ChatGPT claims a fifth of Google's searches as publishers lose the clicks

This week two kinds of number met. OpenAI said ChatGPT now handles about a fifth as many search-like queries as Google. In the same days, two of the largest publishers in the US reported how much search traffic they no longer get: People Inc lost a quarter of its core sessions in a year, and Ziff Davis lost about 37% of its organic Google referrals in a quarter. Questions are moving to AI answers faster than visits are. The rest of the week was about who may read a site, and who only pretends to: a study of AI bots in robots.txt on 14,318 sites, and a hole in the usual way of checking that a visitor is really Googlebot.

ChatGPT says it is a fifth of Google

In an interview written up by Search Engine Journal on 3 November and Search Engine Roundtable on 4 November, OpenAI's Adam Fry said: "ChatGPT's volume of search-like queries is now roughly one-fifth the volume of queries of Google." He counts broadly: "Every time you ask a question and go search the web, that's a search." He also said that "ChatGPT has gone from one billion to 1.2 in just the last two months", meaning weekly users.

Fry did not say which Google figure he compared against, for which period or which countries. Search Engine Journal points out that in February Ahrefs estimated ChatGPT's search volume at about 12% of Google's, by a different definition. And queries are not visits. Ahrefs' traffic tracker, across 118,671 sites in October, put Google at 24.74% of tracked traffic and ChatGPT at 0.36%: about one ChatGPT visit for every seventy from Google.

Publishers count the clicks they lost

People Inc, the publisher of People and Allrecipes, reported its third quarter on 2 November. PPC Land's write-up of 6 November quotes chief executive Neil Vogel: "We've lost nearly 70% of our Google search traffic over the last few years," and "Sessions were down 26% this quarter, but revenue was only down 6%." The company's presentation puts AI Overviews on 82% of the searches it tracks; on the call on 3 November its finance chief, Tim Quinn, said "over 90%".

Ziff Davis, which owns CNET, IGN and PCMag, gave a steeper number. PPC Land reported on 7 November that its organic Google referrals fell about 37% in the third quarter, after a drop of about 23% a year earlier, while Search Console showed its pages appearing in AI Overviews and AI Mode about 2.7 billion times. Chief executive Vivek Shah: "Look, I think our position is we're not planning on a search recovery."

AI visibility: a dashboard from OpenAI, and what mentions are worth

On 6 November Search Engine Roundtable showed a screenshot shared on LinkedIn by Juozas Kaziukėnas: OpenAI appears to be building a tool to "track how often your business is mentioned in ChatGPT conversations." According to the post, it shows mentions and clicks to the business's site as a daily chart, inside a larger Business Hub, only for businesses that verify ownership, and without prompt data, user data or any comparison with competitors. OpenAI has not announced it; Kaziukėnas calls it a work in progress. If it ships, it will be the first AI visibility number that comes from the assistant itself rather than from someone sampling it.

Ahrefs published two studies on 3 November that ask where those mentions come from. In the first, Louise Linehan compared 10,000 US brands' visibility across 260 million non-branded prompts in AI Overviews, AI Mode and ChatGPT with seven kinds of mention elsewhere on the web. "Reddit mentions track AI visibility at 0.75, more than any other factor," and "Press coverage (0.67) doesn't beat general webpage mentions (0.70)." Ahrefs is careful: "These are correlations, so we can't say for certain that more mentions cause more AI visibility."

The second, by Ryan Law, lists OpenAI's publisher partnerships and counts how often ChatGPT cites each partner in Ahrefs' sample of US answers. "Ten partner domains are among the 50 sites ChatGPT cites most often in our sample, and 47 are in the top 500," and the most cited are "mostly review and lifestyle sites, not news sites": vogue.com with 42,108 citations and caranddriver.com with 41,603 lead the list.

Facebook Answers: six million AI pages in Google

On 5 November Search Engine Roundtable reported that Google has indexed about six million pages under facebook.com/fb-answers/. Each one, in Saul Westbrook's description, "consists of an AI-generated summary to explain a specific topic." Glenn Gabe, who says Facebook started filling the directory in July 2026, called it "another example of massive 'Scaled content abuse'". Google has not commented. Barry Schwartz expects the pages not to stay indexed for long now that they are public. The case is worth watching for one reason: it tests whether Google's scaled content policy applies to a site of Facebook's size the way it applies to everyone else.

Technical SEO: who reads your site, and who pretends to

AI bots in robots.txt. Seonix requested robots.txt from 16,220 sites on 3 November and read 14,318 of them. "Of 14,318 sites with a readable robots.txt, 12.2% block at least one AI training bot by name." Bots that fetch pages to answer a question are blocked far less often than bots that collect training data: GPTBot on 8.0% of sites, OAI-SearchBot on 0.7%. Seonix's reading is that a site "can refuse training and still stay visible in AI answers." The same crawl found a valid llms.txt on 3,604 of 14,875 sites that answered clearly, 24.2%. The study reads robots.txt only; it does not test what a site's firewall does.

A fake Googlebot can pass the usual check. SenseGap published log research on 5 November showing that Google's reverse DNS instructions accept hostnames ending in googleusercontent.com, and "Every Google Cloud virtual machine gets one of those hostnames by default." Over eight days, from 28 October to 5 November, its own site received 89 requests from Google Cloud machines claiming to be Googlebot that passed the loose check. Across four sites, every request claiming to be CCBot, DuckAssistBot or Perplexity-User that could be checked was fake, and so were 32% of the GPTBot ones. The fix: accept googlebot.com for Googlebot, and check the IP against the lists Google publishes for each crawler.

Gemini and AI Mode fetch as "Google" and "Google-Lens". Dave Smart of Tame the Bots, on 3 November, found in server logs that Gemini fetches a page someone asks about with the user agent Google, and AI Mode with Google-Lens. Neither rendered the page: both fetched "only the main document, and no resources like images, scripts, or stylesheets." Text that appears only after JavaScript runs is invisible to them. One client's firewall "got spooked by the short nature of the Google user-agent string alone".

Redirects, explained by Google. Search Engine Journal wrote up on 5 November how Gary Illyes and Cherry Prommawin describe redirects. For a 301, "Google takes all the ranking signals of the old page and hands them to the new page." For a 302, "Google assumes you are going back eventually, keeps the old URL in the search results, and basically ignores the new URL." A 308 and a 307 are the modern 301 and 302; "The only major difference is how they handle post data." And JavaScript and meta refresh redirects "force the browser to download the page, parse the code, wait for the script". Prommawin's advice was the calmest line of the week: "First of all, relax. No matter which redirect you choose, one of your pages will stay in search and be part of AI training."

A teapot that stayed in the index. When VLC's site returned HTTP status 418, "I'm a teapot", Gary Illyes explained on 5 November that Google reads it like most 4xx codes, as not found, but that "Search temporarily protected the URL from falling out of the index right away" because the page is important. With nothing fresh to show, the result lost its snippet and favicon. Google also rewrote its help text for "Crawled – currently not indexed" on 3 November: it is "a common message for many sites" and "doesn't necessarily indicate an issue that needs attention."

Rank trackers: Google slows its own result links

Search Engine Journal reported on 5 November that DataForSEO, Serper and SerpApi all see 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", and by the 15th "there could be a 40 second wait"; DataForSEO saw "up to 120 seconds for 1 URL". Serper first noted it at 1:30 UTC on 30 October. "Neither DataForSEO nor Serper links to a Google announcement regarding the restriction." If your rank reports show Google redirect URLs instead of your pages, this is why.

SE Ranking added data, on 3 November, to the September episode of results that had nothing to do with the query. Across 10,000 US desktop keywords, the share with mostly unfamiliar results was 0.09% on 16 September and 57.3% on 17 September, peaking at 67.2% on 24 September. In 84% of cases the odd results matched one word of the query, and 96.5% of the affected blocks started at position 9 or lower. SE Ranking says results became unstable again at the end of October.

PPC: ChatGPT ads change bidding, and run on research prompts

OpenAI told advertisers, in an email reported by Search Engine Roundtable on 5 November: "Starting Monday, November 9, you can optionally choose a cost target in Ads Manager." Max Bid and Max CPC retire on 1 January 2027; existing campaigns keep their bidding until the advertiser changes it.

Where those ads run was counted by Adthena, which sells ChatGPT ad tracking, in a report written up by Search Engine Journal on 5 November. It found 204,489 ChatGPT ad placements across five categories. In beauty, 83% of ads ran on research or discovery prompts; in apparel and consumer electronics, under half. The number of advertisers in those categories grew from 1,082 in June to 6,333 in August. The report counts where ads appeared, not which were clicked.

Google: AI Mode books the plumber and takes the payment

AI Mode kept turning answers into transactions. On 4 November Google's Robby Stein posted: "Today, we're making it even easier to research & book home services with direct links in AI Mode," in the US, with Angi, Thumbtack, Housecall Pro and Jobber, and ServiceTitan coming. The same day Google Canada brought direct checkout through the Universal Commerce Protocol to AI Mode and the Gemini app, with lululemon, Shoppers Drug Mart, Simons and selected Shopify merchants; the merchant stays the seller of record.

Three sightings, none confirmed by Google: a "Continue with this search?" screen in AI Mode that warns that "Malicious queries could result in unsafe or harmful actions", which Barry Schwartz reads as a block on bots and scrapers; a pop-up asking people to connect more Google apps to Personal Intelligence; and AI Mode missing in one person's incognito Chrome.

In Europe, Glenn Gabe shared a Search Console email sent about two months after the fact: from 30 August, past manual actions under the site reputation abuse policy "will no longer impact search results shown to users within the EEA." And BrightEdge, in Search Engine Roundtable's write-up of 6 November, found 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.

Updates and links: quiet again

Google's Search Status Dashboard lists no update that started or ended between 4 and 7 November; the last confirmed one is still the September 2026 spam update. Nothing on link building was published in the window that we would ask you to read, so there is no link section this week. The closest thing is Ahrefs' study above: mentions on Reddit and across ordinary web pages track AI visibility at least as well as press coverage does.

Our news: GLLmolfar reads the redirects, the bots and the big sites

Half of this week's technical news is about what a crawler sees: where a redirect ends, what robots.txt tells each AI bot, and whether the server agrees with it. That is what GLLmolfar, our desktop crawler, is for. It is the app called GLLmolfar until this week; version 0.9.31 gave it its short name. Five versions came out between 4 and 6 November, and each one meets a story above.

• Redirects. A crawl started at /blog/ on a site that redirects to /blog now goes on past the redirect instead of stopping at one page. We found that redirect on 7 of the 190 sites in our own research crawl. Redirect chains and redirects that end on an error are reported with the URLs behind them.
• Pages that disappear. A new mode, Re-crawl + find new, asks for every URL of the last crawl first and only then looks for new pages. A page missing next time is a 404 that was seen, not a page a smaller crawl never reached, and crawls of different size are compared only on the URLs both reached.
• AI bots. Seonix read robots.txt. Our study of 190 SEO sites, published on 4 November, also asked the servers: 19 of 147 sites refuse an AI user agent that their robots.txt allows. GLLmolfar runs the same check on your site.
• Big sites. 0.9.33 stores a crawl about five times smaller and writes to disk in batches on its own thread, so a 20,000-page crawl opens in half a second and the window keeps answering while it runs. Every table can now be exported to JSON and XML as well.

Also this week

• Search Console click spikes. Several SEOs saw clicks jump on 31 October and the days after, up to ten times a normal day for some UK clients; Joy Hawkins said she was "fairly confident it's bot traffic" and that it was back to normal on 4 November. Google has not commented. Search Engine Roundtable, 3–4 November.
• Screaming Frog SEO Spider 25.0. Released on 3 November: Network Idle becomes the default for JavaScript rendering, the crawler slows down for hosts answering with 429 and 503, and configuration files move to JSON for the command line and MCP.
• Device filters in Search Console. The Performance report now lets you pick several devices at once. Search Engine Roundtable, 5 November.
• Messaging in Google Maps. Business messages extend to the iOS and Android apps. Search Engine Roundtable, 6 November.
• A visual summary in Bing. One sighting of a button that turns a search result into a generated infographic. Search Engine Roundtable, 6 November.

Sources

Every claim above is dated and linked. Where a number comes from a measurement rather than an announcement, whoever measured it is named.

• Google's search volume is about 5x ChatGPT's, OpenAI says — Search Engine Journal, 3 November 2026, and Search Engine Roundtable, 4 November 2026
• People Inc loses 26% of core web sessions as Google AI Overviews spread — PPC Land, 6 November 2026
• Ziff Davis loses 37% of Google traffic as publishers count the AI bill — PPC Land, 7 November 2026
• ChatGPT tools for business — Search Engine Roundtable, 6 November 2026
• Which types of mentions correlate the strongest with AI visibility? — Ahrefs, 3 November 2026
• OpenAI's publisher partnerships — Ahrefs, 3 November 2026
• Facebook Answers: AI-generated pages in Google — Search Engine Roundtable, 5 November 2026
• AI crawlers in robots.txt: 14,318 sites measured — Seonix, crawled 3 November 2026
• Googlebot, reverse DNS and Google Cloud — SenseGap, 5 November 2026
• The Google and Google-Lens user agents — Tame the Bots, 3 November 2026
• Google explains how redirects affect ranking signals — Search Engine Journal, 5 November 2026
• Google can temporarily protect important URLs from being deindexed — Search Engine Roundtable, 5 November 2026
• Google updates "Crawled – currently not indexed" — Search Engine Roundtable, 3 November 2026
• Google goto result links throttling — Search Engine Journal, 5 November 2026
• Unrelated Google results hit over 57% of keywords overnight — SE Ranking, 3 November 2026
• OpenAI updates ChatGPT ads bidding options — Search Engine Roundtable, 5 November 2026
• Most ChatGPT ads in beauty and travel ran on research prompts — Search Engine Journal, 5 November 2026
• Google AI Mode shows up to four ads — Search Engine Roundtable, 4 November 2026
• Reddit begins testing ads in its own search results — Search Engine Journal, 4 November 2026
• Google Ads expands the AI content label — Search Engine Roundtable, 5 November 2026
• FTC unseals details of the Amazon ads auction surcharge suit — Search Engine Journal, 5 November 2026
• Google confirms issues with Merchant Center — Search Engine Roundtable, 6 November 2026
• Google AI Mode home services booking links — Search Engine Journal, 4 November 2026
• Google brings UCP direct checkout to AI Mode in Canada — Search Engine Journal, 4 November 2026
• AI Mode "Continue with this search?", the Personal Intelligence pop-up and AI Mode removed in a test — Search Engine Roundtable, 3–6 November 2026
• Site reputation policy in the EEA — Search Engine Roundtable, 4 November 2026
• Google Shopping in the EU after the DMA — Search Engine Roundtable, 6 November 2026
• Google Search Status Dashboard
• Search Console clicks spike, multi-select device filters, Maps messaging and Bing AI visual summary — Search Engine Roundtable, 3–6 November 2026
• SEO Spider 25.0 — Screaming Frog, 3 November 2026
• We crawled 190 SEO sites: what they tell AI bots and what their servers do — GetLoopLoop, 4 November 2026
• GLLmolfar — GetLoopLoop, desktop crawler, preview