A publisher's site is a large archive under a small number of templates: article, category, tag, author, date archive, and pagination of each. The archive pages exist because the CMS creates them, not because anyone decided to, and on most sites they outnumber the articles. Each one is a thin page or a duplicate of a category, and together they are the largest structural finding on any content site.
Content sites also feel Google's core updates more than any other kind, because those updates are about content quality. That makes telling an update from a regression the single most valuable measurement skill a publisher can have: a template change that shipped the same fortnight as an update will be blamed on the update, or the update on it, unless the two are separated deliberately.
The assistants are the newer problem. AI Overviews and answer engines quote publishers without sending the click, and the only way to know how often is to ask the questions yourself, since Search Console will not tell you.
The findings that recur on this kind of site
- Tag pages in their thousandsOne tag, one article, one indexable page. Thin content created by the CMS on every publish, which nobody chose and nobody prunes.
- Pagination with the page number missing from the title'News — Page 2' rendered as 'News'. Duplicate titles across every paginated archive, fixed once in the template.
- Author and date archives duplicating the categoryThe same articles listed three ways, and canonicals pointing at each other after a theme migration that half-updated them.
- Old articles with no internal linksPieces from five years ago that were once linked from the home page and are now reachable only from page 84 of the archive — effectively orphaned, some of them still earning impressions.
- Article schema missing after a redesignThe missing-schema finding on every article because the new theme did not carry the markup over. One of the more common things a redesign silently drops.
What to change first, and what to compare it against
- Fix the archive templates before touching a single articleNoindex the tag and date archives, put the page number in paginated titles, resolve the canonicals. Three template changes cover most of the findings, and they rank above any article-level fix by the impressions they touch.
- Separate content changes from template changes in timeShip a template fix, freeze a control set, wait the window. Then start the content work. Two kinds of change in one month produce one uninterpretable chart.
- Judge article improvements by section, against untouched sectionsRewrite the pieces in one category and leave a comparable category alone. If both rise at the next update, it was the update. The method is the same as for any other site; publishers simply have more sections to use.
- Track citation rate alongside impressionsA fixed set of questions your articles answer, asked weekly to the assistants. Being cited without the click is a real outcome and it needs its own number — share of voice rather than traffic.
Questions from people running one
Should old articles be deleted?
Pruned, not deleted by default. An article with no impressions in sixteen months and nothing to say is a candidate for removal or merging; one with even modest impressions is earning something and should be updated and linked rather than cut. Check Search Console per URL before deciding; the archive is usually less dead than it looks, and rarely as alive as the publisher hopes.
A core update dropped our traffic 30%. Was it the redesign?
Possibly, and it is checkable. Crawl the site now and compare the SEO-relevant fields against the last crawl before the redesign: canonicals, schema, titles, indexability. If a template lost something in the same fortnight as the update, the fix is concrete. If nothing changed on your side, the update is the explanation and the response is editorial, not technical.
Everything above is the method, usable with any crawler and a Search Console export. Rankelle runs it on a schedule: the crawl, the grouping into causes, the watch for the fix landing, the frozen control set and the verdict four weeks on — including the verdicts that say the work did nothing. One site is free, permanently, because the first verdict takes six weeks to arrive.
Try it on your own site, free
The whole audit of every website you add, no card and no expiry. Pay only when you want the proof.