AI Search news and analysis on Web Pulse: 728 curated articles tracking the latest AI Search developments, tools, and research, updated continuously from vetted sources.
Anygrep, a new MIT-licensed terminal tool from developer kuroko1t, performs semantic search over notes, PDFs and images using Google's EmbeddingGemma 2 model converted to int8 ONNX and run entirely on the CPU, keeping a …
A developer scanned 1,902 local-business websites across the US and UK and found that 24.9% of the 1,355 readable sites already serve an llms.txt file, a figure driven almost entirely by the site builder rather than owne…
Anthropic published its 2026 Usage Policy update on October 8, adding a section titled "Do Not Engage in Deceptive Campaigns or Artificial Activity" that prohibits using Claude to manipulate the sources search engines or…
A 9B typed-answer decision model called nimble matched the shipped 27B general language model qwen3.8 on recall@5 in a reranking test on 800 questions from the public ConditionalQA benchmark, putting a good answer in the…
Google has been poisoning the search engine results pages it serves to rank trackers since September 2026, with 8.0% of SERPs coming back poisoned, according to a Nozzle deep dive by Derek Perkins. Perkins described a th…
Salesforce's Winter '27 release, announced August 31 and generally available October 12, moves agent authorization into the CRM by shipping an API Catalog that lets administrators control which agents may call which comm…
Microsoft is experimenting with its new Microsoft-Decision-1 decision-scoring model in both Bing Search and Copilot, following the model's launch last week in Microsoft Foundry and on OpenRouter. Rangan Majumder said on …
A developer analyzed the robots.txt and server behavior of 740 of the top 1,500 Tranco-ranked domains, finding that 14.7% block OpenAI's training crawler GPTBot while only 7.8% block OAI-SearchBot, the agent that control…
Apify's RAG Web Browser returned 372,137 characters of extracted content across 8 substantive usable records out of 10 search results, versus Jina AI Search's 253,621 characters across 6 usable records, in a head-to-head…
A developer has published a guide to writing an llms.txt file, a plain-text Markdown document placed at a site's root that tells AI assistants which pages to read first, using a four-part format of title, summary, and li…
A free tool called the Gemini Grounding Gap Checker, built by an SEO practitioner, runs a user's query through Gemini five times and counts how often each search, source and claim recurs, then compares those claims again…
Researchers at ETH Zürich demonstrated that a fictitious camera promoted through a webpage with a concealed white-text prompt injection raised its recommendation rate in Microsoft's Bing Copilot from 34% to 59.4%, slight…
A developer behind DOS AI, an AI assistant for small businesses, built a sub-300-line script that queries a web-search-enabled model with 35 buyer-style questions to measure whether AI answer engines name and cite the pr…
A developer describes a two-stage, tenant-bounded retrieval architecture for recruiting search that enforces isolation as a correctness requirement rather than a latency trade-off. The approach applies tenant and permiss…
A developer's Day 11 explainer walks through how embedding models convert text into dense vectors of roughly 768 to 1,536 numbers so that semantically similar passages cluster together, and demonstrates cosine similarity…
A developer built an AI Citation Readiness Checker that scores pages against eight structural conditions for generative search engine citation, then validated it against four external pages after realizing all prior test…
Alham Fikri Aji's new HyperBrowseComp benchmark found that five evaluated AI systems failed to answer 244 of its 423 multilingual search questions using their built-in web search, leaving 58% unsolved. The 423-question b…
A developer's reference implementation, "Building an AI Search Visibility & Brand Analyzer with Gemini, BigQuery, and Google Search Grounding," treats AI search visibility as a measurable engineering problem by running r…
Testing multilingual AI search requires scoring retrieval against judged source passages before evaluating generated answers, according to a how-to guide that separates language direction from locale and applicability. T…
Terraphim released a preview of a browser-based Markdown editor that runs as WebAssembly with a bundled local knowledge graph, letting users write, annotate and search against the same graph, according to the company's d…