THE CHECK IS THE START.
Get from “not working”
to your next step.
Start with the simplest comparison. Avoid changing several settings at once: you won't know which change made the difference.
Website owner? Help AI helpers read and use your site →
Works on one device, not another?
- Check the spelling and open the website's homepage directly.
- Try a private browser window. If it works there, an extension or saved browser data may be involved.
- Compare Wi-Fi with mobile data. A difference points toward the connection or provider, but does not identify the exact cause.
- Check the service's official status page. A working homepage does not prove that every service is healthy.
- If the browser shows a certificate warning, stop. Do not disable security checks to reach the page.
Do not buy a VPN just because our checker can reach a website. Many local problems have free fixes, and our result does not prove a regional block.
What does the response code mean?
| Code | What to do |
|---|---|
| 200–299 | The request succeeded. Compare the actual failing page and function in your browser. |
| 300–399 | The server redirects elsewhere. We follow up to five safe HTTPS redirects; the main result describes the tested final destination or explains why we stopped. |
| 401 / 403 | Authentication or access restrictions. Do not infer a global outage. |
| 404 | The requested page was not found. Check the address. |
| 405 | The server does not allow the usual HEAD method. We can retry with a header-only GET for 404, 405 or 501 responses. If the check is still restricted, compare in your browser. |
| 429 | Too many requests. Wait before trying again. |
| 500–599 | The server or an upstream service reported a problem. Owners should inspect logs and recent changes. |
Your own website is failing?
- Record the time, exact browser error and affected page. Avoid sharing private URLs or access tokens.
- Check your hosting provider's status and your own deployment history.
- Check domain registration, DNS records and HTTPS certificate separately.
- If a recent release caused the failure, use your documented rollback procedure.
- After recovery, add independent monitoring and a tested alert route.
This tool does not diagnose every cause and is not a replacement for your provider's logs. Read exactly how we check.
Help AI helpers read and use your website
Start with useful public content and clear controls. A website answering, an agent reading its content, and an agent carrying out an action are three different things.
Not tested by our quick check. Our checker reads response headers, not page contents or another website's WebMCP tools. This is a guide for owners, not a score for the domain you checked.
- Make important answers easy to find. Publish clear text, descriptive headings and links between related pages. Keep structured data (machine-readable descriptions, such as an article author) consistent with what people can see. Google says its usual search requirements apply to AI Overviews and AI Mode; no special AI file is required. Being eligible does not guarantee inclusion. Google's guidance.
- Make controls understandable. Give form fields visible labels and buttons specific names, such as “Check website”. Keep important controls stable while a page loads. Test the actual task with a keyboard as well as a mouse. This helps people and gives automated helpers clearer cues. Chrome's agentic browsing checks.
- Choose search access separately from training. OpenAI uses OAI-SearchBot for search and GPTBot for model training. You can allow the search bot while blocking the training bot. ChatGPT-User handles some user-requested visits and is a separate case. Review the provider's current rules and published addresses rather than trusting a bot name alone. OpenAI's bot descriptions.
- Check your hosting protection too. Your robots.txt file tells crawlers which pages they may visit, but does not reveal every block. Your hosting provider or CDN (the service in front of your website) may challenge or block a request separately. Inspect its rules and verified-bot settings; keep protections against abusive traffic. Cloudflare's AI bot controls.
- Add an optional content guide. An llms.txt file can summarize your site's purpose and link to useful public documents. Keep it accurate. Missing this optional file does not prove that agents cannot read your site, and adding it is no search-ranking guarantee. Chrome's llms.txt explanation.
- Describe actions when your site needs them. WebMCP is an experimental way to describe browser actions, such as searching or submitting a form. A readable article does not need it. Preserve ordinary controls, explain inputs, and ask for confirmation before sensitive actions. WebMCP is separate from search inclusion. Chrome's WebMCP guide.
How do I check whether a change helped?
Record one real task, the page and the result before changing it. For example: can a person or supported agent find a contact address, or complete a public search? Repeat the same task after one change. A better task result does not prove better search visibility.
Lighthouse's experimental agentic browsing checks require Chrome 150 or later; WebMCP checks also require origin-trial registration. Its report lists passed checks and individual findings, rather than a weighted 0–100 score. Name the browser, version, method and anything not tested when sharing a report. Lighthouse's method and limits.
About this service
CheckWebsiteNow checks public website responses. No ads, external analytics or session recordings are loaded. Optional usage measurement records coarse steps and a random visit identifier on our server for up to seven days. Read our operator and privacy information or contact us.
Help agents read your public page
The separate content check reads public homepage HTML and two optional files. It never runs scripts or signs in. Unverified findings stay unverified; a training opt-out is a policy choice, not a quality failure.
What the six observations mean
- Public page: Did our request receive a complete supported document? A refusal may be a protection rule, not an outage.
- Structure: Descriptive titles, headings and text can help readers understand a page. Static HTML cannot prove what a full browser renders.
- Bot rules: We inspect published robots.txt rules for the tested final path. A published permission does not prove a provider can access the site. Keep search and training choices separate.
- llms.txt: An optional guide to your public content. Missing it does not make a page unreadable.
- Metadata: Optional structured hints can explain your content; they are not a ranking promise.
- WebMCP: Static form or script hints cannot prove that tools register or work in a real browser. Ordinary articles do not need agent actions.
Improve your real public content first: useful text, clear headings, labeled controls and working links. Add optional files only when they help your audience. Do not remove security controls or expose private data to chase a score.
Sources: OpenAI bot roles, Google robots rules, Chrome WebMCP. Our bounded inspection is not a complete standards crawler, an official Lighthouse score or a remote MCP server.