No Core Update Since May: Reading Unconfirmed Volatility

The most common diagnostic error in an SEO’s week is attributing movement to an update that does not exist. As of 24 September 2026 the most recent Google core update began on 21 May 2026 and ran for 11 days and 21 hours. There has not been another one in June, July, August or September. Anything that moved in those four months moved for some other reason.

What Google has confirmed in 2026

The full list is short, and it is worth knowing by heart because it is the only authoritative record. From the ranking history on the Google Search Status Dashboard, with Google’s own durations:

• February 2026 Discover update — 5 Feb 2026, 21 days 17 hours.
• March 2026 spam update — 24 Mar 2026, 19 hours 30 minutes.
• March 2026 core update — 27 Mar 2026, 12 days 4 hours.
• May 2026 core update — 21 May 2026, 11 days 21 hours.
• June 2026 spam update — 24 Jun 2026, 2 days 1 hour.
• August 2026 spam update — 18 Aug 2026, 2 days 16 hours.

Two core updates all year, both in the first half, and the dashboard notes that all times are US/Pacific. Third-party volatility trackers have reported spikes through August and September that correspond to none of these. That gap between “nothing confirmed” and “the rankings moved” is where most of the bad decisions get made.

“Unconfirmed update” is a category error

Google has never undertaken to announce every change. Its guide to Search ranking systems states that it regularly improves its ranking systems through testing and evaluation and provides notice of updates “when those might be useful to content creators and others”. Notice is a courtesy extended when it would help, not a log. Google has also said for years that it makes changes to Search continuously, of which the named updates are a small subset.

So “unconfirmed update” describes the observer, not the event. It means a third-party tracker registered variance. That variance might be a systems change, or it might be sampling: most trackers watch a fixed keyword set through a fixed set of locations, and a results page that adds or removes a feature registers as rank movement across every site below it without any site having changed.

Movement without a named update

The ordinary causes are unglamorous and account for most of what people bring to a support queue. Index refresh, as pages are recrawled and re-evaluated at their own pace. Query interpretation shifting, so a term that used to return comparison pages starts returning product pages. Results page composition changing, which alters clicks with no change in position at all. Competitor movement, including a large site publishing into your topic. And your own site: a plugin update that altered canonical output, a template change, a crawl budget shift after a migration. The ranking systems guide is also explicit that the systems work at the page level and that good site-wide signals do not mean everything ranks well, nor poor ones that everything ranks badly, which is a useful corrective when a single URL moves and the instinct is to diagnose the domain.

A triage order that stops you chasing ghosts

Verify the data first. Search Console has had repeated reporting incidents, and a chart can be wrong before a site is. Then segment by country and device, because a single-market change is almost never an algorithm story. Then isolate the query class: if informational queries moved and transactional ones did not, that is a results page composition question. Then look at impressions against clicks, because a flat impressions line with falling clicks is a SERP problem and a falling impressions line is an eligibility problem. Only at that point does the page itself become the subject. The systems-level context for that sequence is in what changed when the helpful content system merged into core, and the reporting side in making SEO decisions from GA4.

When waiting is the right call

During a confirmed rollout, waiting is correct: Google publishes a start date and a duration, positions move while it runs, and changes made mid-rollout cannot be attributed afterwards. Outside a rollout the opposite applies. There is nothing to wait for, and a quiet period with no confirmed core update is the best available window for work whose effect you actually want to measure — consolidation, internal linking, the coverage decisions in what the topical authority research shows. Four months of silence is not a reason to hold still.