Market analysis

· 35 min

Google, first half of November 2026: a core update the trackers struggle to see

Google started its November 2026 core update on 10 November and, on 18 November, it was still rolling out. In the same weeks Google made its results harder to watch from outside, put a new model behind AI Mode, and told sites to let its own agent through.

Topic radarSelect a topic to filter the archive. The radar shows how this issue's words divide across the topics.

Signal readout

Topic mix

  • AI: 28%
  • Google & Search: 48%
  • PPC: 7%
  • Analytics: 15%
  • Brands: 2%
  • AI28%
  • Google & Search48%
  • PPC7%
  • Analytics15%
  • Brands2%

Share of the issue's words about each topic, counted section by section.

Why it matters

Read this core update from your own side. Note 10 November as confirmed and 14 November as reported; do not read AI impressions on 9–11 November as a ranking signal; and compare clicks and impressions for the same landing pages before the update and after the dashboard shows it complete, not on a day in between.

The short answer

What
A reading of Google Search and its rivals in the first half of November 2026: the November 2026 core update, scaled content and SEO poisoning, how hard Google became to observe from outside, Gemini 4 Hydrogen and the changing AI Overview, what AI Mode does to clicks, Search Console, crawlers, agents and CDNs, share and click data, courts and EU rules, and Bing.
Why
Because the first core update since June began while the instruments used to watch it were being shut out, and because Google, in the same days, published more about who may fetch a site — its crawler, its agent, its CDN advice — than about how it ranks one.
Who
SEO practitioners and site owners reading a core update, publishers, technical SEOs who run CDNs, firewalls and bot rules, and anyone who reports search performance.
Where
Global, with the UK, France, the European Economic Area and US courts treated separately where the sources do.
When
1–18 November 2026, with the last days of October where events began. The core update was still rolling out when this was written, on 18 November.
How we know
From the Google Search Status Dashboard, Google Search Central and Google's product blogs, the studies' own pages (SE Ranking, Seonix, SenseGap, Tame the Bots, SE Visible, SparkToro, Sistrix, Semrush), and Search Engine Roundtable, Search Engine Journal and PPC Land where no primary page could be opened — in which case the text says so. It builds on our October analysis and the three weekly digests of November. Every source is listed at the end.

In plain words

Google is in the middle of one of its big reshuffles of search results, which started on 10 November and can take two weeks. The tools people use to watch their Google rankings are struggling to see it, because Google is making it harder for robots to read its results. Google also put a new AI model behind its chat-style AI Mode, and a study found that people who used AI Mode clicked through to websites much less often. Meanwhile more robots visit websites on people's behalf, and Google asked site owners not to block its own one. The safest numbers right now are the ones from your own site and your own Search Console.

At the end of October, our analysis of the month said the page had changed more than the ranking, and that the instruments used to watch Google from outside could no longer be trusted. The first half of November tested both claims at once. On 10 November Google started its first core update since June, the kind of event the whole SEO industry is built to measure. It arrived in a fortnight when Google was slowing the links on its own results pages, testing a sign-in wall after the first search, and fighting scrapers in court, and when one well-known visibility index wrote "No usable data" against one of the update's first days.

The page kept changing too. Gemini 4 Hydrogen became the default model in AI Mode on 12 November, AI Mode stopped saying which model answers, and the AI Overview changed shape at least six times in sightings Google did not confirm. A preregistered field experiment, published as a preprint, put a number on what AI Mode does to clicks. And Google published more about access than about ranking: how to move to a CDN without dropping out of Search, how to tell Search traffic apart, and how to let its new Google-Agent through a firewall.

This analysis goes by theme: the core update, how far it can be seen from outside, spam and SEO poisoning, the AI page, what owners can see, crawlers and agents, clicks and share, courts and rules, and rivals. Then it checks October's watch list item by item, gives our reading, a note on our own crawler, and what to watch for the rest of November.

DateWhat happenedStatus
30 OctShared robots.txt token for AI training; Bing splits training from groundingAnnounced
2 NovBrand AI Overviews move to the right column on desktopSpotted, no comment
5 NovGoogle slows its /goto result links; scrapers report waitsReported, no announcement
5 NovAbout six million Facebook Answers pages indexedSpotted, no comment
9–11 NovLogging error cuts AI impressions in Search ConsoleConfirmed by Google
10 NovNovember 2026 core update startsConfirmed; still rolling out on 18 Nov
11 NovSearch Central on CDNs and crawlingGoogle documentation
12 NovGemini 4 Hydrogen the default model in AI ModeAnnounced
16 NovSearch Analytics API: branded, multimodal, filter listsAnnounced
16 NovAI Mode drops the model name; Carbon reaches AI ProConfirmed by Google
17 NovGoogle-Agent help documentGoogle documentation
17 NovSign-in demanded after the first searchSpotted; "may be testing"
17 NovFBI: Gootloader's SEO poisoning tied to over $28 millionFBI alert, via SER
The first half of November 2026 in Google Search: the events this analysis rests on.

The core update: started on 10 November, still rolling out

Google's Search Status Dashboard lists the November 2026 core update as starting on 10 November 2026 at 10:06 US/Pacific. Its first entry reads: "Released the November 2026 core update, which applies globally and to all languages," and "The rollout may take up to 2 weeks to complete." Google's post, as quoted by Search Engine Roundtable, called it "a regular update designed to better surface relevant, satisfying content for searchers from all types of sites." When we read the incident page on 18 November it had no completion entry: as of 18 November the update is still rolling out. SISTRIX, in a note last modified on 16 November, said the same: "It is currently rolling out and is expected to take up to two weeks."

It is the first core update since the one that ran from 21 May to 2 June, and the first named update of any kind since the September spam update ended on 8 October. PPC Land did the arithmetic on the calendar: two weeks from 10 November ends on 24 November, three days before Black Friday. Search Engine Roundtable counts 23 core updates with a median rollout of 14 days; the longest, March 2024, took 45.

Core update rollouts, days (Google Search Status Dashboard)

December 202518.1
March 202612.2
May 202611.9
Nov 2026, still open8.0
December 2025, 18 days 2 hours; March 2026, 12 days 4 hours; May 2026, 11 days 21 hours (via Search Engine Roundtable and PPC Land). November 2026: about eight days from 10 November to 18 November by our count, with no end entry yet.

What the update did is harder to say than usual. Some tools showed movement on 8 and 9 November, before the announcement, but Barry Schwartz wrote then that "it is hard to trust the tools since they seem a bit off." On 16 November he described chatter as quieter than for most core updates, which he partly put down to third-party trackers struggling, and named one day: "The tools, in aggregate, also seemed to show more volatility on Saturday, November 14th, than other days." Glenn Gabe, quoted there, wrote: "I believe the update landed on Friday the 13th." Forum reports ran both ways, from one site "more than 20% down" to another up roughly 10% over a weekend. SISTRIX's daily note for 12 November reads "No usable data"; on 16 November it was cross-checking large-domain candidates, with a possible rise for the Daily Mail in the UK and no figures.

Our reading: the start date is confirmed and the end date is not. Between them, the only dates anyone can defend are 10 November, from Google, and 14 November, from the aggregate of tools that are, by their own operators' account, partly blind. A core update reorders results over days; a report that names a winner or loser before the dashboard shows an end is reporting a moment of a rollout, not its result.

It is hard to trust the tools since they seem a bit off.

Barry Schwartz, Search Engine Roundtable, on ranking volatility before the announcement, 9 November 2026

Watching from outside: Google closes the windows

October's problem was poisoned results: trackers were served pages that looked normal and were wrong. In November Google added friction of other kinds. Search Engine Journal reported on 5 November that DataForSEO, Serper and SerpApi all saw Google limit how fast its /goto result links resolve: on a page with 18 links the first 5–6 open at once, "the 7th link may result in a 1 second wait", by the 15th "there could be a 40 second wait", and DataForSEO saw "up to 120 seconds for 1 URL". Serper first noticed it at 1:30 UTC on 30 October. None of them linked to a Google announcement.

On 17 November Brodie Clark of SerpAlerts reported being made to sign in before he could see any results, even in incognito, and said he had seen the screen for a few days; normally the prompt appears only after several queries. Google's Rajan Patel answered: "I don't believe it's meant to happen on the first page of results like this but we may be testing something here..." Earlier in the month AI Mode had shown some users a "Continue with this search?" screen warning that "Malicious queries could result in unsafe or harmful actions", which Schwartz read as a block on bots and scrapers. Google did not confirm that either.

The effects spread. Brave said on 11 November that Google's throttling "created a bot problem for Brave Search", with spikes of "several thousand requests per second", and began encrypting its own result links. In Google's lawsuit against the scraping service SerpApi, a judge dismissed the amended complaint on 10 November with leave to amend by 30 November; Search Engine Roundtable notes that "the judge did not rule that SerpApi's scraping practices are lawful."

The data that did come out was about the past. SE Ranking showed on 3 November that in September the share of 10,000 US keywords with mostly unfamiliar results went from 0.09% to 57.3% overnight and peaked at 67.2% on 24 September, and said results became unstable again at the end of October. Nozzle, our October benchmark, published no new figure in the period our sources cover. Search Console itself was not a perfect refuge: Schwartz noted on 16 November that its performance reporting "has been a bit laggy", and Google confirmed that "a logging error caused a decrease in impressions on the Generative AI performance report" on 9–11 November, adding that "this issue affects data logging only."

Spam, scale and SEO poisoning

Google's October documents drew the line around scaled pages, invented authors and pages users generate on someone else's domain. November produced the test case. On 5 November Search Engine Roundtable reported that Google had indexed about six million pages under facebook.com/fb-answers/, each of which, in Saul Westbrook's description, "consists of an AI-generated summary to explain a specific topic." Glenn Gabe, who says Facebook began filling the directory in July 2026, called it "another example of massive 'Scaled content abuse'". Google has not commented. The question it poses is the one October left open: whether the scaled content policy applies to a site of Facebook's size the way it applies to everyone else.

Another source of user-generated pages arrived on schedule: from 16 November every discoverable Discord server gets a public web page with AI-written channel summaries, and from mid-December some public messages may appear in web search, Discord said on 30 October.

The most serious item was not about rankings at all. An FBI FLASH alert dated 16 November, reported by Search Engine Roundtable on 17 November (we did not open the alert itself), describes Gootloader using SEO poisoning on "over 1000 infected WordPress sites and 216 domains controlled by the Gootloader administrator", which contributed to more than $28 million in ransomware payments as of October 2026. Bill Hartzer's comment: "It is not often the FBI writes the letters SEO into a cyber alert." For a site owner the lesson is plain: a hacked WordPress site can be used to rank pages that are not yours.

John Mueller's answers in the period were about quality more than spam. Of pages left out of the index he wrote, "Usually this is a sign that Google's quality systems are unconvinced about your site" (10–11 November), and "A site move doesn't improve the site's overall quality" (17 November). The same day he warned against answering with a 200 status when a site is down: the error text can appear in results, and identical error pages can be de-duplicated out of the index. In Europe a quieter change surfaced: a Search Console email, shared by Glenn Gabe and reported on 4 November, says that from 30 August past manual actions under the site reputation abuse policy "will no longer impact search results shown to users within the EEA."

The page: a new model, unnamed, and an AI Overview that will not sit still

On 12 November Robby Stein called Gemini 4 Hydrogen "the biggest model upgrade to Search yet": "Hydrogen is now available globally in Google Search for all users," in every language AI Mode supports. Google says it handles multi-part questions and typos better, runs a wider query fan-out, shows more images, tables and custom tools, and connects people "to a broader range of helpful links than before." Stein added on X, via Search Engine Journal, that Hydrogen is also "being used for your hardest questions in AI Overviews". Google gave no numbers on traffic or on how often answers link out. It follows Gemini 3.5 Flash as AI Mode's default in May, Gemini 4 Argon for paid subscribers on 13 October and Gemini 4 Carbon for Ultra subscribers on 10 November.

Then the label went. On 16 November Search Engine Roundtable showed the AI Mode header reading only "AI Mode", without the model's name, and Rajan Patel confirmed it: "Intended, but we're thinking through the right way to bring back an indicator for your selected model." Gemini 4 Carbon, which reached AI Mode for Google AI Ultra subscribers on 10 November, now reaches AI Pro subscribers too ("Not a bug."). Temporary threads, which do not use past context and are not saved to history, were spotted on 14 November; Search Engine Roundtable reports that Stein announced the global rollout on 17 November.

The AI Overview changed in public view and without announcements. On 2 November, brand searches on desktop showed it in the right-hand column where the knowledge panel used to be, with the brand's own site leading the results; PPC Land found Zillow's AI text giving a founding year of 2006 beside a fact row saying 2004. On 10 November a local pack inside an AI Overview lost its buttons, photos and reviews. On 11 November definitions appeared fully expanded and, in a test, source icons moved below the text. On 13 November AI-generated images were back for some users, and Glenn Gabe saw a results page without the AI Mode tab. And on 17 November Brodie Clark saw desktop AI Overviews show three citation cards on the right, unfaded, by default — a reversal of an earlier removal. None of these was confirmed by Google.

AI Mode also became a place to transact: direct booking links for US home services and direct checkout in Canada on 4 November, and on 16 November an Ask Maps that can "call local businesses for you". And ads moved deeper: on 10 November ads in AI Mode reached ten more countries and multiturn ads, which "show up on responses to follow-up user queries", went live in the US; on 12 November Gabe captured a single answer with "Four different sponsored placements", in Juozas Kaziukėnas's count. Brazil, Mexico and the UK are still due this quarter.

What AI Mode does to clicks: the first randomised number

Until now the evidence on AI Mode and clicks has been observational. On 17 November PPC Land wrote up a preprint by Wang, Gleason, Bart, Wilson and Metaxa (University of Pennsylvania and Northeastern) that ran a preregistered randomised field experiment through a browser extension with about 1,100 US users recruited on 17–19 March 2026. After a three-day baseline, each user spent seven days in one of three arms: AI Overviews hidden, Google as it is, or every search redirected to AI Mode. We read PPC Land's account, not the paper.

Against ordinary Search, the AI Mode arm clicked through to external sites 18.8 points less often (95% interval −22.2 to −15.3), or 14.9 points less if each conversation counts as one search, about 2.27 fewer external clicks a day. The share of users who clicked an ad fell 42.7 points. Use of Bing, DuckDuckGo or Yahoo rose 11.2 points, and trust and satisfaction fell. Hiding AI Overviews showed no significant effect (+2.7 points, p = 0.124), but the filter that hid them broke during the study, so that arm says little.

AI Mode against ordinary Google Search: fall, percentage points (field experiment, preprint)

Users clicking an ad42.7
Users clicking Reddit21.2
Clicks to other sites18.8
Users clicking news12.5
Users clicking Wiki9.9
Wang, Gleason, Bart, Wilson and Metaxa, arXiv 2608.18352, unreviewed preprint, via PPC Land, 17 November 2026. About 1,100 US users in three arms (386, 356, 358). "Clicks to other sites" is the click-through rate to external sites; "Wiki" is Wikipedia.

Try it on your own page

See what Vectorscope finds on a page of yours

Everything above is measurable on your own work, and the check takes about a minute. Paste one address and read what comes back.

Open Vectorscope

Free · no account · one at a time

The caveats matter. It is a preprint, not peer reviewed; users were pushed into AI Mode rather than choosing it — at baseline AI Mode was used in 0.6% of their searches, and AI Overviews appeared in 36%; and PPC Land notes that Google added shopping ads to AI Mode in February 2026, which the paper does not address. But it is the first estimate in which the only thing that differs between groups is the surface, and it points the same way as the publishers' own numbers below.

What site owners can see

Search Console did not gain what October's watch list asked for. On 2 November Gary Illyes answered talk of clicks and queries in the AI report: "The team is super interested in giving more data, data that makes sense, but we don't have anything to announce yet." The report still counts impressions; AI clicks stay mixed into the ordinary Performance report. Search Engine Journal points to an outside deadline: the UK competition regulator's requirements for Google, which include clicks for AI Overviews and AI Mode, apply from 3 December.

What it did gain was in the API. On 16 November Google brought three things already in the interface to the Search Analytics API, "following ecosystem requests to make this data available for custom reporting and automated workflows": a branded classification that splits branded from non-branded queries ("this data is only available for eligible properties"), a webMultimodal type with performance from Google Lens, Circle to Search and image uploads, and filter lists for several countries or devices in one request. On 12 November a new help page listed the referrers Search traffic arrives with; the header carries "only the scheme and host", AI Overviews and AI Mode are not mentioned, so they cannot be told apart from ordinary results, and Google says the method "isn't exact."

Two sites showed what the existing data can do. Ziff Davis said Search Console showed its pages appearing in AI Overviews and AI Mode about 2.7 billion times in the third quarter, while its organic Google referrals fell about 37%. Cyrus Shepard switched his site out of Google's generative AI features for September and estimated, in Search Engine Journal's write-up of 9 November, that they had sent about 21% of its clicks, at a click-through rate of about 0.9% against 0.4% for the site as a whole — "more as a directional yardstick than Google gospel".

Who may fetch your site: crawlers, agents and CDNs

This was the half-month's biggest theme in Google's own documentation. It began on 30 October, when Google said it will support a shared robots.txt token that lets a site "set broad preferences across participating AI providers" — User-Agent: __usage_ai-training with Disallow: / — and aims to do so in 2027. Bing published the same day and went further, splitting AI training from AI grounding: training can be refused in robots.txt or on the page, grounding only on the page, and from 1 March 2027 NoArchive and NoCache stop counting as either. Bing's own line: "Neither choice guarantees inclusion, traffic, or payment."

On 11 November Cherry Prommawin and Gary Illyes wrote on the Search Central blog about moving to a CDN. A faster CDN can raise Google's crawl rate — "Fast and stable responses speed up crawling" — but a new one starts cold, because "Caches only shield you on the second visit"; if the origin needs relief, answer 429 or 503 for hours, a day or two at most. The bigger risk is a blanket challenge or CAPTCHA that stops Google's verified crawlers. PPC Land added from Google's video that then "products disappear from Google Shopping and ads products stop working."

Your pages, shopping listings, and your chicken pictures will begin to drop out of Search.

Google Search Central blog, on CDNs that challenge Google's crawlers, 11 November 2026

Six days later Google asked for the opposite kind of exception. A help document published by 17 November describes Google-Agent, "the name used by agents hosted on Google infrastructure to navigate the web and perform actions upon user request." Its requests carry Google-Agent in the user agent, come from the ranges in user-triggered-agents.json, and are being tried with Web Bot Auth signatures. Google asks sites to allow it explicitly in CDNs, firewalls and bot-detection services, and PPC Land lists the document's six troubleshooting checks, from rule priority over ASN blocks to proxies that strip signature headers — including the fact that "concurrent requests from multiple users share the same egress pool." The document does not mention robots.txt; PPC Land notes that Google's list of user-triggered fetchers, which includes Google-Agent, says they "generally ignore robots.txt rules." OpenAI's commerce playbook for ChatGPT Wallet, written up by PPC Land on 14 November, asks merchants the same thing for its agents: "websites need to ensure agents are not inadvertently blocked".

Meanwhile, what is actually knocking became clearer. Dave Smart of Tame the Bots found on 3 November that Gemini fetches a page someone asks about with the user agent Google, and AI Mode with Google-Lens, and that neither renders: they fetch "only the main document, and no resources like images, scripts, or stylesheets." SenseGap showed on 5 November that Google's reverse DNS instructions accept any googleusercontent.com hostname, which every Google Cloud machine gets by default; its own site received 89 requests in eight days from such machines posing as Googlebot. Across four sites, every checkable request claiming to be CCBot, DuckAssistBot or Perplexity-User was fake, and so were 32% of the GPTBot ones. HUMAN Security counted the agents: in October Meta's Muse sent 59% of the AI agent requests it saw, agent traffic was about eight times August's, and 1.9% of requests went to checkout.

NameWhat it isWhat to check
GooglebotGoogle Search's crawlerHostname ending in googlebot.com and Google's published IP ranges, not any googleusercontent.com host (SenseGap, 5 Nov)
Google-AgentAgents on Google infrastructure acting for a userRanges in user-triggered-agents.json; allow in CDN and firewall; robots.txt not mentioned (Google, 17 Nov)
GoogleGemini fetching a page a user asked aboutMain document only, no scripts; short user agent some firewalls block (Tame the Bots, 3 Nov)
Google-LensAI Mode fetching a pageMain document only; text that needs JavaScript is not seen (Tame the Bots, 3 Nov)
__usage_ai-trainingShared robots.txt token, not a crawlerSays what a page may be used for; Google aims to support it in 2027 (30 Oct)
ai-grounding:disallowBing meta tag or X-Robots-TagKeeps a page out of Copilot's grounding; page level only (Bing, 30 Oct)
Who fetches a site in the name of Google and Bing, as documented or measured between 30 October and 17 November 2026.

How sites answer is uneven. Seonix read robots.txt on 14,318 sites on 3 November: 12.2% block at least one AI training bot by name, while bots that fetch pages to answer a question are blocked about ten times less often. Our own study of 190 SEO sites, published on 4 November, asked the servers as well and found 19 of 147 refusing an AI user agent their robots.txt allows. And United Airlines put its rule where it can be enforced: since 11 November its terms say "No Automated Agent shall be permitted to access United's site or app without prior written permission from United", while its robots.txt, on 14 November, still named no AI crawler.

Sites blocking an AI bot by name in robots.txt, % (Seonix, 14,318 sites)

GPTBot (training)8.0
ClaudeBot (training)7.4
Google-Extended6.4
ChatGPT-User (answers)1.4
OAI-SearchBot0.7
Seonix, crawled 3 November 2026. Google-Extended is a training control; OAI-SearchBot fetches for answers. The study reads robots.txt only, not what a firewall or CDN does.

Clicks and share: the questions move, the visits do not follow

The half-month produced two kinds of number, and they disagree in an instructive way. On demand, OpenAI's Adam Fry said in an interview written up on 3–4 November that "ChatGPT's volume of search-like queries is now roughly one-fifth the volume of queries of Google", without naming which Google figure. On visits, Ahrefs' tracker put Google at 24.74% of tracked traffic to 118,671 sites in October and ChatGPT at 0.36%. Graphite and Similarweb found that in the US in the second quarter organic search sent 49 times as many visits as AI assistants. SE Visible found that 79.2% of sites got at least some AI referral traffic in September 2026; among sites with at least 1,000 organic visits a month, a typical one got 2–5 AI visits per 100 organic ones. And SparkToro, on 17 November, read 118,163 ChatGPT prompts from 12,321 US panel users: only 6.9% showed research toward a brand to buy from. Rand Fishkin's conclusion was that AI visibility matters a great deal, "But NOT BECAUSE OF CHATGPT."

MeasureFigureWho measured, when
ChatGPT search-like queries against Google's"roughly one-fifth"OpenAI's Adam Fry, via SEJ and SER, 3–4 Nov
Share of tracked traffic, OctoberGoogle 24.74%, ChatGPT 0.36%Ahrefs, 118,671 sites, via SEJ, 3 Nov
Sites with any AI referral traffic, September79.2%SE Visible, 28,500 sites, 12 Nov
Organic search visits per AI assistant visit, US Q249Graphite and Similarweb, via SEJ, 9 Nov
People Inc core web sessions, Q3−26% (revenue −6%)Company results, via PPC Land, 6 Nov
Ziff Davis organic Google referrals, Q3about −37%Company results, via PPC Land, 7 Nov
Reddit search referrals, Octobermore than 1 million fewer a weekShareholder letter, via SEJ, 30 Oct
Browser visits to the top million sites, Q3 2026 vs Q1 2022−12.2%; −29.2% per userSparkToro on Similarweb, 10 Nov
ChatGPT prompts showing purchase research6.9%SparkToro, 12,321 US panel users, 17 Nov
Different methods, panels and periods: read each row on its own terms, not as one series.

The publishers' numbers are the hardest. People Inc's chief executive Neil Vogel said "We've lost nearly 70% of our Google search traffic over the last few years"; its core sessions fell 26% year on year in the third quarter while revenue fell 6%. Ziff Davis's organic Google referrals fell about 37% in the quarter, after about 23% a year earlier. Reddit told shareholders that "search traffic has dropped by more than 1 million referrals on average per week" in October, and its chief executive said "Google is crawling us as hard as ever". In France, where AI Overviews launched on 22 July, one site in Glenn Gabe's cases fell "from about 750 per day to about 400 per day".

Look, I think our position is we're not planning on a search recovery.

Vivek Shah, chief executive of Ziff Davis, on the third-quarter call, as quoted by PPC Land, 7 November 2026

Coverage kept rising where it had been lowest. Sistrix found AI Overviews on more than 80% of the million keywords in its UK index on 11 November, with the newest growth on ecommerce — Go Outdoors from under 2% of its keywords to 87.4% — and the German rate "has just risen over 50%", where SE Ranking had measured 40.4% on German brand keywords in October, a different panel. And which pages the Overview cites still follows ranking more than not: Semrush and performance.io, on 16 November, found across 14,300 US pharma keywords that 83.8% of AI Overview citations came from top-ten organic domains and the number-one domain was cited 68.8% of the time — an association at domain level, with no clicks measured.

Courts and rules

The cases of the half-month were about answering with publishers' pages rather than about ranking them. On 2 November The New York Times and three Tribune papers sued Brave in New York over its AI answers and its ad blocking; PPC Land reports that in the Times's test Brave's LLM Context endpoint returned 1,291 of the 1,295 words of one story. On 29 October the Wikimedia Foundation asked the court in the OpenAI case, in a proposed brief, to hold that copying sources into answers is not fair use, while calling training "likely fair use". Both draw the line Bing drew on 30 October: training is one question, answering with a page another.

Google's own litigation went the other way: its case against SerpApi was dismissed on 10 November, with leave to amend by 30 November. In Europe, BrightEdge found, in Search Engine Roundtable's write-up of 6 November, that in the product grid the DMA opened to comparison sites Google Shopping "appears 16 to 35 times as often as the next-largest service" in Spain, Italy, France and Germany. The UK requirement for AI click data applies from 3 December. In the main US search case, the D.C. Circuit argument October expected "as soon as 9 December" is still ahead, and none of our sources for 1–18 November reports a step by Penske or Chegg or the CMA's decision on a choice screen.

Rivals: Bing offers the off switch, ChatGPT reports its own numbers

Bing spent the half-month giving people ways out of AI answers. On 11 November it let users turn them off with a browser extension, which also makes Bing the default search engine, or with the -ai operator — "We believe Bing users should have a choice in how they search" — and on 16 November Search Engine Roundtable saw a "Web Mode" tab with web pages only and no Copilot answers, which Schwartz could reproduce in some browsers. At the same time Bing was seen placing search and shopping ads at the top of its AI answers (9 November, a sighting), and it remains the engine that splits training from grounding.

OpenAI's moves were about measurement and layout. On 10 November it opened Businesses in ChatGPT and Business Hub, where a verified business sees its mentions, profile views and clicks — with a tooltip that says "Clicks don't confirm website visits or conversions." It is the first AI visibility number that comes from the assistant itself. On 16 November three different ChatGPT results layouts were reported, one of them an AI answer with a "See more" button above web results, which Search Engine Roundtable likened to AI Overviews; Schwartz: "I couldn't replicate any of these findings myself." Each rival is now arguing about the same thing Google is: how much of the answer belongs on the results page.

October's watch list, checked

Our October analysis ended with six things to watch in November. Here is what happened to each by 18 November, from the sources above.

What we said to watchWhat happened, 1–18 NovemberVerdict
A named update, or late-October volatility left unnamedThe late-October movement was never named. A new, confirmed update, the November 2026 core update, started on 10 November and was still rolling out on 18 NovemberHappened
Enforcement of the shared-pages guidanceNo manual actions or Page Indexing drops reported in our sources. About six million Facebook Answers pages indexed (5 Nov), no Google comment; Discord public pages from 16 NovemberNot yet
Ads in AI Mode in the UK, Brazil and Mexico; any AI Mode click numbersTen more countries and US multiturn ads on 10 November; the UK, Brazil and Mexico still "this quarter". No click data from Google; the only number is an outside preprintNot yet
Clicks or queries in Search Console's AI reports"we don't have anything to announce yet" (2 Nov); a confirmed logging error on 9–11 November; branded and multimodal data reached the API on 16 NovemberNot yet
Rank-tracking data back toward normalNo new Nozzle figure; Google slowed /goto links, tested a sign-in after the first search, and SISTRIX had "No usable data" for 12 NovemberWent the other way
Courts and rules: D.C. Circuit, Penske and Chegg, the CMA, Anthropic's policyAnthropic's policy reached its 12 November effective date; the D.C. Circuit date is still ahead; nothing in our sources on Penske, Chegg or the CMA decision. New: NYT v. Brave, and Google v. SerpApi dismissed with leave to amendMostly not yet
Items from "What to watch in November" in our October 2026 analysis of Google Search, checked against sources dated 30 October – 18 November 2026.

What it adds up to

This section is our reading, not a measurement.

  • A core update is now measured mostly from inside. With /goto throttled, sign-in walls in test and a visibility index reporting no usable data on one of the first days, the most reliable account of this update will be each site's own Search Console and crawl, compared before and after the dashboard says it ended.
  • Access became the technical agenda. Google's November documents are about who may fetch a site — a CDN that must not challenge Googlebot, a Google-Agent that should be let through, a training token that is not a crawler. A misconfigured firewall can now cost more than a misjudged page.
  • The AI page is moving faster than its labels. A new default model, a header that no longer names it, and at least six unconfirmed AI Overview changes in two weeks. A screenshot from last week is not evidence of this week's page.
  • The click cost of AI Mode has a first randomised estimate. An 18.8-point fall in click-through to other sites is a preprint, under forced use, but it matches what People Inc, Ziff Davis and Reddit report. Demand moving to AI answers and visits moving with it are different things.
  • Spam is a security problem as well as a ranking one. Six million AI-written pages on one big domain and a ransomware chain built on SEO poisoning in the same fortnight: the scaled-content line Google drew in October will be tested by size and by crime, not only by thin pages.

Our tool: read your own site during a core update

While a core update rolls out and rank trackers swing — or, this month, struggle to see Google at all — your own crawl and your own Search Console are the steadier reading: they tell you which of your pages changed, and which only appear to have moved in someone else's sample. Google's November notes point at the same place from the other side. Its CDN advice and its Google-Agent document are both about what your server actually lets through, which robots.txt does not tell you and only a crawl of your own site, and your own logs, can.

That is what GLLmolfar is for. It is GetLoopLoop's desktop crawler, in preview, for Windows and Mac, and it crawls your site on your own computer. It sorts findings by impact — indexability, redirects, canonicals, internal links, structured data — and checks how pages read to AI search, with an AI-readiness score, a check of what AI bots may fetch under robots.txt, and llms.txt. It groups pages by meaning with an on-device embedding model, for topics, cannibalization and semantic search, without sending the crawl to a cloud AI service. It reads Search Console data read-only, from Google straight to your computer, runs planned crawls, and exports to Excel, Google Sheets and Jira. It also runs an MCP server, so an AI agent such as Claude Code or Codex can start a crawl and read the fix list. Since version 0.9.36 it can also download small Gemma 4 models that run on your own computer, and it has a content studio; the release note has the details.

GLLmolfar preview: an AI agent reads a site's technical SEO through MCP. In English, 14 seconds; the site and its numbers are a demo.

What to watch in the rest of November

  • The end of the core update. Google said up to two weeks from 10 November, which points to about 24 November; on 16 November Search Engine Roundtable expected it about a week later. Only the dashboard's completion entry settles it, and only after it is a before-and-after comparison fair.
  • The sign-in wall. Whether Google's apparent test of sign-in after the first search spreads, ends, or is explained — and what rank trackers say about their data while it lasts.
  • Google-Agent in practice. Whether CDNs and bot-management services ship rules for it and for Web Bot Auth, and whether Google says anything about robots.txt for it.
  • Facebook Answers. Whether the six million pages stay indexed: the clearest test of the scaled content policy on a large domain.
  • SerpApi. Whether Google amends its complaint by 30 November.
  • AI clicks in Search Console. The UK requirement for AI Overview and AI Mode click data applies from 3 December; whether Google adds anything to the AI report before then.
  • Bing's Web Mode and AI Mode ads in the UK, Brazil and Mexico. Whether the one leaves test and the other arrives this quarter as planned.

Questions and answers

Has the November 2026 core update finished?

Not as of 18 November 2026. Google's Search Status Dashboard lists it as started on 10 November at 10:06 US/Pacific, applying globally and to all languages, and the incident page had no completion entry when we read it on 18 November. Google said the rollout may take up to two weeks.

Why do rank trackers disagree about this core update?

Because Google is harder to observe from outside. Scrapers reported Google slowing its /goto result links from 30 October, a sign-in wall after the first search was spotted in mid-November, and SISTRIX reported no usable data for 12 November. Search Engine Roundtable put the most volatility on 14 November; Glenn Gabe believes the update landed on 13 November.

Does AI Mode reduce clicks to websites?

One randomised field experiment says yes: users redirected to AI Mode clicked through to external sites 18.8 points less often than with ordinary Search, and 42.7 points fewer clicked an ad. It is an unreviewed preprint with about 1,100 US users, reported by PPC Land on 17 November 2026.

What is Google-Agent, and should I block it?

Google-Agent is the name used by agents on Google infrastructure that browse and act for users. Google's help document asks sites to allow it explicitly in CDNs, firewalls and bot-detection services, using the IP ranges in user-triggered-agents.json. It does not mention robots.txt, and Google's list of user-triggered fetchers says they generally ignore robots.txt rules.

Can Search Console show clicks from AI Overviews and AI Mode yet?

No. On 2 November Gary Illyes said Google has nothing to announce. The generative AI report counts impressions, and it under-counted them on 9–11 November because of a logging error Google confirmed. AI clicks remain inside the ordinary Performance report.

Sources

Every claim above is dated and linked. Where a number comes from a measurement rather than an announcement, whoever measured it is named; where we read a write-up rather than the original, the write-up is the source listed.

The product

Watch it on every page, every day

One reading tells you where a page stands today. The product asks the same questions of the same assistants continuously, so a change is something you are told about rather than something you go looking for.

Request an invitation

Invite-only while we keep the readings honest

On this topic

Is my page ready for AI search?

Every day Crawls with the Vectorscope reading beside every page, re-checked after each fix. How the product works

Author

Olexandrauthor

15+ years in marketing and SEO. Dozens of products, mostly the US market. Research and measurement.

LinkedIn

Pass it on

Read in your language

Opens a browser translation of this English article. The original source stays in English.

All issues