Analytics

Weekly traffic for apis.io from three independent sources — the CDN edge, a page-view beacon, and fetches of the markdown twins by AI agents. They measure different things, and the gap between them is the point.

Last updated: August 17, 2026  ·  34 weeks of data

Most Recent Week 2026-08-10 → 2026-08-16

AWS CloudFront Edge / CDN layer
6,479,429
Edge Requests
186.702 GB
Bandwidth
8.77%
Error Rate (4xx/5xx)

Server-side totals for every request the CDN served — pages, assets, /api, /mcp, and bots alike. Page views, uniques and per-country breakdowns live in the Google Analytics panel.

Google Analytics 4 Browser / session layer
213,379
Page Views
206,851
Sessions
206,717
Total Users
206,742
New Users
95.4%
Bounce Rate
0m 8s
Avg Session

Top Countries
Singapore 136,688 United States 45,986 Brazil 5,792 Netherlands 4,066 Japan 3,380 China 2,961 United Kingdom 2,507 Australia 2,359

Traffic Over Time

Agent Traffic markdown twins

Markdown Twins Agent layer
6,322
Twin Fetches
3,744
From AI Agents
59.2%
AI Share
22.9 MB
Served

Every provider and API page on apis.io publishes a markdown twin at its own URL — identical bytes to every client, never varied by user-agent. This counts fetches of those documents, read from CloudFront access logs (3,466,764 logged requests scanned). Neither source above can see this: agents run no JavaScript, so Google Analytics reports zero for them by construction, and edge metrics count the request without knowing what it was.
Partial week — counting began 2026-08-13.

Which agents
browser 2,087
amazonbot AI 1,278
meta-externalagent AI 1,219
oai-searchbot AI 1,083
googlebot 405
gptbot AI 161
other 61
other-bot 24
perplexitybot AI 1
chatgpt-user AI 1

Named AI crawlers are marked. Search and SEO bots are shown but not counted in the AI share. browser is almost certainly agents presenting a browser user-agent — humans do not hand-fetch index.md — so the AI share is a floor, not a ceiling.

Daily twin fetches

Weekly History

Week ending Edge Requests Bandwidth (GB) GA Page Views GA Sessions GA Users Req / View Twin Fetches AI Share
2026-08-16 6,479,429 186.702 213,379 206,851 206,717 30.3× 6,322* 59.2%
2026-08-09 6,561,545 159.318 104,859 101,108 100,002 62.5×
2026-08-02 9,548,627 195.505 526,240 502,881 498,730 18.1×
2026-07-26 6,373,099 122.41 124,539 95,693 95,073 51.1×
2026-07-19 5,218,191 74.862 111,200 103,870 105,220 46.9×
2026-07-12 6,018,956 71.209 123,654 115,883 116,467 48.6×
2026-07-05 6,515,760 68.784 229,654 222,191 219,645 28.3×
2026-06-28 5,363,220 77.473 161,481 152,684 149,272 33.2×
2026-06-22 1,881,905 25.596 74,687 68,336 69,888 25.1×
2026-06-21 1,320,865 16.764 75,183 70,441 69,197 17.5×
2026-06-14 2,073,492 57.492 58,598 54,197 54,379 35.3×
2026-06-10 2,359,591 46.849 48,367 45,731 44,508 48.7×
2026-06-03 1,386,946 41.569 13,951 13,113 13,052 99.4×
2026-05-27 1,156,273 44.582 3,352 2,278 2,173 344.9×
2026-05-20 753,178 29.203 6,704 6,166 6,087 112.3×
2026-05-13 713,857 21.126 11,483 11,196 11,118 62.1×
2026-05-06 443,417 10.23 981 432 368 452.0×
2026-04-29 485,605 5.959 1,503 1,301 1,266 323.0×
2026-04-22 472,289 6.1 938 442 400 503.5×
2026-04-15 368,448 6.909 376 253 220 979.9×
2026-04-08 330,832 3.533 321 197 181 1030.6×
2026-04-01 171,944 1.003 249 204 190 690.5×
2026-03-25 128,668 0.698 264 226 213 487.3×
2026-03-18 119,763 0.666 579 357 307 206.8×
2026-03-11 111,269 0.609 345 280 259 322.5×
2026-03-04 227,967 1.314 714 423 352 319.2×
2026-02-25 160,243 0.874 776 483 432 206.4×
2026-02-18 98,957 0.579 915 427 363 108.1×
2026-02-11 94,345 0.554 650 425 380 145.1×
2026-02-04 82,281 0.486 443 346 329 185.7×
2026-01-28 83,499 0.43 466 364 335 179.1×
2026-01-21 50,792 0.264 383 305 293 132.6×
2026-01-14 55,066 0.324 476 347 312 115.6×
2026-01-07 72,047 0.588 513 350 323 140.4×

Why the Numbers Differ

CloudFront counts more
  • Every HTTP request the edge served — pages and every asset (CSS, JS, images, fonts)
  • Bots, crawlers, feed readers, and API / MCP clients hitting /api and /mcp
  • Cached responses served at the edge (never reach the S3 origin)
  • Requests that load before any JavaScript runs
  • Clients with JavaScript disabled or blocked
Google Analytics counts less
  • Only fires when JavaScript loads and executes successfully
  • Ad blockers and privacy extensions suppress the GA tag (significant on developer audiences)
  • Safari ITP and Firefox ETP limit cross-site tracking
  • Bots and crawlers are filtered out automatically
  • Session-based model groups rapid multi-page visits together
Markdown twins count a different audience entirely
  • Agents execute no JavaScript and request no tracking pixel, so Google Analytics reports zero for them no matter what actually happens — this is not a gap that can be closed by fixing the tag
  • CloudFront does count the request, but a distribution-wide metric cannot tell a twin fetch from any other hit — which is why this number comes from the access logs instead
  • A twin fetch has no click-through and never will, so it is reported on its own and never blended into a rate alongside human traffic
  • The same bytes are served to every client. Nothing here varies by user-agent, which is the whole point on a network that publishes the tool auditing that behavior
What to watch: the requests-per-view ratio is as interesting as the raw numbers. Because CloudFront counts every asset on a page (not just the page itself), a single page view is naturally many edge requests. A rising ratio can indicate more bot/crawler traffic, heavier API/MCP usage, or growing ad-blocker usage among our developer audience; a falling ratio can indicate fewer crawlers or improved real-user engagement. Both trends together tell the story of how the site is actually growing.