Stats

A public view of pages served by this site.

Traffic

Period

Loading stats…

Page views

Daily clients

Device mix Browsers only

Desktop

Mobile

Tablet

Page views over time

View chart as a table

Top pages by views

Reported referrers by views

Between blog pages reported transitions

Countries by views

Who fetched these pages by kind and reason

What these numbers mean

Reports can be up to one hour old. The updated time shows when a report was calculated; changing a filter may select a report calculated at a different time.

Reported referrers describe each page request. Within this blog means an internal referrer was sent; No referrer sent means the header was absent or empty; Unusable referrer means a value was received but could not supply a supported hostname. Unknown — earlier collection covers historical records where those cases were merged. These categories are not a count of first-time arrivals.

Page transitions use recognized public internal paths, with query strings and fragments removed. Select a page to see both previous and next blog pages. Time and traffic filters apply to the request for the destination page. Repeated requests count again; same-page referrals are separate. These observations do not establish complete journeys, exits, or a conversion rate. Page-transition prior art.

Referrers are claims supplied with requests and can be fabricated. Only reviewed hostnames are named here; other names are grouped as Other reported referrers, without treating unfamiliar sources as bots. A pinned Matomo spam list and reviewed local rules exclude matching observations from every public metric, including All; the excluded count is shown for your selection. Referral evidence is retained. These rules also apply to past dates, so reported history can change when the policy changes. An approved name does not prove a genuine referral. Referral-abuse policy.

A page view is one recorded page event. Since the August 26, 2026 cutover, that means a successful, non-prefetch GET that served HTML or the Markdown twin of a page at the edge. API requests, assets, redirects, errors, feeds, and other non-page responses do not count. Recording is best effort. Rows marked as mine are excluded from public queries, but that marking depends on server-side configuration and is not currently proven for all of my requests.

Every row carries a classification and the rule that produced it. Browsers: requests matching this site's browser rules, described below. AI agents: declared on-demand fetchers such as ChatGPT-User and clients with a verified Web Bot Auth signature. A declared name or valid signature does not establish that a person requested or read the page. Crawlers: search engines, AI search indexers, AI training crawlers, and link-preview fetchers, all matched on the User-Agent they declare. Automation: browsers running on hosting networks, headless browsers, HTTP clients, generic bot tokens, and browsers too old to send the headers this site checks. The four groups are disjoint and add up to All.

Browsers is not a verified-human count. It holds three kinds of evidence, listed by name in the composition table. Since September 3, 2026: a navigation-shaped request, meaning a Sec-Fetch-Mode of navigate, a Sec-Fetch-Dest of document, an Accept that admits HTML, and an Accept-Language, from a network that is not on this site's list of hosting providers. Between August 26 and September 3, 2026: the User-Agent alone, because the site did not record headers yet; for most of those rows the network was reconstructed afterwards from Cloudflare's zone analytics, and rows that turned out to sit on hosting networks moved to Automation. Before August 26: rows from the former browser beacon, which ran this site's script inside a page. Expect the Browsers series to drop across those two dates. That is the method changing, not readers leaving.

A hosting network is a verdict on its own. A browser on a cloud host that sends every header a person's browser sends is still counted as Automation, because that is where the bulk of the inflation came from. A missing Fetch Metadata header is a verdict only when the User-Agent claims a browser version that always sends it: Chromium 76, Firefox 90, Safari 16.4 and later, including their WebViews. Older claims are listed as Old browsers within the Automation filter; that grouping does not prove automation. A crawler match does not prove that a model read, indexed, cited, or used a page. Filtering by a page, a kind, or a named agent applies the same window and group to every panel on this page.

A daily client is a per-day pseudonymous identifier, not a person. Edge observations use a secret HMAC over the site, UTC date, IP address, and User-Agent; historical beacon rows use their original daily identifiers. Neither method establishes the same client across dates, devices, or networks. Hourly daily-client values are not additive.

Privacy details · Engineering decision · Browser evidence · Audience composition