Encited
Technical overview

How Encited Edge serves your pages

The ideal page for humans, search crawlers and AI crawlers is not the same. With Encited, you can treat each of them as a first-class audience.

  1. 0.1Edge cache responses in 15–30ms
  2. 0.2Cold cache reads in 50–100ms, then copied to the edge
  3. 0.3HTML for search engines, Markdown for AI agents
One URL, four readersPlays each request in turn

But how do these caches come to be?

Below is the full process, from the Encited pre-rendering fleet visiting your pages to post-processing and caching them.

The pipelineOne page, start to finish
01

It starts with your refresh rules

Configure re-caching rules and schedules by URL group, each with its own browser profile.

  1. 1.1Daily, weekly, monthly or quarterly refresh per path
  2. 1.2Assign each URL group its own locale, time zone and device for fully localized snapshots
  3. 1.3Built-in manual re-caching for off-schedule refreshes, no more messy scripts
Fig. 1 · One week of refreshesDue pages join the queue
02

Health check before render

Each page is probed before it renders, and broken pages raise an alert.

  1. 2.1Page alerts for server errors, redirects and blocked pages
  2. 2.2Broken pages are never cached or served
  3. 2.3No render spent on a page that fails its check
Fig. 2 · A healthy page and a failing oneAlternates
03

Then, render fleet visits

Pages are rendered on your configured schedule and attached browser profiles.

  1. 3.1Desktop or mobile rendering per path
  2. 3.2Language and time zone per path
  3. 3.3Localized pages render in their own language
Fig. 3 · Three profiles, one siteActive profile in blue
04

Load and wait until page is settled

Before taking the snapshot, Encited makes sure your page is in its fullest form, with no data left out.

  1. 4.1Waits for the network, event loop and content to settle
  2. 4.2Ignores beacons and connections that never close
  3. 4.3Honors an explicit “ready” signal from your page
Fig. 4 · From navigation to snapshotQuiet periods in green
05

No accidental DDoSes allowed

Render speed follows your server's response time and error rate.

  1. 5.1Speeds up while your server is fast
  2. 5.2Backs off on slow responses or errors
  3. 5.3No load spikes for your real visitors
Fig. 5 · Concurrency follows the originValues for illustration
06

Plain rendered HTML is hard to read, though

On average, modern web pages are 70%+ non-HTML, non-content. Scripts, styles, inline SVGs and decorative markup make the real content hard to read and extract.

  1. 6.1Strip scripts and style blocks
  2. 6.2Strip decorative markup
  3. 6.3Strip inline SVGs
  4. 6.4Pure content for search engines and AI agents
Fig. 6 · Tokens per pageOne product page

Get recommended in AI answers

Optimize for Google, AI and humans without compromises.

Start 14-day free trial Talk to sales