Archive traffic.

Explore the people, search engines, and AI crawlers visiting this collection of websites.

All sites
Sites and sections
Loading report…

Activity overview

Human page views
—

Pages that loaded for verified readers
Visits
—

Verified visits that reached a page
Search requests
—

Search-engine crawlers
AI requests
—

Training crawlers and assistants
Total requests
—

Everyone: readers, crawlers, probes and other automation

Large number: succeededStriped: failed with an error

Activity over time

View daily values

Who's requesting?

Share of all logged requests, and how many of each failed

Visitor behavior

Reading patterns

Pages per visit
—
Observed visit duration
—
Single-page visits
—
Downloadable file responses
—
Pages served in ≤1 second
—
Visits that met an error
—

Duration uses visits with a measurable interval.

Where readers came from

Referral sources at the start of a visit

Site breakdown

Large figures succeeded; the red figures under them failed.

Reading hours

Human page views by weekday and hour · UTC

Most-read pages

All observed history · selected sites

    Errors

    Requests the server answered with a failure (status 400 and above). None of these are counted as page views or visits.

    Errors readers met
    —Failed requests inside verified visits
    Pages that failed for readers
    —Share of the pages readers asked for
    Server errors (5xx)
    —The site itself failed
    Security probes
    —Attempts to find a weakness
    All failed requests
    —Readers, crawlers, probes and other automation

    What failed

    Failed requests by the server's answer

    Who met the errors

    Failed requests by who asked

    Addresses that fail

    Most affected first: addresses readers ran into, then server errors, then the most requested. Selected sites.

    AI collection

    All observed history · selected sites

    Estimated training coverage

    —

    Based on successful collection of pages in the current inventory.

    —Local media fetched in full

    Collection by organization

    Potential model families · share of inventoried HTML pages

    This estimate counts a page as training material when an identified training crawler retrieves it successfully (answer 200); fetches that failed are not counted. Logs show collection; the organization’s final dataset, model version, and use of each page remain unverified.

    Who discovered the same pages?

    Every inventoried HTML page counts once. Compare the reach of verified readers and training crawlers across the observed history.

    How to read this report

    How is page serving time measured?

    Apache times successful page responses for verified readers. The percentage counts those served within one second. Loading images, running scripts, and drawing the page in the browser can take longer. Timing begins when the updated log format is deployed.

    What counts as a downloadable file response?

    Inside verified visits, this counts GET responses for PDFs, ZIP and 7z archives, MP3, FLAC and WAV audio, and EXE downloads. Apache must report a successful 200 or 206 response with bytes. A range response (206) can represent just one piece of a file. One person can produce several responses while streaming or resuming. Cache checks, errors and known interrupted transfers are excluded. Older log formats can include header bytes and lack a connection outcome. A response alone cannot establish that someone saved the whole file.

    What counts as a human page view?

    A page view is one successful HTML page shown by a verified browser. The pictures, style sheets and scripts that a page pulls in are never views, however many there are. Redirects are not views, and the same page requested again within ten seconds is the same view.

    A browser's name proves little: most requests that call themselves Chrome or Safari come from scrapers that fetch one page each from a fresh address. A reader's browser also asks for the page's pictures. So a visit is verified when, after a page request, the same visitor is served one of the page's pictures or its video within 30 seconds, or when the server answers that the visitor's saved copy of the page is still current. Style sheets, scripts and the site icon are not enough, because scrapers and scanners fetch those too. A visitor that asks for robots.txt or pulls six pages inside five seconds is treated as automation; one that probes for known software flaws is counted with the security probes.

    On each host, a visit ends after 30 minutes of inactivity or at UTC midnight. Once a visit is verified, its later pages count even when the browser already had the pictures saved. Visits count toward the section where they began; views count where they occurred. Readers who block pictures, and returning readers whose browser had everything saved, can go uncounted; automation that loads pictures at a human pace can be counted. Duration measures the interval between page requests in timed visits. Time spent on the final page is unknown.

    What do "succeeded" and "failed" mean?

    A request failed when the server answered with an error (status 400 or above). Everything else succeeded, including redirects and "not changed" answers to a browser that already has the page. The large number in each box at the top counts what succeeded; the line below gives all requests and how many failed, and the bar shows the two side by side, failures striped.

    For readers the comparison is between pages: page views count pages that loaded, and the line below adds the pages that failed for verified readers. A visit counts as successful when the reader reached at least one working page; a verified reader who met only errors is a failed visit.

    What counts as a security probe?

    A request that only makes sense as an attempt to find a weakness: asking for secrets and configuration files (such as /.env or /.git/config), for software these sites have never run (WordPress logins, phpMyAdmin, Joomla administration, router and server consoles), for backups and database dumps, or carrying attack code in the address (database injection, script injection, climbing out of the site's folders). Unusual methods such as CONNECT and PROPFIND also count.

    A request for a probe address counts as a probe whoever claims to send it. A visitor that presents a browser's name and probes is a scanner in disguise, so everything it asked for that day counts as probing, including its first look at the home page. Self-declared tools and crawlers are counted request by request. The sites' own old addresses (forum pages, the retired pool interface, the Village's old forms) are not probes, even though old programs and links still request them and fail.

    How are errors counted?

    Every request the server answered with status 400 or above is a failed request, whoever sent it. They are grouped by the answer: not found (404), retired on purpose (410), refused (401 and 403), other request errors, and server errors (5xx), where the site itself failed. Failed requests never count as page views, and a visit in which a reader reached no working page is not counted as a visit.

    "Errors readers met" are the failed requests inside verified visits: a page that would not load, or a picture or file missing from a page. Requests a browser makes on its own, such as the one for a site icon, are listed but left out of that count because the reader sees nothing fail. Most other failures come from security probes and from old programs still calling addresses that were retired.

    The list of failing addresses covers the latest 30 days. An address is named only when it is a known public page, was linked from the site itself, or failed at least three times; rarer addresses and all security probes are merged, and query strings are never recorded. A folder with many failing files appears once, ending in an asterisk.

    How are bots classified?

    Crawlers that name themselves are classified by the user agent they report. Search indexing, training collection, and other agent activity are counted separately. Requests with a browser's name that were never verified as a reader's visit are reported as unverified browsers; nearly all of them are automation. Security probes are counted separately from all of these. These identities can be imitated. OpenAI identifies GPTBot as a training crawler and Anthropic identifies ClaudeBot as a model-development crawler. User-initiated fetches are excluded from training coverage. OpenAI crawler guide · Anthropic crawler guide

    What does coverage measure?

    The headline percentage compares distinct, currently inventoried public HTML pages with successful training-crawler fetches. A gallery page can count even if the crawler never downloaded the photograph. The separate media figure counts local images, audio, video, and PDFs retrieved in full. Different sizes of the same image count as separate files. Google Drive originals and YouTube streams are outside the server logs. Totals count a page or file once even when several companies collect it. Organization percentages can overlap. The coverage panel uses all observed history regardless of the activity date filter.

    What do the logs include?

    The reports use server-side Apache logs. Cached reads may never reach the server. The first and last observed days can be partial. A date with no records can also indicate a gap in logging. All dates are UTC.

    Download the aggregate JSON