Operator & privacy

Impressum.

CheckWebsiteNow is operated by:

Marcus Wolf
Stiegstrasse 11
8602 Wangen
Switzerland
info@marcuswolf.org

CheckWebsiteNow is the name of this service. This website provides a limited technical observation and practical guidance. It does not guarantee availability, diagnose every fault or offer emergency support.

Privacy: the essentials

Contact Marcus Wolf at the address above for questions or to request access, correction or deletion of personal information. Tell us which information concerns you; do not send passwords or secret links. Requests are handled subject to the applicable law.

Checking a website

Your browser sends the public domain to our checker. Pasted links are reduced to the domain; credentials are rejected. With every Check Website, our server runs two checks at the same time. The quick response check resolves public addresses, reads only the HTTPS response headers of the homepage and follows at most five safe HTTPS redirects; it does not inspect page contents. The agent readability check reads a few bounded public files from the final homepage address: the homepage HTML, robots.txt, llms.txt, /.well-known/api-catalog, /.well-known/ai-catalog.json, /.well-known/mcp/server-card.json, a sitemap and one Markdown version of the homepage. All of them share a ten-second retrieval budget with at most nine HTTPS requests and no more than five redirects per file. Every file after the homepage is fetched only from the same host; a sitemap on another host is never fetched. The homepage is read up to 1 MiB (a larger homepage is analysed from its first 1 MiB), every other file up to 64 KiB. No page scripts, form submissions, linked pages, MCP or API endpoints or actual agent actions are executed. Retrieved documents are processed in memory and discarded; they are not saved to a content database. Reports use bounded observations, not a copy of the page. The checked website receives our checker requests (User-Agent UtilityStatusProbe/1.0 for the quick check and User-Agent CheckWebsiteNowContentProbe/1.0 for the agent readability check), not your browser address. The destination and name-resolution provider receive the connection data needed to answer the check. Avoid entering private or sensitive addresses.

If you paste an error message instead of a domain, the text is sent to our server only to find matching entries in our error library. Do not paste private links, passwords or personal data.

Results of both checks are held in server memory for up to sixty seconds to avoid repeated requests; when one of our limits refuses a fresh check, a result up to five minutes old may be shown instead, marked as reused, and the help for websites blocked in your country uses the same result for up to five minutes; older results are deleted from memory within a minute. For the curated status pages, the last result per listed service (for example whether github.com answered, with the time) is kept in our server’s database so it survives a restart, replaced by the next check of that service and shown for up to 24 hours; it describes the public service, not you or your visit. We do not store a permanent database of the domains you check. The five recent checks, comparisons and sound preference stay in this page’s memory and disappear when you reload. If you choose a light or dark theme with the switch in the header, only that choice (“light” or “dark”) is saved in your browser’s local storage on this device; it contains no identifier and is never sent to us. No persistent analytics cookie or browser identifier is created.

Delivering and protecting this service

The service runs on the operator’s existing server and uses Cloudflare to deliver HTTPS securely. Cloudflare processes necessary network information, including your IP address, according to its privacy policy. Its network can process data in countries outside Switzerland. The checker uses your IP address (for IPv6 your /64 network) for request limits, only in its working memory: these entries are deleted at the latest about two minutes after your last request (24 hours for “Was this helpful?” answers, see below; salted fingerprints for the source totals until the end of the UTC day, see “Where visitors come from”) and are not included in our usage events. The application and gateway do not keep routine access logs. Infrastructure providers may retain their own security data.

Optional usage measurement

This optional usage measurement is off until you enable it. The anonymous daily source totals (see “Where visitors come from”) are separate and contain no identifier. The additional agent-content check is not included in the quick-check usage funnel. If you agree, we record coarse steps such as opening the page, starting a check, seeing a result or clicking a partner link, plus a random identifier for this visit, a time, a test/agent/visitor label and an experiment variant. These records never include the entered domain, your IP address, page contents or recordings. Labels are not verified counts of real people. You may turn measurement off at any time; this stops further events.

Public-service event records older than seven days are deleted during the next hourly cleanup, server start, event or report. Local development records use a separate database and are not public visitor data. We do not make copies of the public event database for long-term backup. Experiments store decisions and evidence references, not visit identifiers.

These optional events are used to find confusing steps and evaluate improvements. They stay on the operator’s server. No PostHog delivery or session replay is enabled. We may earn a commission if you buy through a labelled partner link. Labelled partner offers can be shown whether or not you enable optional measurement. Checking the service works without optional measurement. Anonymous visit identifiers are not linked to contact messages; reload to start with measurement off.

Where visitors come from

When a page opens, our server counts which website linked to it and the campaign tags in the link. We keep only the name of the linking website (for example “news.ycombinator.com”, never the address of the linking page) and the values of utm_source, utm_medium and utm_campaign (for example “reddit”). We store only daily totals per name or tag, kept for 90 days, with no IP address, no full address, no identifier and no cookie. Visits with neither a linking website nor a campaign tag, recognisable bots, internal network names and our own pages are not counted. So that one connection is counted only once per day, and one network only a few times, our server keeps salted fingerprints of your IP address (for IPv6 your /64 network) and of the larger networks it belongs to, together with the kind of source, in its working memory until the end of that day (UTC). These fingerprints are deleted within about a minute after midnight UTC or when the server restarts, and are never written to disk. The totals show how many different connections came from each source per day, not people. Anyone can send a false linking website, so the totals are indications, not proof.

The same counter also counts use of our MCP server (the interface AI assistants call at /mcp), so we can see which tools are used and decide what to improve. After a successful answer it adds one to the daily total of the method or tool name, for example “initialize”, “tools-list” or “check-website”; answers that report a tool error are not counted (we read only that error flag). It never stores what was asked or answered (no website names you check, no error texts), no IP address and no identifier. Each connection counts each method or tool at most once per day, with the same kind of fingerprints, network limits and 90-day retention as above.

“Was this helpful?” answers

At the end of our guides, status pages and check results you can answer “Was this helpful?” with yes or no. This is voluntary feedback, separate from the optional usage measurement. We store only the totals of yes and no answers per page (for a check result: per kind of result, such as “503 Service Unavailable”), with no IP address, no domain you checked, no free text and no time of the answer. So that one person cannot easily vote many times, your browser notes in this tab’s session storage that you answered a page; the note disappears when you close the tab and is never sent to us. Our server also counts only one answer per page and day from one connection and limits how many answers come from the same network. For this it keeps your IP address (for IPv6 your /64 network) and the larger networks it belongs to, together with the page you rated but not your answer, in its working memory only. These entries are deleted at the latest 24 hours after your last answer (the clean-up runs every minute) or when the server restarts; they are never written to disk. Totals below 30 answers per page are too small to decide anything and are not used for decisions.

Advertising (Google AdSense)

This website uses Google AdSense to pay for running the free checker. Every page loads Google’s AdSense code from pagead2.googlesyndication.com, also while no advertisement is shown yet. Google, as an independent provider, receives technical data such as your IP address, browser and device information and the page you are viewing, and may set cookies or use similar technologies to show ads, limit how often you see them, measure them and prevent fraud. Data can be processed in countries outside Switzerland and the EU, including the USA.

Google’s consent message for visitors in the European Economic Area, the United Kingdom and Switzerland is configured in our AdSense account. When Google delivers the message, it offers advertising choices. The account setting alone does not prove that the message appears for a new visitor; its delivery has not yet been verified in a fresh visitor session. We do not treat a missing message as consent. Google’s code still loads on every page as described above, and may process data or set cookies even while no advertisement is shown. You can block Google’s code and cookies with your browser’s privacy settings or an ad blocker. The checker works with or without them. If the choice is missing or cannot be changed, contact us using the address above. Our own optional usage measurement is separate and stays off unless you turn it on.

Labelled VPN partner offers may appear on the home page before you run a check. In a check result they follow the result and free troubleshooting steps. These offers are separate from Google AdSense. Their visibility does not depend on optional usage measurement. Following a partner link opens the provider or its affiliate network, which receives the connection information needed to handle the visit and may use its own attribution technologies under its privacy policy. We do not send your checked domain or report in these partner links. A commission is possible only under the partner’s terms; a link click is not proof of a purchase.

Google explains its processing in How Google uses information from sites that use its services and its privacy policy. You can manage personalised advertising in Google’s ad settings.

Contact and external links

If you email us, your message and email address are used to answer and handle your request. Keep messages relevant and avoid secret data. We retain correspondence while needed for the request and any applicable obligations. External links open the other provider’s website and are subject to its policies. Provider status links are helpful references, not extra live measurements.

What the signals mean

Read the method, colours and limits. One measurement point cannot prove a worldwide outage. For a website that fails for you, start with the free troubleshooting steps.