Re-indexing & AI-crawl requests

When you act on a recommendation and ship changes to a page, the answer engines only reflect it once their underlying indexes re-crawl and re-fetch that page. This section explains what Lokrix can nudge, what it cannot, and how long to wait before re-scanning.

Why re-indexing matters for AI visibility

Most engines do not read your page live at answer time from scratch — they ground on an index. ChatGPT retrieves via the Bing index (around 87% of its citations match Bing's top ten results); Copilot also grounds on Bing; Gemini and Google's AI surfaces ground on Google's own index. So a change is invisible to those engines until Bing or Google re-crawls the page. Perplexity and Grok run live search and can pick changes up sooner, but still depend on the freshness of the results they retrieve.

What Lokrix can and cannot do

Lokrix does not have a private channel to force any engine to re-read your page — no such button exists in the underlying systems. What it does is surface the standard, official levers and track when a change should be visible:

  • Google re-indexing — request re-crawl through Search Console's URL Inspection / indexing tools for pages you own, using your connected Search Console account.
  • Bing / Copilot — submit URLs via Bing Webmaster Tools so the Bing index (which ChatGPT and Copilot ground on) picks up the change.
  • Crawl hygiene — confirm the page is reachable, not blocked by robots rules, and returns fresh content, so a re-crawl actually captures your edits.

Note that llms.txt has no measured effect on whether engines cite you, so Lokrix never treats it as a re-indexing lever.

Timing and re-scanning honestly

Re-indexing is not instant and the exact lag is outside Lokrix's control — it can be days for Google or Bing to re-crawl and reflect a change in AI answers. Because answers are non-deterministic, a single scan right after a change can mislead. Wait for the index to catch up, then run a fresh scan so the new Monte-Carlo presence probability reflects the update with a proper confidence interval rather than one lucky or unlucky run.