Wissenswertes

Key Takeaways: Rudzki (Peec AI) analyzed 5 million query fanouts across ChatGPT, Perplexity, and Grok (April, 2026) and found: Perplexity commonly just simplifies your query - not much insight for optimization (1.4 fanouts per prompt) ChatGPT commonly adds brands, comparisons, and reviews - crucial for optimization for AI search (2.1 fanouts per prompt) Grok runs a full research brief with year modifiers and site: operators - the richest signal of all if you want to optimize

Key Takeaways: Konitzny analyzed one prompt in 10 different markets in ChatGPT, Perplexity and AI Mode ""I need a juicer that produces little waste and is not very loud. Price doesn't matter."" Retrieval via Peec AI 10 Markets: US, DE, FR, NL, UK, IT, DK, AUT Listicles: biggest share on average, but massive variance by market → dominate the US. In Germany and the UK, they account for roughly half that share (US: 53.9% / DE: 25.9% / UK: 25.7%) Reddit: almost entirely a US phen

Key Takeaways: On 16th of April, Claude Opus 4.7 was released Wojcik cited the AMD report about performance issues with Claude Opus 4.6 BridgeBench reported that accuracy on the hallucination benchmark of Opus 4.6 dropped to 68,3% vs. 83,3% before He outlines that it is a natural rhythm of hype cycles that push providers to allocate infrastructure to newer models taking power away from other models He proposes to adjust flexibly to the time of day: e.g. during end of business

Key Takeaways: Segonzac (RESONEO) published a full technical study on how ChatGPT works and what changed when GPT-5.4 was introduced on 04.03.2026 and GPT-5.3 became the new default: Average domains dropped by -20% - fewer domains share the visibility in ChatGPT answers - which is also supported by Crawl log data from ChatGPT User Agents Query Fan-Out changed and includes site: Parameters to filter highly relevant or authorative sites Single tool web.run evolved into 12 o

Key Takeaways: Indig analyzed over 21.000 ChatGPT citations from different industries in depth to cover how AI cites sources based on content length and which pages/domains are picked etc. His findings / statements: Text with a question mark is 2x more likely to be cited (18.9% vs. 9.4%) 43.2% of sources cited by ChatGPT rank in Google position #1 ChatGPT retrieves about 6x more pages than it cites Top 10 domains take 46% of all citations in a topic, Top 30 at 67% While pages

Key Takeaways: writesonic analyzed 50 prompts (re: SaaS) in ChatGPT comparing GPT-5.3 Instant (new default) and GPT-5.4-Thinking (new premium) vs. GPT-5.2 Instant and GPT-5.2 Thinking and found: in GPT-5.4 Thinking a total of 53% of citations were from brand websites , compared to GPT-5.3 Instant's 8% a major shift (it cited reddit.com , techradar.com , forbes.com and others instead ) GPT-5.4 Thinking cites totally different sources as GPT-5.3, most of them not from Googl

Key Takeaways: Konitzny analyzed Microsoft Copilot Ads source code and found Ad Formats and Selection Reasons for Ads Copilot Ads Ad Formats : multimedia products tourActivity propertyPromotion text miniShowroom mainShowroom miniProduct Copilot Ads SelectionReasons reveal how personalized ads will be delivered: current-search-term recent-searches age gender website-visits interests custom-audience similar-audience profile-targeting The observed focus on current search terms

Key Takeaways: Konitzny analyzed ChatGPT code and found traces of ChatGPT's Ad System called 'Bazaar' Ad Types: single vs. multi advertiser ads single_advertiser _ad_unit multi_advertiser _ad_unit Ad design: will contain Brand name and Image Brand name Website Title Description text Link Image Actions/Interactivity with Ads: users can hide/block advertisers and ask ChatGPT about the ad CHATGPT_BAZAAR_PLACEMENT_ACTION_TYPE_ HIDE_ADS_PLACEMENT CHATGPT_BAZAAR_PLACEMENT_ACTION_TY

Key Takeaways: Landwehr analyzed Query Fanouts from ChatGPT when running Prompts in German, French, Portuguese, Italian, Spanish, Dutch, Polish and Turkish using local IPs and found that very often English Fanout Queries are triggered In 60%+ of cases, at least one of fanout query is in English. No matter the user’s language and location! Pure local brands in non-English markets are structurally disadvantaged . Brands from English-speaking countries start with a global edg

Key Takeaways: Ray analyzed websites (US) that were hit by an unconfirmed Google Update in January and found: Dropping in Google SEO also negatively impacted their LLM Visibility The data shows a broad decline in both SEO traffic & AI search citations: every subfolder in the study (11 of 11) experienced a drop in both Google organic traffic and total AI search citations , with a significant average citation decline of -22.5% . Google’s AI Mode (-23.8%) and ChatGPT (-27.8%)

Key Takeaways: Michalik created 3 listicles for a made up matcha tea that got listed in most major AI chats Claude and AIO cited the fake Matcha quite frequently ChatGPT and AI Mode also cited the fake product from time to time Gemini did the best job in not mentioning the fake product Now all chats caught up, still, way to easy to game in the short run A while ago Fishkin reported that only in <1% of the runs of the same prompt the KI chat returns the same list of recommen

Key Takeaways: Landwehr analyzed AI Search responses for pnewind(.)com - a former wind energy company website that has been misused by Casino Affiliates using Google cloaking and access blocking His findings suggest: OpenAI is scraping Google Grok is somehow using Google Claude is using Brave Search Perplexity might use Brave Search (but states that it uses its own index) Konitzny points out that Grok is using Brave for image search - based on the code run in Chrome. Ray con

Key Takeaways: Segonzac analyzed the behavior of AI Search Agents on how they view websites The Good - Static HTML is visible to everyone (your safest bet) - <noscript> fallbacks work for ChatGPT, Claude & Gemini The Surprising - Only 3 crawlers execute JavaScript​ : Bing Copilot, Grok & DeepSeek - Bing Copilot is the ONLY one that can read iframes and Shadow DOM - Some crawlers identify themselves honestly (ChatGPT, Claude, Gemini), others operate in full stealth mode (

Key Takeaways: ChatGPT is trying to build their own Knowledge Graph , Shopping Graph, Local Data, and more From the release notes: "Answers now also highlight important people , places , products , and ideas . You can tap any highlight to open a side panel with key facts and trusted sources, making it easy to understand the context without asking follow-up questions." Seems like ChatGPT realized that it has to become more like Google Inlcuded image contains "sample photo" wa

Key Takeaways: SEMrush analyzed 217,000 prompts across Google AI Mode, Perplexity, and ChatGPT Search: SearchGPT references Reddit most often . 12.6% of its answers contain Reddit links, though these links appear later in responses (avg. position 7). Perplexity gives Reddit higher prominence. Only 3.5% of Perplexity answers include Reddit, but those links appear much earlier (avg. position 3). Google AI Mode uses Reddit moderately. Reddit appears in 9% of responses, usua

Key Takeaways: Landwehr and Wells analyzed PeecAI data on 20M+ Query Fanouts (QFOs): Fanout queries used to have an average length of 6 words . Mid November they started to become longer. They peaked at 16 words in early December. Since Christmas, they are stable at 12 . Global change : observed across multiple languages and countries 10+ countries in North America, Europe, and Asia The structure of these new fanout queries suggests they are optimized for vector search - not

Key Takeaways: Metehan analyzed Common Crawl (CC) data and found a correlation to citation frequency in LLM Chats such as ChatGPT, Perplexity and others: Most major LLMs were trained on CC data (64% of models studied, 80%+ of GPT-3 tokens) CC prioritizes high-authority domains in its crawling via Harmonic Centrality These same domains tend to be cited most frequently by LLMs Feel free to use his tool to analyze Common Crawl authority scores for your domain or industry:...

Key Takeaways: OpenAI officially introduces Ads to monetize ChatGPT- first to ChatGPT logged out and Go users in the US - ChatGPT Pro, Business and Enterprise remain ad-free for now OpenAI to reportedly charge on an impression (pay per view, CPM) basis not CPCs for ChatGPT ads, require $1M spend for each advertiser and it will be done manually on some level Our ads principles from OpenAI's statements Mission alignment : Our mission is to ensure AGI benefits all of humanit

Key Takeaways: On January 11th Google launched the Universal Commerce Protocol : UCP will soon power a new checkout feature on eligible Google product listings in AI Mode in Search and the Gemini app, allowing shoppers to check out from eligible U.S. retailers right as they’re researching on Google UCP is built to work across verticals and is compatible with existing industry protocols like Agent2Agent ( A2A ), Agent Payments Protocol ( AP2 ) and Model Context Protocol ( M

Key Takeaways: Long (Nectiv) analyzed ChatGPT Query Fan-Out via 8.500+ prompts across different industries in October : Total Number Of Searches: 2,648 Percentage Of Search Instances: 31% Average Number Of Searches: 2.17 Average Words Per Query: 5.48 A search triggers 1 to 4 fan out queries that are mostly 3+ words long with an average of 5,84 words Compared to Google (3,4 words per query) the ChatGPT fan-out queries are significantly longer and more long-tail The main n-gram