Two Google changes, one Search Console chart
Crawlmind Engineering··5 min read
An attribution window is the stretch of dates you compare to decide what caused a change in your traffic, and this November Google gave sites two overlapping causes inside the same window.
The first is the November 2026 core update. Google's Search Status Dashboard shows it began on November 10, 2026 at 10:06 Pacific, applies globally and to all languages, and "may take up to 2 weeks to complete." As of today the dashboard has no completion entry. Search Engine Journal notes it is the third core update of 2026.
The second is a model change. On November 12, Google made Gemini 4 Hydrogen the default model in AI Mode for all users worldwide, two days into the core update rollout. Google's own announcement on the Search blog says Hydrogen's reasoning improves its query fan-out technique and connects people to "a broader range of helpful links than before." Per SEJ's account of Robby Stein's post on X, Hydrogen is also being used for the hardest questions in AI Overviews. Neither Google nor Stein gave counts for links or sites, and "hardest" was not defined.
These are different mechanisms. A core update changes how Google's ranking systems assess pages. A model swap changes how an AI surface expands a question, which sources it retrieves, and which of those it cites. Both can move the same numbers in the same week. If you read your Search Console chart for the second half of November as one story, you will likely blame the wrong cause.
#Why the two changes blur together in reporting
Search Console now has a generative AI performance report that counts how many times links to your site were shown in AI Overviews and AI Mode. Google's help page says it rolled out to all websites worldwide as of August 31, 2026. It reports impressions, not clicks or queries. The same page says the report includes data from the Web search type in the Performance report, so AI feature appearances and classic results are not two clean, independent pools.
That matters for both changes at once:
- If the core update lowers a page's standing, fewer of its links may surface in classic results and in the AI features that draw from the same index.
- If Hydrogen's fan-out generates different sub-queries, your page may be retrieved for questions it was never retrieved for, or dropped from ones it used to win, with no change to its ranking.
- A drop in Web search impressions starting November 10 could come from either change, or from both pulling in opposite directions.
Google's core update guidance is written for one change at a time. It says to confirm the update has finished, wait "at least a full week after a core update completes," then compare that week with a week before the rollout began, and to look at Web, Images, Video and News separately. It does not address a second system change landing mid-rollout.
#The calendar makes it worse
If the rollout runs the full two weeks, it ends around November 24. A full week after that runs into the start of December. That comparison week contains Thanksgiving (November 26), Black Friday and Cyber Monday, a point SEJ raised in its weekly roundup. For retail, travel and anything gift-adjacent, demand in that week is not comparable to an ordinary week in early November.
So the clean comparison Google recommends will mix three effects: ranking changes from the core update, retrieval and citation changes from the model swap, and holiday demand.
#How to separate them
None of this requires new tooling. It requires logging dates and splitting data you probably already have.
Annotate both dates, not one. Mark the core update start and the Hydrogen switch in AI Mode as separate events in whatever reporting you use. When the dashboard posts a completion date, add it. A chart with one annotation invites a one-cause explanation.
Split by surface before you split by page. Compare the generative AI report against the full Web search type trend. If AI Mode and AI Overview impressions moved sharply from the day of the model switch while the overall Web trend moved gradually from the start of the core update, the model swap is the more likely driver for the AI-feature change. If both move together from the core update start, ranking is the likelier cause. This is a direction, not proof, because the AI numbers sit inside the Web numbers.
Check citations directly, on a fixed query set. Search Console tells you that a link was shown, not which answer showed it or why. Run a stable set of prompts against AI Mode on a schedule and record which URLs are cited. A query that used to cite your page and now cites a page you have never competed with points to a change in fan-out or retrieval. A query where the same set of sources reshuffled in order looks more like ranking. Because AI answers vary between runs, compare several runs per query rather than one before and one after.
Use a holiday baseline. For seasonal businesses, compare the post-update week with the same calendar week last year as well as with the pre-rollout week. Neither comparison is perfect. Together they keep you from reading normal Black Friday behavior as an algorithm effect.
Hold off on content rewrites. A page that lost AI Mode citations because Hydrogen expands questions differently does not need a new tone. It may need an answer to a sub-question it never covered. A page that lost ranking in the core update may need a broader quality review, which Google's guidance describes as a deeper self-assessment after a large, sustained drop. Those are different fixes. Picking one before you know the cause wastes the work.
#What Crawlmind does with overlapping changes
Crawlmind's citation tracking records which URLs each engine cited on each run, so a change in source selection can be read against dated events rather than a single before-and-after snapshot. For this window, the useful markers are the core update start, the model switch, and the completion date once Google posts it. The goal is not to decide which cause is bigger. It is to avoid treating a retrieval change as a ranking problem, or the reverse.
#The short version
Two things changed in Google Search two days apart, and the standard recovery advice assumes one. Annotate both dates. Compare AI-feature impressions with the overall Web trend. Track cited URLs on fixed prompts. Use a holiday-aware baseline. Then diagnose.
Related field notes
November 21, 2026 · 5 min
Shallow pages lose in AI search first
Google's head of Search says AI answers leave little room for pages that only restate facts. How to find your shallow pages and decide what to do with each.
November 20, 2026 · 5 min
Google now writes your citation card blurb
AI Overview citation cards are testing AI-written page summaries. The text under your link may no longer be text from your page.
November 20, 2026 · 5 min
More links in AI answers are not more clicks
Google added more inline links to AI Mode and AI Overviews but published no click data. Here is how to tell whether the change reached your pages.
Share or discuss
New posts, no spam. Roughly monthly. Unsubscribe with one click.