Google's advice is to wait a full week after a core update finishes, then compare that week with the week before the rollout began. This year that advice sets a trap. According to SEJ's weekly roundup, the November 2026 core update started on Nov. 10 and can take up to two weeks. The first full week afterward probably covers U.S. Thanksgiving, Black Friday and Cyber Monday.
Put that week next to an ordinary early-November week and you will see a jump (or a dip, for a content site that loses weekend readers) that says almost nothing about rankings. A second confounder sits on top of it. Google made Gemini 4 Hydrogen the default model in AI Mode on Nov. 12, two days into the rollout, and says it also uses the model for some AI Overview questions. Here is how to build an audit that holds up when someone asks how you reached your conclusion.
Step 1: Start a change log before you open any report
Write the dates down first. Memory gets fuzzy by mid-December. Keep the log in a shared sheet or your reporting tool, and give the two Google events separate entries. SEJ's roundup says the rollout began at 10:06 a.m. PT on Nov. 10 and was still in progress on Nov. 12.
| Date | Event | Source | Notes | | --- | --- | --- | --- | | Nov. 10, 2026 | Core update rollout starts | Google Search Status Dashboard | Third core update of the year | | (fill in) | Core update rollout completes | Google Search Status Dashboard | Do not guess; wait for the dashboard | | Nov. 12, 2026 | Gemini 4 Hydrogen becomes AI Mode default | Google Search announcement | Also used for some AI Overview questions | | Nov. 26, 27, 30 | Thanksgiving, Black Friday, Cyber Monday (U.S.) | Calendar | Seasonal demand | | (your dates) | Your own changes | Your team | Sales start, email sends, price changes, template edits, migrations, outages, out-of-stock events |
The last row matters most. If you launched a sitewide promotion on Nov. 20, get that on record before you credit or blame Google for anything.
Step 2: Pick weeks that match in weekday and in season
The update started on a Tuesday, so use Tuesday-to-Monday weeks. Every window then has the same shape. Pull these four in the Search Console Performance report, using the date compare option and a custom range.
- Baseline week this year: Tue Nov. 3 to Mon Nov. 9, 2026.
- Post-rollout week this year: the first full Tuesday-to-Monday week after Google marks the rollout complete. If it finishes by Nov. 23, that is Nov. 24 to Nov. 30.
- Baseline week last year: Nov. 4 to Nov. 10, 2025. That is exactly 364 days earlier, so weekdays line up.
- Post-rollout week last year: Nov. 25 to Dec. 1, 2025. The 364-day shift puts Thanksgiving 2025 (Nov. 27) in the same slot as Thanksgiving 2026 (Nov. 26).
Search Console keeps 16 months of data, so last year's windows are still there. Add a fifth window as a cleanup check: Dec. 1 to Dec. 7, 2026 against Dec. 2 to Dec. 8, 2025. If the rollout runs long, shift every post window forward and keep the 364-day offset.
Set the search type to Web and filter to the country that drives your holiday sales. A U.S. holiday spike won't show up the same way in UK or Australian traffic, and mixing them blurs both signals.
Step 3: Split branded from non-branded queries
Branded demand follows your own marketing, email, ads and word of mouth. It tracks the holiday calendar closely and tells you little about a core update. Ranking changes show up in non-branded demand.
In the Query filter, choose Custom (regex) and enter your brand name, common misspellings and product-line names, separated by pipes. Run every window twice: once with that filter and once with the "Doesn't match regex" option. Search Console hides anonymized queries from query-filtered views, so the two segments won't add up to your unfiltered total. Record the unfiltered total as its own line. This pairs well with how to evaluate aeo research before changing your seo explained.
Step 4: Group pages the way your business works
Create page groups with the Page filter, using URL patterns. An ecommerce site might use category pages, product pages and buying guides. A publisher might use evergreen guides, news and tools. You will reuse these groups when you verify the result.
Note which groups are most exposed to AI features. Informational guides usually sit closer to what AI Overviews and AI Mode answer than product pages do. That is an inference, not a measured fact for your site, but it gives you a testable split.
Step 5: Run the naive and the adjusted comparison
The numbers below are hypothetical and illustrative only. They come from no real site. The adjusted view asks a different question. Instead of "did traffic rise?" it asks "did traffic rise as much as it did in the same weeks last year?"
| Segment (weekly clicks, hypothetical) | This year, baseline | This year, post week | Naive change | Last year, baseline | Last year, post week | Post week vs. seasonal expectation | | --- | --- | --- | --- | --- | --- | --- | | Branded | 4,000 | 6,000 | +50% | 3,600 | 5,400 | 0% (expected 6,000) | | Non-branded | 6,000 | 7,000 | +17% | 5,400 | 8,100 | -22% (expected 9,000) | | Total | 10,000 | 13,000 | +30% | 9,000 | 13,500 | -13% (expected 15,000) |
The naive comparison says clicks rose 30%, so the update must have been kind to you. The adjusted one tells a different story. Last year the same two weeks produced a 50% lift in both segments. Apply that to this year's baseline and you would expect 15,000 clicks, not 13,000.
The branded line matches expectation exactly, so the shortfall sits entirely in non-branded queries. Year over year, non-branded clicks fell about 14% in the post week (7,000 against 8,100), while the baseline week was up 11%. That pattern is worth investigating. It is not proof, and Step 7 explains why.
Step 6: Account for the AI Mode model change
Search Console has no clean AI Mode filter. SEJ's roundup notes that Search Console counts clicks on AI Overview and AI Mode inside the overall total, so you can't subtract them out. Google also said Hydrogen connects users to a broader range of links, but its announcement gave no link counts. Don't assume that helps or hurts you. This pairs well with ai performance review questions for seo teams that work in depth.
You can still look at the daily shape. Chart non-branded clicks by day from Nov. 3 and mark Nov. 10 and Nov. 12. A step change that starts on Nov. 10 and keeps going points toward the core update. One that begins on Nov. 12 deserves a second look. Since the core update is still rolling out and Hydrogen is available everywhere in Search, treat this as weak evidence.
Your page groups carry more weight. If informational pages drop while product pages hold, AI features are a plausible part of the story.
If you claimed a ChatGPT Business Hub profile, its Insights show mentions and clicks for the past 28 days. SEJ's roundup says a recorded click is not a confirmed visit, so compare it with the referrals in your own analytics before you rely on it.
Step 7: Verify the conclusion with cross-checks
A single table can mislead you. Before you report anything, run these checks.
- Page-group check: Does the non-branded shortfall appear across several page groups, or in one section you changed on Nov. 18? One group points to your own change or a local issue. Several groups point to something broader.
- Query-group check: Split non-branded queries into informational, commercial and navigational-adjacent terms. A drop concentrated in question-style queries suggests AI feature overlap. A drop in commercial terms suggests ranking movement.
- Cleanup-week check: Repeat the adjusted comparison for Dec. 1 to Dec. 7. A real shift tends to persist once the holiday rush ends. A gap that closes was probably timing.
- Source check: Compare Search Console clicks with organic sessions and conversions in your analytics. If they diverge, suspect tracking, consent or attribution changes before rankings.
- Own-change check: Scan your log for site releases, stockouts, price changes, indexing problems and outages in the same windows.
Also read: related topic: how to audit brand presence on reddit: a worked example
A conclusion is defensible when the non-branded gap shows up in several groups, survives the cleanup week, and doesn't line up with one of your own changes. Even then, write "likely affected" rather than "caused by."
What this method cannot tell you
No method cleanly isolates a core update from seasonality or AI feature changes. Last year's lift is only a proxy for this year's, and your own promotions may have shifted. Last year's windows may also have had their own search changes, which you would then bake into your baseline. Consumer demand can change year over year too, and a 364-day offset can't correct for that.
There is an industry-level problem as well. SEJ's roundup reports that Ziff Davis CEO Vivek Shah said organic Google referrals to its properties fell about 37% year over year in the third quarter, and that the company is "not planning on a search recovery." If your whole sector is shrinking, a year-over-year dip may be structural rather than update-driven. Segment-level year-over-year numbers can't tell those two apart.
When to act and when to keep waiting
Act quickly on anything you would fix regardless of the update: broken templates, accidental noindex tags, slow pages, indexing errors, promotions that didn't launch. Those are your own problems, and holiday traffic makes them expensive.
Hold off on rewriting content, pruning pages or changing site architecture based on one holiday week. Wait until the rollout is marked complete, the cleanup week is in, and the page-group and query-group checks agree. If they still do, plan changes as a measured test with a log entry, so you can see what moved.
What to expect next
Google's Search Status Dashboard will show when the rollout ends, and that date sets your post-rollout window. The cleaner read should come in the first or second week of December, once holiday demand settles. Keep the change log open until then and add every site change to it, however small.



