-up answers, so you can see how it works before adding keys.
Monitoring
Run audits and AI visibility checks on a schedule with avc monitor or from the dashboard. Every run is kept, so you can see history and a diff of what changed.
Send a message to Slack or Discord with AVC_WEBHOOK_URL, or by email with AVC_ALERT_EMAIL and your SMTP details, when something gets worse.
Run the audit in CI and upload the results as SARIF, so issues show in GitHub code scanning. The workflow file is in the repository README.
Privacy
Licence: AGPL-3.0. Source: github.com/Devs-Core-Company/ai-visibility-checker
Questions
Straight answers, including the ones where the honest answer is “nobody can measure that”.
Yes. It is open source under the AGPL-3.0 licence and costs nothing to download or run. Using it on your own or your clients' sites carries no obligations; if you change it and offer it to others, including as a hosted service, you must share your source. The only cost is the AI visibility tracking, which sends your prompts to OpenAI, Anthropic, Google or Perplexity on your own API keys. Your provider bills you directly, and you can set a spend cap per run and per month. Mock mode shows the full report with made-up answers at no cost.
The web checker at the top of this page tests any public URL from outside, on a sample of pages, and asks four AI engines a fixed set of questions once. The app runs on your machine for a site you control. It crawls every page up to your limit, runs 640+ technical SEO, AEO/GEO and agentic-web checks, tracks your own prompts over time, reads Search Console, GA4 and your server logs, and can apply fixes to WordPress or a local repository.
No. Site audits, AI crawler checks, GEO scores and fixes work without any keys. Keys are only needed for AI visibility tracking, which asks ChatGPT, Claude, Gemini and Perplexity your prompts. You can use keys from each provider, or one Vercel AI Gateway key for all of them. Set AVC_LLM_MOCK=1 to try the tracking reports without keys.
Only to the services you configure. The app crawls your website, and calls the AI providers, Google, Bing or WordPress only if you connect them. Results stay in a local SQLite file. There is no Devs Core account and the app does not send anything to Devs Core.
macOS, Windows and Linux, with Node.js 22.5 or newer. You can also build and run it with Docker. Chrome or Chromium is optional and is used to render pages that build their content with JavaScript.
Yes. The app includes an MCP server. Run avc setup-claude (the one-line installer does it for you) and it is added to the Claude desktop app, Claude Code in every folder, and Cursor. Quit the Claude app fully and reopen it, then ask in plain words: audit your site, explain a failing check, or apply a fix to the repository you have open.
Fixes name the file to change for Next.js, WordPress, Shopify, Astro, Webflow and other common stacks, with copy-paste code and a prompt you can hand to Claude Code or Cursor. On WordPress the app can apply a fix for you and roll it back with one click. On a local code repository it writes a patch for you to review.
Yes. If you would rather not do the work yourself, Devs Core builds server-rendered sites AI can read, fixes crawler access and structured data, and connects your logs and analytics. Book a discovery call and bring your report.
More free tools
Is GPTBot, ClaudeBot or PerplexityBot blocked? robots.txt rules plus a live fetch as each crawler.
Find and validate your llms.txt and llms-full.txt against the llmstxt.org format.
How the assistants choose which businesses to name, why most sites never appear, and what to fix first.
Bring your report. Devs Core builds server-rendered sites AI can read, fixes crawler access and structured data, and connects your logs and analytics.
Not your site, or just a quick look? Check any URL with the free web checker.