Changelog.

What changed in pangoling, one release a week. RSS feed

Week of 26 October 2026

Better data for every site, fewer false alarms

This week's update gives webhook sites the same data as git sites, checks generated titles for typos, makes the site check quieter and more accurate, makes the plan and ranking numbers match what the site really has, and starts measuring how AI assistants talk about your site without any setup.

New

  • Starter questions: when a site tracks no AI-search questions yet, we pick them from the searches people already find the site with, within your plan's question count. This can be switched off per site.

Improvements

  • Webhook contract v2: the payload now carries the first-publish date, the modified date, keywords, related pages, entities, canonical address, reading time, breadcrumbs and FAQ data. Existing receivers keep working: they ignore keys they do not know.
  • Related pages: the pages a piece links to are listed with their titles, so a site can build its related section from data instead of matching text.
  • Typo check: a generated title or description is checked against the page before it goes live. A misspelled version of a word on the page is retried once with the word named, then refused, and the page keeps its current title.
  • Unknown words: on pages with enough text, a word the page never uses is held to the same retry and refusal.
  • Topical plan: a keyword that Search Console or the rank check already ties to a page on your site now shows that page instead of asking you to write it. Weak matches (far down the results, or with almost no impressions) stay as a hint.
  • Tracked keywords: the stale and tracked counts in reports now cover only the keywords you follow, not every keyword ever pulled.
  • Rank checks: the keyword list now shows why the last rank check skipped a site, such as used-up credits or the monthly search-data budget.
  • Reality check: it now says when no AI-search questions are tracked, so a missing measurement is not read as a zero.
  • Norwegian: question templates now work for sites set to Bokmål or Nynorsk.

Fixes

  • Fixed the site check reporting pages as unlinked when it had only looked at a small sample of the site.
  • Fixed language-link warnings on pages that exist in one language only.
  • Findings that have not been re-checked for three weeks now show as low priority instead of staying at full severity.
  • Fixed one very large keyword list using up the whole rank-check run and leaving other sites unchecked.

Week of 19 October 2026

Refreshes that publish themselves

This week's update is about keeping live pages current without extra work. Refreshes now go live on their own after the quality check, and the weekly email only reports moves that are real.

New

  • Automatic refreshes: on sites where we have turned on automatic refresh, a refreshed page goes live after it passes the quality check. A page that fails stays as it was.
  • Nightly internal links: pages Google has not found yet get links woven in from related pages every night.
  • IndexNow: Bing and other search engines are told about new and updated pages when they publish.

Improvements

  • Weekly email: ranking moves are verified before the email goes out, and junk queries no longer appear in it.
  • Title rule: live pages now follow the title rule of sentence case and 60 characters or fewer.
  • Refresh quality: a refresh is edited in place, repaired before it is held, and the live text is restored if it is held.
  • Sources: cited sources are checked again before a queued piece goes live.

Fixes

  • Fixed refreshes being held for old failures instead of new ones.
  • Fixed action lines in the weekly email losing their action.
  • Fixed a page being reported as unranked after a single blank reading.

Week of 12 October 2026

Seeing everything your site exposes

This week's update is about the full footprint of your site. pangoling now reads every URL a site shows to search engines, groups them by template, and says which ones to keep, fix or cut.

New

  • Footprint audit: reads every indexable URL on your site and groups them by template, so one fix covers a whole family of pages.
  • Thin, duplicate and noindexed findings: listed per template on Optimize, under Technical.
  • Cut list: after a Google update, pangoling proposes which weak pages to remove or merge, each with a reason. Nothing changes until you approve it.

Improvements

  • Update verdicts: when a site has too few clicks to judge a Google update, the verdict now reads impressions instead.
  • Footprint audit: slows down when your server asks it to, and skips paths that are not pages.
  • Measure: every checkpoint now compares against an expected baseline, so a lift is judged against normal movement.

Fixes

  • Fixed links on pages that pangoling adopted from your own site files not being picked up for refreshes.
  • Fixed false findings in the site audit and AI search runs.
  • Fixed counts that disagreed between pages.