324 items, newest first. Page 7 of 14.
ChatGPT's citation rates vary sharply by topic, creating visibility gaps for policy, finance, and multilateral brands.
Claude Opus 5 scores 30.2% on ARC-AGI-3, nearly four times GPT-5.6 Sol, signalling a qualitative reasoning gap.
Opus 5's price-performance edge could shift enterprise model routing and redraw LLM citation patterns.
Ahrefs launches a free llms.txt generator backed by a 137,000-domain study to improve brand visibility in LLM answers.
Anthropic's Claude Opus 5 leads quality rankings at Opus 4.8 prices, reshaping LLM workload economics.
OpenAI's health feature splits answer quality by subscription tier, creating a citation-quality risk for institutions.
Google AI Mode serves ads on 1 in 3 commercial queries, but cited sources are rarely the advertisers paying for them.
Google's Flash updates cut token costs by 65%, but the missing Gemini 3.5 Pro cedes frontier ground to rivals.
Google cites its own properties #1 and #2 in AI Mode, crowding out third-party brand content.
Google claims AI Search drives billions of weekly clicks. The data behind the claim doesn't exist in the public domain.
Chinese generative search surfaces only 8.3% of cited brands in final answers, per a new arXiv study of 160,000+ records.
Germany rules AI Overviews are publisher content, reshaping how brands must think about AI citation.
Search Console's social tracking hides AI click loss and turns verified profiles into model training data.
RAG benchmarks measure the wrong thing for agentic AI search, a new arXiv study finds.
A critical arXiv survey finds GEO's most-cited gains depend on conditions that rarely hold in live AI systems.
Invoca: ChatGPT referrals produce higher lead conversion on calls than organic or paid search.
A researcher bypassed Claude's exfiltration guard to leak user memories via web_fetch, exposing a structural flaw in AI assistant security.
Where source records are thin, LLMs hallucinate most and GEO manipulation costs least.
Anthropic finds Claude shifts tone and values by language, exposing gaps for global brands in LLM visibility.
ChatGPT access is linked to a 9% fall in Google search volume, cutting brand discovery from both ends.
New research exposes how most AI citation rankings lack the data quality to support reliable brand decisions.
RAG systems score clean on hallucination checks while citing the wrong entity's evidence, at rates up to 87%.
GPT-5.6 launches in three sizes. One shared knowledge cutoff makes pre-training authority more important than ever.
GPT-5.6 is now Microsoft 365 Copilot's default model. Better reasoning rewards structured content and exposes poor knowledge governance.