Google volatility (20–26 Oct): what to do this week

There is no confirmed Google core update between 20 and 26 October 2026. What you do have is ranking movement, a lot of forum noise, and trackers that are lying more than usual this season. If you run SEO in Spain or Catalonia and saw swings on 20–24 October, skip the “did the core ship?” loop. Separate four things: what sits on the official panel, what is industry interpretation, which part of the tracker signal may be a poisoned SERP, and what you check in Search Console this week without reopening the whole evergreen drop guide.

The update map lives in Google algorithm updates in 2026. The drop method lives in diagnosing organic traffic drops. This piece is only the 20–26 October briefing.

Confirmed versus noise

Confirmed (official source). The Ranking history on the Google Search Status Dashboard, checked on 26 October 2026, lists no ranking incident in October. The most recent one in 2026 is 24 September, lasting 13 days and 16 hours (Ranking history). The September 2026 spam update started on 24 September at 09:15 US/Pacific and ended on 8 October at 01:00 US/Pacific. At 01:37 PDT Google wrote that the rollout was complete on 8 October. It applies globally and to every language (official incident). If it hit you, recovery sits under the spam updates docs and the spam policies. Search Central is clear: fixing the issue can help, but systems may take months before they trust the site again. That months-long lag is why a flat week after 8 October does not close the case by itself.

Not confirmed (industry). Search Engine Roundtable logged chatter and sensor spikes on 16 October, 20–21 October and again on 24 October. Barry Schwartz wrote on 25 October that, given the mix of tools and comments, he believes there was some ranking adjustment on Friday and Saturday. That is his reading. It is not a Search Central announcement and not a new row on the dashboard.

Real gains and losses are happening. Broken sensors are happening too. Often both at once. Do not open a “core recovery plan” this week only because Mozcast or Semrush Sensor went red on a Saturday.

Timeline: September spam versus the 20–26 week

Search Engine Roundtable put the spam update close at 4:37 am ET on 8 October (8 October coverage), consistent with the official 01:00 US/Pacific mark. Do not blend that cycle (24 Sep–8 Oct) with the 20–24 October noise.

If your drop started in late September and has not reversed, the documented explanation is spam or quality, not an October core the panel never published. Roundtable described several volatility waves during the rollout (late September and early October); that fits a two-week spam deployment, not an “invisible core” this week. If the movement is fresh from 20–24, work from Search Console and human spot-checks before you name it.

One useful nuance for leadership reports: the spam update has finished rolling out. That does not mean your traffic is back. Google says that after you fix issues, automated systems can need months to reassess the site. Confusing “rollout complete” with “my curve should already be flat” is a common mistake this week. If someone asks “when do we recover?”, answer with the spam-policies timeline and your own remediation date, not with a made-up core calendar.

Why rank trackers are lying this season

Google has hardened scraper blocking. Barry Schwartz summarised it on 12 October: broad IP blocking, more captchas and injection of incorrect data into suspicious SERPs (Third-Party Google Rank Tracking Struggling). Derek Perkins (Nozzle) has documented the third point best.

The Poisoned Google SERPs study, in the 21 October 2026 update we checked, counts a SERP as poisoned when page 1 or page 2 is poisoned. Figures from that dated version:

• 22.6% of SERPs judged on both pages were poisoned since 13 September, against a 0.28% baseline before.
• Peak of 43.5% on 30 September.
• It eased to 4–6% in mid-October, then climbed again: 25.9% on 19 October and 35.7% on 21 October.
• On poisoned SERPs, the tracked site is missing from the top 10 in 60.8% of cases, against 2.7% at baseline.
• In the country table, Spain shows 4.8% poisoned SERPs since 13 September (808,086 SERPs judged), below the US (8.8%) or Canada (11.0%) (Nozzle country table). 4.8% is still plenty of noise for a weekly report.

Those numbers come from Nozzle’s 21 October update.

The page looks normal. The results are not. Social networks and forums fill slots that used to hold vendors or local listings. A rank report can show a crash that a user in Madrid or Barcelona never sees. That is the practical risk for an ES client call: you walk in with a red chart and the live SERP on their phone does not match. Ahrefs, in a statement Roundtable covered on 22 October, talked about 5% to 20% of results affected depending on the provider; Semrush and others said they had added filters. None of those statements turns the sensor into a primary source ahead of Search Console.

Protocol for this week:

1. Do not treat a one-day tracker drop as a verdict.
2. Cross-check 10–20 critical queries by hand (incognito and signed-in, mobile, ES location). Prefer the same device class your money queries actually get.
3. Read clicks, impressions and average position in Search Console with a Spain filter. Export the compare view so the debate later is about numbers, not memory.
4. If the tracker and GSC diverge, trust GSC and the spot-check first. Log the discrepancy; do not “average” the two signals into a fake consensus.

If the underlying problem on your site is spam or scaled content, poisoned trackers do not clear you. Go back to the policies and to site reputation abuse and scaled content.

How to read Search Console this week

Google publishes the method in Debugging drops in Google Search traffic (official guide). Our operational guide is diagnosing organic traffic drops. We are not rewriting it. Only the cut for 20–26 October.

Time anchor. Compare 20–26 Oct with the previous week and with the same window in 2025. Filter Spain. Separate Web, Images, Video and Discover. A drop “since the spam” (late September) is not the same hypothesis as a hit on Friday 24. In the Performance report, use compare mode: last 7 days versus the previous 7, and a year-over-year cut if your sector is seasonal (travel, education, autumn retail). Write the inflection date on the ticket before you name a cause.

Country and type. Filter Spain. Separate Web, Images, Video and Discover. Roundtable has been carrying complaints about irregular Discover and old pieces in News for days; do not mix that channel with classic web ranking when you brief a client. A Discover dip can look like a “site-wide crash” if you leave the search-type filter on All.

Shape of the chart. Impressions and clicks fall together: look at algorithm, technical, spam, security or demand. Impressions stable and clicks down: titles, snippets, rich results or more SERP competition, including AI inside Google. A small position drop (from 2 to 4) can move CTR a lot without the site “dying”. That pattern shows up often after layout changes in the SERP, not only after a named update.

Manual actions and security. Open the manual actions report and security issues before you stand up a core war room. If there is a manual action, the story changes completely: you stop talking about volatility and move to documented remediation. Skipping this step wastes a day of speculation.

Generative AI. If you watch AI Overviews / AI Mode impressions, the report exists; clicks still sit mixed inside Web. Detail in Search generative AI performance reports and last week’s link context in AI Overviews and AI Mode links (October 2026). Do not invent an AI-only click column that Search Console still does not offer.

If you cannot point to the inflection date in GSC Spain, you do not yet have an “update story”. You have a tracker anecdote.

Pichai (22 Oct): agentic layer, not a ranking factor

On 22 October Search Engine Journal published Roger Montti’s reading of Sundar Pichai’s interview with Alex Heath (SEJ, 22 Oct 2026). Pichai describes a product vision: an agentic layer that unifies chat, tasks and Search (“one box”), with cloud agents that keep context across surfaces. He talks about making agentic capabilities more available in Search “in a pretty deep way” over the coming months.

None of that is a core update, a ranking change, or a KPI you can put in Friday’s standup.

For a Spain team it is roadmap context. If you already watch performance against agents or Lighthouse in agentic scenarios, keep the useful thread in agentic navigation, PageSpeed and Lighthouse. Do not open a ticket that claims “the algorithm changed on the 22nd”.

AI search investment: Haynes, SEO Works and Indig

The same week produced three pieces Roundtable grouped on 22 October (AI search studies). They do not say measuring LLM citations is pointless. They say volume and conversions from external assistants are still small next to Google organic in the published samples.

Marie Haynes (21 Oct). She analysed 12 months of GA4 on 100 sites, with more than 145 million sessions. ChatGPT sessions were 0.62% of Google organic volume; all LLMs combined, 0.73%. On the 55 sites with conversions, ChatGPT converted at 13.8% versus 3.02% for Google organic, but Google held 96.5% of total conversions in the analysis and Google organic generated 27× more conversions than all LLMs combined (GEO is a distraction). Her reading: the search AI that moves business today is still mostly Google’s (AI Overviews, AI Mode, ML ranking), not the ChatGPT referrer.

The SEO Works (benchmark, 95 sites, Sep 2025–Aug 2026). Organic delivered 39× more traffic than AI assistants. For every 100 organic conversions, AI referrals delivered 3.2. AI conversion was slightly better in aggregate (7.18% versus 6.08%), but worse than organic on 23 of 63 sites with enough sample (AI Referral Traffic Benchmark). In their arithmetic, a +5% organic lift added more conversions per month than doubling AI traffic. Watch the scope: the study measures assistant referrals (ChatGPT, Gemini, Copilot and so on), not AI Overviews clicks, which arrive in GA4 as organic.

Kevin Indig (19 Oct). Across 12 properties he could not reliably detect hidden AI Direct traffic into the long tail; engaged Direct stayed flat while total Direct inflated with bots (Does AI drive more direct traffic?). Showy correlations between AI visibility and Direct did not survive spurious-correlation controls.

How to use this week. Measure LLM referrals and citations if the channel matters to you. Keep the sample sizes in view: Haynes at 100 sites, SEO Works at 95, Indig at 12. None of them is a census of Spain. Do not reallocate Spain SEO budget because a “GEO” slide asks for it. If you want the measurement frame, it is in measure AI visibility in SEO. If the debate is scaled generated-content spam, go back to the policies and the site-reputation post.

GBP: Business Home, government ID and reports

Google Business Home (22 Oct). Roundtable describes a beta portal that joins Business Profile, Merchant Center, YouTube and Ads. Official help is blunt: Business Home is currently a closed beta experience, available only to select business categories in the US and the UK (About the Business Home beta). It is not a documented rollout for Spain. Do not open a local-process migration to Business Home this week. Roundtable also covered the announcement on 22 October.

Government-ID verification (20 Oct). Roundtable reported a test seen on Local Search Forum: upload a photo of a government ID (20 October coverage). The official verification help, checked for this briefing, still lists phone or SMS, email, live video call and postcard (Verify your business on Google). It does not include government ID. Treat it as a test, not as a procedure for Spain.

Advanced performance reports (23 Oct). There is a test of richer reports (engagement, reviews, trends, audience) covered by Roundtable (performance reports). If it appears for you, look at it. If it does not, that is not an incident. Steady local work still sits in advanced Google Business Profile.

Limitations and exceptions

• A missing row on the dashboard does not prove Google changed nothing. It only means there is no published ranking incident with that name on the check date.
• Schwartz’s view of a 24 October adjustment is interpretation, not confirmation.
• Nozzle figures measure SERPs served to its tracking infrastructure, not every real user’s SERP in Spain.
• Haynes, SEO Works and Indig use different samples (100, 95 and 12 sites). They are not a census of the Spanish market.
• Business Home and the ID test are not documented as available in Spain in the open sources for this week.
• This briefing is not individualised professional advice. It does not promise ranking recovery, AI citations or traffic.

What to do and what to ignore

Do this week

1. Read the official spam update incident and the Ranking history with the team (10 minutes).
2. In GSC, filter Spain; compare 20–26 Oct with the previous week and with last year; export top pages and queries with the largest click deltas.
3. Open manual actions, security and indexing before you talk about a “core”.
4. Spot-check 15 money and brand queries on mobile ES; note where the tracker disagrees.
5. If the drop starts in the 24 Sep–8 Oct window, review spam policies and quality, not an October core that is absent from the panel.
6. Keep Discover and News separate from classic Web in the client or leadership story.
7. Mark Business Home and government-ID verification as out of scope for ES unless new help-page evidence appears.

Ignore or park

• Declaring an “October core update” in a client report without a dashboard row.
• Rewriting the whole site because a volatility sensor went red on a Saturday.
• Migrating GBP processes to Business Home in Spain this week.
• Cutting organic SEO budget only because ChatGPT converts “better” in someone else’s sample: look at the volume.
• Mixing Pichai’s agentic-layer announcement with a 20–24 ranking change.

Concrete next step

Pick ten URLs that concentrate clicks in Spain. For each one, note in a table: clicks and impressions 13–19 Oct versus 20–26 Oct, average position, whether the tracker shows a crash GSC does not confirm, and whether the URL came onto the radar during the September spam update. Add one column for search type if Discover or Images muddy the Web story. In about 90 minutes you have a prioritised diagnosis. Fix this week only what the table flags (indexing, soft 404, obvious cannibalisation, misleading title, page touched by spam). After 14 days, repeat the cut. Do not use “wait for the core” as a plan.

If the inventory is large and you need to prioritise by margin or content risk, SEO consulting scopes the backlog without turning every Roundtable headline into a sprint. Drop detail still lives in the diagnosis guide; the update map, in 2026 updates. Keep this briefing as the weekly filter, not as a substitute for those two evergreen pieces.