# Last Week in AI Search: October 19 Published: 2026-10-19T11:00:42.745Z · Source: AEO Wire (https://anythingengineoptimization.com/weekly/) Source date: 2026-10-19 ## The Story ### NYT and BBC block ChatGPT. WSJ and AP block Claude. Pick your AI, pick your news. (https://anythingengineoptimization.com/item/2026-10-19-sites-with-openai-deals-let-chatgpt-s-search-in-and-keep-claude-s-out/) One AI gives you the original reporting. Another gives you someone's rewrite. Ask ChatGPT and Claude to find the same week's AI coverage. One links directly to Reuters, AP and The Wall Street Journal. The other turns up secondhand accounts, a news tracker, or nothing. That difference may look like a search-quality problem, but it's an access problem. A publisher's crawler rules can determine whether an AI's own search bot can retrieve its reporting. Licensing arrangements may help explain why those rules differ. We tested both AIs across 12 major news sites, then examined the crawler rules of 52 sites. The results raise a consequential question: Who gets to decide which reporting an AI can read, and what happens when the answer depends on whom a publisher does business with? Same question, two AIs Late Sunday night we gave ChatGPT and Claude the same prompt for twelve news sites, in each company's regular app, a new chat each time: > Find this week's AI coverage from [site]. Link the articles. Try it yourself. It takes a minute. ChatGPT linked articles straight from 9 of the 12 sites. Claude did it for 4. It was one run each, so treat it as a first look, not a stable measure. - Both reached The Washington Post, Axios and TechCrunch. - Only ChatGPT reached The Wall Street Journal, Reuters, AP, The Guardian, The Atlantic and The Verge. All six are OpenAI partners. Claude found two WSJ stories, but only by way of a news tracker. - Only Claude reached CNN. - Neither reached The New York Times or the BBC. ChatGPT failed on exactly the three sites whose robots.txt blocks OpenAI's search bot: the Times, the BBC and CNN. For the BBC it offered secondhand links instead, including one to a lookalike domain, bbc.com.im, which it labeled as unofficial. Then we asked about two scoops from last week, without naming a site. The Verge broke the story of Yelp, Delta and two banks blocking AI agents (https://anythingengineoptimization.com/item/2026-10-17-yelp-delta-and-banks-block-ai-shopping-and-booking-agents/). Reuters broke Reddit's fight over Perplexity's lawyers (https://anythingengineoptimization.com/item/2026-10-17-reddit-moves-to-bar-ex-law-firm-from-defending-perplexity/). ChatGPT linked both originals. Claude named The Verge and Reuters correctly, but couldn't link either. It cited retellings instead. In our test, Claude linked secondhand accounts of stories it couldn't link to directly. That matters because every retelling is another chance to lose context or introduce errors. A lookalike domain raises a different concern: whether the source is what it appears to be. We didn't test the answers for accuracy. We traced their sources. When an AI couldn't link to the original reporting, readers got someone else's version of it. Did the apps tell you? Yes. Both said when they fell back to a copy, and ChatGPT flagged the BBC mirror as unofficial. Then we asked two broader questions without naming an outlet, each about a scoop Claude can't reach. ChatGPT cited The Wall Street Journal and The Verge directly. Claude credited both by name but linked aggregators, and said so at the bottom: "I couldn't read them directly here." Once, it blamed a paywall instead. The warning was there. You had to read to the end to find it. Three locks, not one A site can turn an AI away in three places. Our test hit all three. The AI company. Anthropic says its bots obey robots.txt, the file that tells bots what they may fetch. Claude's developer tools refused to fetch The Verge, Reuters or the BBC. The robots.txt file. Across 52 major news, trade and reference sites, OpenAI's search bot is disallowed on 16. Claude's is disallowed on 28, and Perplexity's on 30. The server. Claude's search still listed CNN pages, pulled from a search index. But when Claude's developer tools asked CNN's server for an article, it answered "451 Unavailable For Legal Reasons." The locks don't always agree, and Anthropic doesn't publish how its tools decide. Who holds the keys The gap lines up with OpenAI's licensing ties. Seventeen of the 52 sites have them: 16 announced deals, plus Reuters, which is on OpenAI's partner list without announced terms. - With OpenAI ties (17 sites): 16 allow ChatGPT's search bot. 3 allow Claude's. - No ties (35 sites): 20 allow ChatGPT's search bot. 21 allow Claude's. These aren't stray blocks. Thirteen sites allow ChatGPT's search bot and disallow Claude's, all of them OpenAI partners, and 11 name Claude's bot in their rules. Training bots split the same way. Whether the deals are the reason, the files can't say. Who licenses to whom, from public announcements: - OpenAI: AP, News Corp (the Wall Street Journal), The Guardian, The Washington Post, The Atlantic, Vox Media (The Verge), Condé Nast (Wired), the FT, Axios, Reddit and others. - Google: more than 400 publications in its News AI pilot (https://anythingengineoptimization.com/item/2026-10-17-google-s-news-ai-pilot-expands-past-400-publications/), plus AP and Reddit. - Microsoft, Amazon and Meta: Reuters, AP, News Corp, People Inc and others; Amazon also has The New York Times. - Perplexity: revenue-share programs with Le Monde, El País, The Washington Post and Condé Nast. - Anthropic: none of these publishers has an announced deal. Google is the exception. Almost no site blocks Googlebot, because that means leaving Google Search. So Google pays for extra rights instead. Its News AI pilot covers Google News, not AI Overviews, where clicks are being lost. The paid door Licensed partners rarely use the front door. They get feeds and APIs. That's how Reddit can block every bot and still license its posts to Google and OpenAI. A block isn't pointless when there's an API around it. It can be part of what makes the API worth paying for. For brands, AI visibility starts before ranking. An AI can only cite what it can reach. What the record says to do Know which doors are open, and to whom. - Check how each AI sees you (AI citation). Don't assume one engine's sources match another's. Run our prompt with the outlets that cover you, then ask your customers' questions in each AI and compare what each cites. Engines already differ for many reasons. Access is now one more. - Know which bots the outlets that cover you allow (crawler access). Anthropic says its bots won't fetch a site that disallows them, so Claude may only repeat that site's story secondhand. Our item lists all 52 sites (https://anythingengineoptimization.com/item/2026-10-19-sites-with-openai-deals-let-chatgpt-s-search-in-and-keep-claude-s-out/), bot by bot. - Set your own rules per bot (crawler access). Training, search and user-triggered bots do different jobs, so decide each one (https://anythingengineoptimization.com/best-practices/#crawler-policy). A block is a request, not a lock. OpenAI says its user-triggered fetcher may not follow robots.txt. Perplexity says its own "generally ignores" it. Sites block Claude's search bot far more often than ChatGPT's, and Claude's is the one that says it obeys. This issue covers October 12–18; we ran the tests after the week closed, just after midnight Eastern on Monday, October 19: free ChatGPT and claude.ai (Sonnet 5.5), web search on, one new chat per question, plus Claude's developer tools. One run each, and AI answers vary from run to run. robots.txt counts are the rules each site states. Our 55-site panel was fixed in September and isn't a sample of the web; three sites refused our fetch outright. Deals come from company announcements and major outlets. The data shows where rules and deals line up, not why. Disclosure: this wire uses Claude in its own production. Read the item → (https://anythingengineoptimization.com/item/2026-10-19-sites-with-openai-deals-let-chatgpt-s-search-in-and-keep-claude-s-out/) ## The Follow-Up ### Rewriting a page for AI can shift election answers, if the rewrite names a target (https://anythingengineoptimization.com/item/2026-10-16-study-geo-rewriting-can-shift-ai-election-answer-exposure/) A new benchmark, VoteGEO-Bench, tested GEO rewriting on election questions about the 2024 US presidential race and Japan's 2026 election. Rewriting a page without a target rarely raised how often AI engines showed it. Rewriting that named a target and strengthened the argument did. A bigger effect came from the user: telling an engine you're a partisan voter can shift what it shows more than rewriting a single page can. It extends last week's Story (https://anythingengineoptimization.com/item/2026-10-12-the-week-on-the-wire/), where eight of ten AI search tools cited a made-up topic within a week. Same lever, now on elections. ## The Numbers - 97% of the pages ChatGPT's own search found weren't in Google's top 50 for the same searches. University of Toronto researchers replayed 90 days of buying questions. (preprint) Read → (https://anythingengineoptimization.com/item/2026-10-14-study-chatgpt-s-own-search-explains-citations-better-than-page-content/) - Only 14.4% of free, logged-out ChatGPT answers ran a web search at all. Ahrefs counted 66.6 million answers in August and September. (Ahrefs study) Read → (https://anythingengineoptimization.com/item/2026-10-12-ahrefs-chatgpt-triggers-a-web-search-on-just-14-of-prompts/) - Asda ranked 8th among the UK brands ChatGPT mentions most in July. It ranked 606th for citations. (Semrush study) Read → (https://anythingengineoptimization.com/item/2026-10-13-semrush-asda-s-chatgpt-mentions-don-t-match-its-citations/) - On holiday-shopping questions, ChatGPT sent 51.5% of its citations to retailers' newsroom and help pages. Google's AI Overviews sent 18.1%. (BrightEdge study; four retailers) Read → (https://anythingengineoptimization.com/item/2026-10-16-brightedge-google-ai-overviews-chatgpt-cite-different-kinds-of/) - 78% of e-commerce brands never appeared in AI shopping answers for their own category. A handful of brands took 75% of the slots. (CommerceV3 study; ChatGPT, Gemini, Perplexity) Read → (https://anythingengineoptimization.com/item/2026-10-17-commercev3-ai-shopping-answers-repeat-the-same-brands/) ## The Rest - For about a week, ChatGPT stopped tagging its links with utmsource=chatgpt.com. Adobe saw about 31,000 daily ChatGPT visits counted as direct traffic instead. Profound says the tags were back on about 99% of answers by October 16. Read → (https://anythingengineoptimization.com/item/2026-10-14-chatgpt-removes-utm-tracking-from-its-outbound-links/) - Since September 13, Google has served rank trackers swapped-out results pages, per the rank-tracking firm Nozzle. That's 8% of pages, and about a quarter on September 27–29. Most return to normal within three days. A one-day drop in late September is more likely this than a real ranking change. (Nozzle study) Read → (https://anythingengineoptimization.com/item/2026-10-18-google-feeds-rank-trackers-fake-results-to-block-ai-scraping/) - Google's results plus OpenAI's API recovered only about a quarter of the pages the ChatGPT app actually cited, per researchers at Xi'an Jiaotong University and the National University of Singapore. Many tracking tools rely on that pairing. Read → (https://anythingengineoptimization.com/item/2026-10-14-study-openai-s-responses-api-captures-only-a-quarter-of-chatgpt-s/) - The SPUR publisher coalition passed 80 members (https://anythingengineoptimization.com/item/2026-10-17-spur-coalition-adds-nyt-afp-economist-and-atlantic/), adding The New York Times, AFP, The Economist and The Atlantic. It wants AI companies to report when they use members' work. No AI company has signed on. Read → (https://anythingengineoptimization.com/item/2026-10-17-spur-coalition-adds-nyt-afp-economist-and-atlantic/) - Google is adding more links to AI Overviews and AI Mode. It's also testing a section that puts a topic's official website at the top of some answers. Read → (https://anythingengineoptimization.com/item/2026-10-15-google-adds-sources-section-more-links-to-ai-overviews-and-ai-mode/) - Google launched Universal Cart in AI Mode and Gemini. Shoppers can fill one cart from several retailers and check out with Google Pay. Read → (https://anythingengineoptimization.com/item/2026-10-15-google-launches-universal-cart-in-ai-mode-and-gemini/) - Barn2, a WordPress plugin maker, spent $477 on ChatGPT ads over six weeks. It got 370 clicks and no sales. Over the same weeks, visitors from ChatGPT's regular answers made five purchases. It's one small business. Last week we reported that OpenAI's own ad case studies came with no spend or sample size (https://anythingengineoptimization.com/item/2026-10-12-the-week-on-the-wire/). Read → (https://anythingengineoptimization.com/item/2026-10-18-barn2-spends-477-on-chatgpt-ads-logs-zero-conversions/) - Amazon added a claim against Perplexity's shopping agent: that it used customers' copied login cookies to reach Amazon from Perplexity's own servers. Perplexity denies it. Read → (https://anythingengineoptimization.com/item/2026-10-17-amazon-adds-new-claim-against-perplexity-s-shopping-agent/) - Coming Oct 30 — In re Mosaic LLM Litigation (Databricks): Hearing before Judge Breyer on whether the case can be decided without a trial (summary judgment) The docket → (https://anythingengineoptimization.com/docket/) - Correction, 2026-10-17 — Chartbeat: Google traffic to publishers fell 40% in a year (https://anythingengineoptimization.com/item/2026-10-17-chartbeat-google-traffic-to-publishers-fell-40-in-a-year/): This item originally said small publishers lost about 60% of their search referral traffic, medium publishers 47% and large publishers 22%, and attributed that breakdown to Chartbeat's 2026 Publisher Playbook. The playbook gives no breakdown by publisher size. The 60% figure comes from earlier Chartbeat data that Axios reported in March, and Search Engine Journal says the playbook doesn't break down its data by size. The breakdown has been removed from the item, its Q&A and its summary. --- All 43 items from October 12–18, dated and sourced: the week on the wire (https://anythingengineoptimization.com/item/2026-10-19-the-week-on-the-wire/) · State of Search (https://anythingengineoptimization.com/state/) · Corrections (https://anythingengineoptimization.com/corrections/) · Archive (https://anythingengineoptimization.com/archive/). Seen something the record should hold? Reply to this email; tips are credited by name. ## The full wire 43 items published October 12–18. Every item dated, sourced and linked. An item can carry more than one lane, so lane totals exceed the count. ### Research · 13 - Oct 13 — Semrush: Asda's ChatGPT mentions don't match its citations (https://anythingengineoptimization.com/item/2026-10-13-semrush-asda-s-chatgpt-mentions-don-t-match-its-citations/) — PPC Land (vendor study) - Oct 13 — Claude recommends different brands by language, study finds (https://anythingengineoptimization.com/item/2026-10-13-claude-recommends-different-brands-by-language-study-finds/) — Search Engine Journal (single source) - Oct 13 — Profound: Sonnet 5.5 asks for context 5x more than Sonnet 5 (https://anythingengineoptimization.com/item/2026-10-13-profound-sonnet-5-5-asks-for-context-5x-more-than-sonnet-5/) — Profound research blog (vendor study) - Oct 13 — Study: ChatGPT's search beats content for citations (https://anythingengineoptimization.com/item/2026-10-14-study-chatgpt-s-own-search-explains-citations-better-than-page-content/) — arXiv (study) - Oct 13 — OpenAI's Responses API misses most ChatGPT citations (https://anythingengineoptimization.com/item/2026-10-14-study-openai-s-responses-api-captures-only-a-quarter-of-chatgpt-s/) — arXiv (study) - Oct 14 — Profound ties missing ChatGPT UTM tags to Intelligent UI (https://anythingengineoptimization.com/item/2026-10-14-profound-ties-missing-chatgpt-utm-tags-to-intelligent-ui/) — Profound research blog (vendor study) - Oct 15 — Study: gap-search method cuts homogenization in GEO content (https://anythingengineoptimization.com/item/2026-10-15-study-gap-search-method-cuts-homogenization-in-geo-content/) — arXiv GEO (study) - Oct 15 — Typeface survey: marketers plan to scale back SEO for AEO (https://anythingengineoptimization.com/item/2026-10-15-typeface-survey-marketers-plan-to-scale-back-seo-for-aeo/) — PPC Land (vendor study) - Oct 15 — Study: GEO rewriting can shift AI election-answer exposure (https://anythingengineoptimization.com/item/2026-10-16-study-geo-rewriting-can-shift-ai-election-answer-exposure/) — arXiv GEO (study) - Oct 16 — Google AI Overviews vs ChatGPT holiday citations (https://anythingengineoptimization.com/item/2026-10-16-brightedge-google-ai-overviews-chatgpt-cite-different-kinds-of/) — BrightEdge Blog (vendor study) - Oct 17 — Chartbeat: Google traffic to publishers fell 40% in a year (https://anythingengineoptimization.com/item/2026-10-17-chartbeat-google-traffic-to-publishers-fell-40-in-a-year/) — Chartbeat (vendor study) - Oct 17 — CommerceV3: AI shopping answers repeat the same brands (https://anythingengineoptimization.com/item/2026-10-17-commercev3-ai-shopping-answers-repeat-the-same-brands/) — CommerceV3 (vendor study) - Oct 18 — Claude-SearchBot blocked more than OpenAI's crawler (https://anythingengineoptimization.com/item/2026-10-19-claude-searchbot-blocked-far-more-than-oai-searchbot-study-finds/) — OpenAEO (vendor study) ### Google Search · 12 - Oct 13 — Google tests pharmaceutical ads in AI Mode and AI Overviews (https://anythingengineoptimization.com/item/2026-10-13-google-tests-pharmaceutical-ads-in-ai-mode-and-ai-overviews/) — Adweek - Oct 13 — Google adds Gemini 4 Argon to AI Mode's model menu (https://anythingengineoptimization.com/item/2026-10-13-google-adds-gemini-4-argon-to-ai-mode-s-model-menu/) — Search Engine Journal - Oct 13 — Eli Lilly runs first prescription drug ad in Google AI Mode (https://anythingengineoptimization.com/item/2026-10-13-eli-lilly-runs-first-prescription-drug-ad-in-google-ai-mode/) — Adthena (vendor study) - Oct 14 — Google's own hosts can't name its documented 64MB PDF crawl limit (https://anythingengineoptimization.com/item/2026-10-14-google-s-own-hosts-can-t-name-its-documented-64mb-pdf-crawl-limit/) — PPC Land - Oct 14 — Google tests AI Mode sidebar on main search results page (https://anythingengineoptimization.com/item/2026-10-14-google-tests-ai-mode-sidebar-on-main-search-results-page/) — Search Engine Roundtable (single source) - Oct 15 — Google launches Universal Cart in AI Mode and Gemini (https://anythingengineoptimization.com/item/2026-10-15-google-launches-universal-cart-in-ai-mode-and-gemini/) — Google Keyword Blog (company announcement) - Oct 15 — Google adds sources section to AI Overviews, AI Mode (https://anythingengineoptimization.com/item/2026-10-15-google-adds-sources-section-more-links-to-ai-overviews-and-ai-mode/) — Google Keyword Blog (company announcement) - Oct 16 — Google drops quota-increase form from Indexing API docs (https://anythingengineoptimization.com/item/2026-10-16-google-drops-quota-increase-form-from-indexing-api-docs/) — Google Search Central documentation updates (company announcement) - Oct 17 — Google confirms Nov 13 cutoff for AI Plus Pro access (https://anythingengineoptimization.com/item/2026-10-17-google-confirms-nov-13-cutoff-for-ai-plus-pro-access/) — 9to5Google - Oct 17 — Google's News AI pilot expands past 400 publications (https://anythingengineoptimization.com/item/2026-10-17-google-s-news-ai-pilot-expands-past-400-publications/) — Google (company announcement) - Oct 17 — Google tests Personal Intelligence ads inside AI Mode (https://anythingengineoptimization.com/item/2026-10-17-google-tests-personal-intelligence-ads-inside-ai-mode/) — Search Engine Land - Oct 17 — Google feeds rank trackers fake results to block AI scraping (https://anythingengineoptimization.com/item/2026-10-18-google-feeds-rank-trackers-fake-results-to-block-ai-scraping/) — Nozzle (vendor study) ### Engines · 6 - Oct 12 — Ahrefs: ChatGPT triggers a web search on just 14% of prompts (https://anythingengineoptimization.com/item/2026-10-12-ahrefs-chatgpt-triggers-a-web-search-on-just-14-of-prompts/) — Ahrefs Blog (vendor study) - Oct 13 — Bing Webmaster Tools adds Topics Portfolio to AI Performance (https://anythingengineoptimization.com/item/2026-10-13-bing-webmaster-tools-adds-topics-portfolio-to-ai-performance/) — Bing Blogs (company announcement) - Oct 17 — ChatGPT ad targeting uses prompts, email and phone (https://anythingengineoptimization.com/item/2026-10-17-chatgpt-ad-targeting-uses-prompts-email-and-phone/) — 404 Media - Oct 17 — Yelp, Delta and banks block AI shopping and booking agents (https://anythingengineoptimization.com/item/2026-10-17-yelp-delta-and-banks-block-ai-shopping-and-booking-agents/) — The Verge - Oct 18 — Barn2 spends $477 on ChatGPT ads, logs zero conversions (https://anythingengineoptimization.com/item/2026-10-18-barn2-spends-477-on-chatgpt-ads-logs-zero-conversions/) — Barn2 - Oct 18 — OpenAI's ChatGPT ad team reorganized seven times in a year (https://anythingengineoptimization.com/item/2026-10-18-openai-s-chatgpt-ad-team-reorganized-seven-times-in-a-year/) — TokenPost ### Law & Money · 6 - Oct 13 — News/Media Alliance: exclude AI from Google choice screen (https://anythingengineoptimization.com/item/2026-10-13-news-media-alliance-exclude-ai-from-google-choice-screen/) — News/Media Alliance (single source) - Oct 17 — Amazon adds new claim against Perplexity's shopping agent (https://anythingengineoptimization.com/item/2026-10-17-amazon-adds-new-claim-against-perplexity-s-shopping-agent/) — The Fashion Law - Oct 17 — Meta narrows AI-scraping clause in its business terms (https://anythingengineoptimization.com/item/2026-10-17-meta-narrows-ai-scraping-clause-in-its-business-terms/) — The Verge - Oct 17 — Reddit moves to bar ex-law firm from defending Perplexity (https://anythingengineoptimization.com/item/2026-10-17-reddit-moves-to-bar-ex-law-firm-from-defending-perplexity/) — Reuters - Oct 17 — SPUR Coalition adds NYT, AFP, Economist and Atlantic (https://anythingengineoptimization.com/item/2026-10-17-spur-coalition-adds-nyt-afp-economist-and-atlantic/) — The SPUR Coalition - Oct 17 — UK publishers urge Burnham to act on deceptive AI bots (https://anythingengineoptimization.com/item/2026-10-17-uk-publishers-urge-burnham-to-act-on-deceptive-ai-bots/) — News Media Association ### Tools & Vendors · 5 - Oct 12 — Zeta Global acquires Senso, its GEO engineering partner (https://anythingengineoptimization.com/item/2026-10-12-zeta-global-to-acquire-senso-bringing-its-geo-engineering-in-house/) — Zeta Global - Oct 14 — Peec AI adds Views, custom dashboards for AI search tracking (https://anythingengineoptimization.com/item/2026-10-14-peec-ai-adds-views-custom-dashboards-for-ai-search-tracking/) — Peec AI blog (company announcement) - Oct 14 — ChatGPT removes UTM tracking from its outbound links (https://anythingengineoptimization.com/item/2026-10-14-chatgpt-removes-utm-tracking-from-its-outbound-links/) — Search Engine Roundtable - Oct 16 — Profound relaunches Agents, powered by its AI Marketer Aim (https://anythingengineoptimization.com/item/2026-10-16-profound-relaunches-agents-powered-by-its-ai-marketer-aim/) — Profound research blog - Oct 16 — ChatGPT's missing UTM tags return, Profound tracking shows (https://anythingengineoptimization.com/item/2026-10-16-chatgpt-s-missing-utm-tags-return-profound-tracking-shows/) — Profound research blog (vendor study) ### AEO · 1 - Oct 18 — Which AI can read which site: OpenAI deals draw the line (https://anythingengineoptimization.com/item/2026-10-19-sites-with-openai-deals-let-chatgpt-s-search-in-and-keep-claude-s-out/) — AEO Wire crawler panel (first-party data) Read the complete archive → (https://anythingengineoptimization.com/archive/) Why it matters: The one place on the wire where the week is read rather than reported — compiled from the published record, framed by the editor, checkable line by line. ## What this answers **How many items did the AI-search wire publish this week?** 43 items between 2026-10-12 and 2026-10-18, across the lanes AEO, Research, Google Search. Each is dated and sourced on the wire. Canonical: https://anythingengineoptimization.com/item/2026-10-19-the-week-on-the-wire/ From Anything Engine Optimization (AEO Wire) — https://anythingengineoptimization.com/ · Standards: https://anythingengineoptimization.com/standards/