Anthropic is offering new details about how it plans to watermark text written by Claude — including what happens when someone edits the output.

The company uses a simple example to explain the technology. If Claude can choose between words such as “overcast” and “grey” to describe the weather, it can make those small choices in a pattern. A reader wouldn’t notice. But someone with the right detection tool could identify the pattern as a watermark.

Anthropic described the system in a Friday TechCrunch blog post answering questions about the plan, which it announced earlier in the week to comply with the European Union’s AI transparency rules. We covered the original announcement here.

The company plans to use SynthID Text, a watermarking technology developed by Google DeepMind. Anthropic also says it will release an API to detect the watermark.

The system has limits, particularly once people start changing Claude’s words.

“Light editing probably won’t remove the watermark completely,” Anthropic said. But a complete rewrite in which every word is replaced would remove it. Anthropic said in such a case, it becomes harder to describe the finished product as AI-generated.

Whether a watermark can be detected also depends on how much Claude actually wrote. If someone gives Claude a human-written document and asks it to make only minor edits, Anthropic said that leaves “very little (if anything) for the watermark to attach to.”

Code presents a similar problem. Code gives a model fewer stylistic choices than prose. Anthropic said it can still watermark choices, including some comments and names within code, but the technology should have “a negligible effect on the actual code produced.”

The company is also drawing a distinction between watermarking and statistical AI detectors.

Those tools look for patterns associated with AI writing and estimate whether a model produced a piece of text. A watermark works differently, using a pattern deliberately inserted when the text was generated.

An AI detector makes a statistical judgment. A watermark provides evidence that a particular system marked the text when it was created.

Not everyone is happy about the change. TechCrunch reported that dozens of Claude users on X said they had canceled subscriptions over the watermarking plan, citing Business Insider.

For news organizations, the technology could make questions about AI use easier to investigate.

An editor could check whether copy submitted by a freelancer carries Claude’s watermark. But competitors, critics and others with access to Anthropic’s detection system could check published work as well.

That makes newsroom policies around AI use more consequential. Publishers may need clear rules about when it is permitted, how much AI-generated material can appear in published work and when that use should be disclosed.

Claude won’t necessarily be the only model leaving marks behind. Anthropic says other major AI developers that signed the same European code of practice are also working on watermarking systems.

What remains unclear is whether those systems will work together. Publishers find themselves checking text against a different system for every AI company.

The post Anthropic says Claude’s text watermark survives light editing but not rewrites appeared first on The Media Copilot.

Gaurav Misra spent about $50,000 to build and run a television news operation for a day, without hiring a single journalist.

About $30,000 of that went to Claude Code credits as Misra, the CEO of AI video company Mirage, built the operation himself. The result was the Mirage News Network, a 24-hour experiment in using artificial intelligence to produce something resembling a traditional television news channel.

The broadcast drew about 50,000 viewers on X and the livestream post generated 824,000 impressions, according to Variety. Mirage paid several accounts to promote the stream, meaning some of that reach was purchased.

The Mirage News Network ran for 24 hours on Aug. 11. Four AI-generated anchors, styled after television broadcasters from the 1980s and given names including “Tom Callahan” and “Marcus Sterling,” delivered stories licensed from Reuters that were already several days old.

Mirage used its Avatar X model to create the anchors, while ChatGPT’s image generator and Google’s Gemini text-to-speech technology helped produce other parts of the broadcast. The humans who appeared on screen came from Reuters video reports and stock footage.

Even the guests were partly synthetic.

Mirage recruited more than a dozen people, many of them venture capital professionals, who allowed the company to use their likenesses and approved scripts for their AI doubles to deliver. Billy McFarland, the organizer of the failed Fyre Festival, was among them.

Mirage does not employ journalists. Misra, a former Snap and Microsoft engineer, told Variety that he instructed the system to follow “general journalistic standards,” rather than a specific set of guidelines such as AP style or the Society of Professional Journalists’ code of ethics.

The company did disclose its use of AI. An on-screen graphic identified the anchors, graphics, guests and advertisements as artificially generated.

But the experiment also showed some of the problems that can emerge when a news operation has little human oversight. At one point, the broadcast incorrectly identified a woman appearing in a Reuters report as President Donald Trump. Mirage later posted a correction on X rather than correcting the error during the broadcast.

Dan Axelrod, chair of the Society of Professional Journalists’ ethics committee, told Variety he considered Mirage’s experiment a news delivery project rather than a news network. He called it “a false and misguided devaluing and deskilling of journalism.”

His committee proposed revisions to SPJ’s code of ethics the same week, as the organization considers how longstanding journalistic principles should apply as newsrooms adopt new technology.

Even some people inside Mirage treated the project more like an experiment than the future of television news. Engineer Jason Silberman described it on X as what happens “when the CEO starts tokenmaxxing too hard and spends $50k in a day on a marketing stunt.”

There is also little evidence so far that large numbers of Americans want AI to become their primary source of news.

A Gallup survey released in May found that 7% of Americans said they relied on AI for news to at least a moderate degree. Nearly 4 in 10 said a news organization’s use of AI would make them trust its reporting less.

For publishers, Mirage may be more useful as an experiment in the economics of automated news than as a model for replacing cable television.

Smaller AI-native news operations are already testing versions of the idea. Runtime Wire operates with one human editor and says its system rejects 97.7% of the stories it considers, leaving at least some editorial judgment with a person.

Disclosure is another unresolved issue. Mirage prominently labeled its synthetic broadcast, but research into AI labeling has shown that disclosure alone does not guarantee audiences will notice or understand those labels.

At the same time, the financial pressure pushing publishers toward automation is real. Bauer Media, for example, stopped commissioning freelance fiction writers for Take a Break as it began using AI to help produce stories internally.

Misra told Variety that AI-generated newscasts are “probably inevitable in the future.”

The post Mirage’s $50,000 AI news channel runs 24/7 without journalists appeared first on The Media Copilot.

A PR agency representative offered guaranteed placement on a website impersonating Forbes for $10,000, according to a recorded sales call obtained by Press Gazette. Forbes later confirmed that forbesla.com has no connection to the company and said it is taking action against it and another site posing as a Forbes publication.

The two fake Forbes sites publish dozens of stories a day under bylines with little or no apparent presence elsewhere online, Press Gazette first reported. Reporters Nathan Pine and Liam Redmond appear to have no social media profiles. AI detection tool Pangram classified one of Redmond’s articles as entirely AI-generated.

Press Gazette also reported that Identifai flagged the byline photo of Harriet Caldwell, described by Forbes Liechtenstein as an experienced news editor, as AI-generated. Her articles were also flagged as likely to have been generated by AI.

The sales pitch extended beyond the fake Forbes sites. Baden Bower, a public relations agency that promises clients guaranteed media coverage, was the company whose representatives offered the placements documented by Press Gazette. One representative, Zac, offered placement in the Los Angeles Times, said The Guardian and The New York Times were the only publications where he could not secure coverage, and quoted $9,900 for a Bloomberg TV interview with an Emmy-winning journalist. He also suggested paid articles could help applicants seeking U.S. skilled worker visas.

A second Baden Bower representative, Camilo, said he was unaware of Forbes LA and that the agency works only with licensed regional Forbes editions. He quoted $25,000 for Forbes US, $2,000 for Forbes Georgia and $100,000 for The Wall Street Journal.

Asked whether paid placements are labeled as promotional, Camilo said, “Some are, some are not.” He also claimed Baden Bower has commercial relationships with more than 1,980 publications, including Press Gazette. Press Gazette said it has never published a sponsored article from the company.

Much of the content on the fake sites promotes companies, including Baden Bower itself. One Forbes Liechtenstein article, which Pangram classified as entirely AI-generated, projects that the agency will reach 7,600 clients and 90,000 published stories by the end of 2026 and quotes company leader AJ Ignacio touting its commercial results.

Baden Bower then cites that article on its own website as evidence of its standing, alongside pieces from an AI-generated Philippines edition of Fast Company and an apparent fake German edition of Vanity Fair.

The sites are not new. The Internet Archive shows Forbes LA online since January 2025 and Forbes Liechtenstein since March 2025. The London office listed by Baden Bower has a single contact number, a mobile number that appears to be disconnected.

For publishers, the risk is to the masthead itself. Fake sites can cheaply trade on the authority of established news brands, turning their names and reputations into promotional inventory. The model echoes fake AI news networks built to game search results and fabricated AI expert profiles reaching real reporters, but extends the problem into the PR industry.

Sarah Waddington, CEO of the Public Relations and Communications Association, said the practice violates industry standards on transparency, honesty and earned credibility. Former BBC journalist Omar Hamdi, who first noticed founders sharing fake Forbes coverage on LinkedIn, said the promotional tone of the articles raised his suspicions.

For newsrooms, the simplest defense may be basic verification: check that the publication behind a citation actually exists before trusting it.

The post Fake Forbes sites sold guaranteed coverage for $10,000, PR pitch shows appeared first on The Media Copilot.

The New York Post is putting its founder’s name on its latest push into artificial intelligence. Starting Tuesday, readers using the New York Post and California Post apps will have access to Hamilton, a suite of personalization tools built around an AI chatbot, according to Axios.

The name is a nod to Alexander Hamilton, who founded the paper in 1801. The tools are designed to help readers discover more stories and receive recommendations based on what they read, according to Ariscielle Novicio, chief technology officer and senior vice president of product and digital strategy at New York Post Media Group. The company plans to bring Hamilton to its websites later.

Hamilton includes four features: the Hamilton Search chatbot, the Post Express personalized briefing, the Picked For You recommendation engine and Post Voices, a tool for discovering commentary. Each draws from current and archived reporting produced by the two newsrooms. The system does not generate original stories or make editorial decisions.

The system runs on Google Cloud, using its Gemini Enterprise Agent Platform and Gemini models. The Post sets the editorial rules and controls the product, while Google provides the infrastructure, AI models, search and recommendation technology, Novicio said. New articles are indexed and made searchable within seconds of publication, allowing the apps to offer recommendations based on the latest coverage.

Google Cloud already works with publishers on backend infrastructure, but this marks the first time it has powered an AI chatbot for a news partner in North America, according to Michael Clark, president of Google Cloud North America. The distinction is significant. Google’s Gemini Enterprise platform is moving deeper into newsrooms, providing technology that publishers can build directly into their products rather than simply sending readers their way.

Novicio drew a clear line around the relationship. It is a standard enterprise cloud deal, she said, not a Google News partnership or a content-licensing agreement. The Post pays Google for the technology, and the arrangement is separate from Google Search. The Post’s archive, which powers Hamilton, is also not being used to train Google’s general-purpose AI models.

That distinction reflects how News Corp., the Post’s parent company, has approached AI. It has signed licensing deals with OpenAI and Meta while suing Perplexity over what it says is the unauthorized use of its journalism. Paying for cloud technology to build its own AI products offers a third approach.

The launch comes as publishers look for new ways to hold onto readers while AI-driven changes cut into search and social referral traffic. Some news organizations are responding by building chat and search tools designed to keep readers on platforms they control. The Washington Post launched an AI chatbot in 2024, Time introduced one with its Person of the Year coverage, and Gannett and The Independent have worked with Taboola on generative search tools built around their own archives.

For now, Hamilton has the advantage of living inside the Post’s apps. The harder test will come on the open web, where readers have plenty of other places to ask questions — and little reason to stay unless Hamilton gives them one.

The post New York Post launches Hamilton AI chatbot in effort to retain readers appeared first on The Media Copilot.

At last week’s Black Hat security conference in Las Vegas, OpenAI disclosed new details about a hacking incident involving its own rogue AI agents. Reporters raced to publish. WIRED was among the fastest, but an outlet called RuntimeWire beat it by more than three hours, according to WIRED.

RuntimeWire had no reporter at the conference. In fact, it has no reporters at all.

The site is an AI operation run by Austin entrepreneur Ryan Merket. After spotting an OpenAI executive posting about the talk on X, Merket fed a transcript of the livestream to his AI agents while the session was still underway. RuntimeWire published about six minutes later.

Most days, Merket is even less involved. His AI tools find stories, then draft, edit, fact-check, illustrate and promote them. If an agent determines that a story poses little legal risk, it can publish without his review. Some stories are also translated, turned into a daily podcast or read by synthetic voices.

Since May, RuntimeWire has published nearly 2,000 stories by scanning court databases, forums, company filings and social media for tech news. The trade-off, however, is quality. Its OpenAI story included a typo in the subhead and muddled key details. The operation also costs about $100 a day. Merket told WIRED he managed it by iMessage from Big Bend National Park, where the site published more than 80 articles in a week.

RuntimeWire is not alone. Dakota Carrasco, a BlackRock analyst, runs an agentic newsroom called The Dissent in his spare time. Its invented reporter personas cover subjects including San Francisco City Hall and Giants games. Carrasco remains anonymous behind the bots and acknowledges that the articles do not yet consistently link to their sources.

The risks are significant. Merket is relying on machines to decide what is true, what is newsworthy and what could create legal problems. Northwestern professor Nicholas Diakopoulos, who leads the university’s Computational Journalism Lab, describes the current moment as experimental and questions how much demand exists for bot-written news sites. He is also skeptical that mainstream journalists would surrender control over the wording and framing of their work.

But research from his lab points to one potential advantage. In a forthcoming paper, Diakopoulos and a colleague found that chatbots including ChatGPT and Claude surfaced AI-written sources 16 percent of the time across four topics. That could create an audience for AI newsrooms if assistants increasingly direct users to synthetic articles.

For publishers, the immediate challenge may be less about quality than cost and speed. An operation running for about $100 a day can compete with established tech outlets on live events and data-driven scoops — precisely the kinds of routine coverage that already strain newsroom budgets. As we have tracked at The Media Copilot, established publishers from Business Insider to The New York Times are bringing AI into their workflows. Merket is taking the idea further by using AI to break news, not simply repackage it.

Pete Pachal, founder and CEO of The Media Copilot, sees a limit. Building relationships with sources, he says, will remain a human job. But for covering product launches or combing through large datasets, he sees agentic newsrooms as a natural evolution of the technology.

Merket says RuntimeWire follows journalistic standards, including contacting subjects for comment and publishing corrections. So far, it has issued three. But he has also retracted accurate scoops about startups when their founders asked him to, calling it a favor from one founder to another.

That may be where the difference between running an AI news operation and running a newsroom becomes clearest. A reporter’s job is not only to get the story right and publish it quickly. Sometimes it is also to publish something a source would rather keep quiet.

The post Inside the AI newsroom that beat WIRED to the punch appeared first on The Media Copilot.

Runtime Wire has published 1,627 articles since launching in May, after scanning 71,796 potential stories and publishing just 2.3% of them, according to a profile in Talking Biz News.

The site covers the AI economy, from funding rounds and model launches to developer tools, infrastructure, executive moves and court filings. Its stories have drawn 205,000 pageviews.

And the newsroom has exactly one person on staff.

“I’m the founder and the only human,” said Ryan Merket. “The newsroom itself is software, a pipeline of AI systems that scans sources, researches, writes, edits, fact-checks, and produces the video and audio around the clock.”

Merket describes his job as part editor-in-chief, part engineer. He sets editorial standards, reviews coverage each day and builds the systems that handle the volume. He previously served as CTO at Microsoft for Startups and worked at Reddit and Amazon Web Services.

The pipeline monitors about 50 sources, including vendor blogs, changelogs, research feeds, social media and tips. Off-topic material, such as sports or crypto price chatter, is filtered out before scoring.

The remaining stories are evaluated by an AI curator against a written editorial standard: Is the story new? Does it matter to people building with AI? Does it come from a live primary source? Has Runtime Wire already covered the underlying event?

Two additional layers check for duplicate coverage and re-reports.

What happens next shows where the automation actually sits. A research step fetches and reads primary sources, including attached images, which are transcribed so image-only claims can be checked. Merket says the research brief is anchored to the current date, helping prevent old news from being presented as breaking.

A writing model with live web search drafts the headline, summary and story. An editor model then reviews the piece for depth and accuracy and can send it back for more research. A separate fact-checking pass tests specific claims against the live web.

For stories Merket assigns himself, he serves as the editor and reviews them before publication. In the automated lane, no human reviews the copy before it goes live.

Runtime Wire has not generated revenue yet. Merket says its monetization infrastructure is already live, including sponsored placements in code editors such as VS Code and Claude Code, a jobs board and site sponsorships.

For now, he is running ads against house inventory to tune pacing, quality scoring and bot filtering before selling placements more broadly. The strategy is B2B-first, targeting developer-tools advertisers while keeping the site free for readers.

For newsrooms, the more revealing number may be the 2.3% publish rate. Runtime Wire is not betting on volume for its own sake. Its pitch is that most of what it scans never gets published.

That sets it apart from the AI content farms that have drawn scrutiny from watchdogs such as NewsGuard, which has tracked hundreds of unreliable AI-generated news sites. Whether a one-person newsroom can maintain that level of editorial filtering as it scales remains the bigger question.

The post Runtime Wire runs an AI newsroom with one human and 1,627 published stories appeared first on The Media Copilot.

A translation of a mass shooter’s journal written in Faux Cyrillic. A scraper that pulled every public meeting minute from a Texas county website. Tens of thousands of leaked emails from a Chinese surveillance firm, made searchable with a large language model.

Each of those reporting efforts won or was a finalist for a Pulitzer Prize this year. And each relied, in some way, on AI.

Nieman Lab’s Andrew Deck reports that eight Pulitzer honorees disclosed AI use to the judging committee in 2026: five winners and three finalists. That’s the most since the Pulitzer Prizes began requiring AI disclosures in 2024—and it points to a shift in how journalists are using the technology.

In the two years before that, reporters mostly used older forms of machine learning — embedding models for data visualization, for example, or pattern-recognition tools to analyze satellite imagery. This year, commercial large language models did more of the heavy lifting, especially when reporters faced mountains of documents.

The approach was similar across this year’s Pulitzer winners. At The Wall Street Journal, computational journalist John West and his colleagues built a custom scraper to collect records from Kerr County, Texas, after deadly summer floods. They then used an internal tool called WSJPT to summarize every page and flag references to previous flooding events. Reporters still read each flagged section themselves.

“We aren’t obviating the need for human investigation of a pile of documents,” West said. “Instead, we’re trying to sort the pile so the most relevant stuff is right at the top.”

At The Minnesota Star Tribune, engineer Dana Chiueh used an enterprise ChatGPT account to analyze screenshots of a Minneapolis church shooter’s journal. The model identified the writing as Faux Cyrillic and produced a first-pass translation of more than 600,000 words.

The team then used Google’s NotebookLM to identify recurring themes and had two Russian-language academics at St. Olaf College verify key passages. Their review caught several errors, including one that mischaracterized the shooter’s motivation.

The reporting won the Pulitzer Prize for Breaking News.

The Associated Press used large language models to make tens of thousands of leaked documents searchable for its Pulitzer-winning investigation into American technology companies’ role in China’s surveillance state.

But reporters did not simply trust what the models surfaced. They manually reviewed documents flagged by AI, independently checked the accuracy of AI-generated summaries and did not quote from those summaries, said AP investigative journalist Garance Burke.

The New York Times flipped the usual workflow. For its investigation into the Securities and Exchange Commission’s retreat from crypto enforcement under the second Trump administration, reporters manually read and classified more than 10,000 documents. They then used OpenAI’s GPT-5 to conduct a second pass, flagging discrepancies for reporters to review.

In this case, AI was used to check the humans — not the other way around.

For newsrooms, perhaps the more revealing detail is what readers never saw.

The Wall Street Journal did not disclose its use of AI in the flood investigation at publication. West said the tools functioned essentially as a more sophisticated search system.

That gap between disclosing AI use to a prize committee and disclosing it to the public is at the center of an increasingly important debate over newsroom standards. It’s a tension that has surfaced repeatedly as news organizations figure out when AI use is significant enough to tell readers about.

Pulitzer administrator Marjorie Miller said the industry now has a clearer understanding of where AI can be used appropriately, such as data collection and analysis, and where its use raises more questions, including writing and editing.

Next year, after controversies over AI-generated text in prize-winning literary works, the Pulitzers will add an AI disclosure question to their book entry forms.

“AI is here to stay,” Miller said. The committee, she added, will continue asking entrants to demonstrate that a human produced the work.

The post Eight Pulitzer awardees disclosed AI use this year, a new record appeared first on The Media Copilot.

Starting August 2, the European Union will require companies to disclose when images, audio and text have been artificially generated or manipulated to look authentic, as the Guardian reports. The rules are part of the EU’s AI Act, and they could have a particular impact on publishers and other organizations that are increasingly using AI to produce content.

Under the new requirements, people must be told when they are interacting with an AI chatbot or viewing manipulated media. Synthetic content must carry a visible label as well as a digital watermark identifying it as AI-generated.

The rules apply to new AI systems entering the European market beginning August 2. Existing systems will have an additional four months to comply. And for publishers, there is another important provision: Text about matters of public interest must be labeled as AI-generated when there has been no human editorial oversight.

Companies that fail to comply could face penalties of up to €15 million.

There are some exceptions. Personal content is exempt, as are works that are clearly artistic, satirical or fictional. Labeling AI-generated content created before the rules took effect is encouraged, but not required.

The European Union says the goal is to make it easier for people to distinguish authentic material from synthetic content. The rules follow the June 10 publication of a voluntary Code of Practice on Transparency of AI-Generated Content, which lays out how AI providers and deployers can comply with the AI Act’s labeling requirements.

Sergey Lagodinsky, a Green member of the European Parliament, told the Guardian that the rules are about more than protecting the public.

“It is a matter not only of customer protection, it’s also a matter of democracy protection,” he said. “Making transparent this information is something which we need to preserve our democracy and the authenticity of facts online.”

But some technology companies and industry groups say the rules have become too broad.

The Computer and Communications Industry Association argues that guidelines issued in July expanded the definition of a deepfake beyond what lawmakers intended when they passed the AI Act in 2024.

“The label was meant to flag deceptive content,” said Boniface de Champris, CCIA Europe’s AI policy lead. “That distinction has been removed, so now almost everything gets labeled — a landscape in an advert is the same as a manipulated political speech… That doesn’t protect anyone, it just burdens commercial activity across the board.”

De Champris says the biggest changes may not be obvious on social media, where platforms already label much AI-generated material. He expects the rules to have more consequences for advertising, film and publishing — industries that are using AI at scale but do not always tell audiences when they do.

That could put news organizations and other content producers in the EU in a new position. AI-generated images, synthetic voiceovers and machine-written text about matters of public interest may now require disclosure under the rules. What was once a question of editorial policy could become a matter of compliance.

Some major technology companies have already taken steps in that direction. TikTok says it requires creators to label realistic AI-generated media and has labeled more than 3 billion pieces of content. Google says its SynthID watermarking system has been used on more than 100 billion images. And more than 180 organizations, including Google and Meta, have signed a voluntary code of practice supporting the EU’s transparency requirements.

Henna Virkkunen, the European Commission’s executive vice president for tech policy, says the broader rules are meant to strike a balance. She described AI as a “transformative technology” capable of bringing “extraordinary benefits,” while warning that the most advanced systems could create risks “on an entirely new scale.”

For publishers and other content producers, the next phase will be figuring out exactly where those lines fall — and what AI disclosure looks like when it becomes a legal requirement rather than an editorial choice.

The post EU makes AI labels mandatory on authentic-looking images, audio and text appeared first on The Media Copilot.

James Allen of Billpin was quoted as a finance expert by outlets including CBS News, the New York Post and Business Insider. Qwoted later barred his account after flagging repeated AI-generated submissions and finding that he had failed its verification process, according to Press Gazette’s investigation.

The case is a stark warning about the source pipeline behind everyday newsroom copy. Qwoted, a U.S. platform that connects journalists with sources, says it has banned hundreds of agencies, executives and experts for conduct ranging from fabricated identities to AI-generated pitches and false claims of expertise.

Billpin’s website is now blank, and Allen’s LinkedIn profile has disappeared, Press Gazette reported. The publication said Allen did not respond to requests to verify his identity. Press Gazette concluded the operation appeared designed to win media mentions that could boost search rankings, or backlinks.

A quote in a respected publication can generate a link that helps an otherwise obscure website gain visibility in search results. That creates an incentive to manufacture expert identities that appear credible enough to earn media coverage.

Qwoted CEO Dan Simon said generative AI has dramatically accelerated an already difficult problem. “AI has just poured petrol on the disinformation fire,” he told Press Gazette.

The platform says it now handles more than 15 violations a day and blocks about 50 prospective clients from registering each day. Shelby Bridges, Qwoted’s head of user experience, said the volume of low-quality pitches began rising rapidly after ChatGPT’s public release in November 2022.

Simon described a source ecosystem that increasingly blurs the line between legitimate expertise and fabrication. “You’ve got real agencies representing fake experts with fake expertise. You’ve got real agencies with fake experts and real agencies with real experts with fake expertise, and fake agencies with fake experts with fake expertise.”

One recurring pattern is a purported expert responding to queries on subjects with no obvious connection to their background. Press Gazette identified examples of sources offering commentary on finance, cleaning, human resources, pet sitting and mental health.

Qwoted combines human review with signals such as unusually high pitching volumes. It also uses Pangram, an AI-content detection service, to flag potential abuse. But AI detection is only one signal. A model score cannot establish that a person is fictional, that a quote is false or that an expert lacks relevant experience. Qwoted’s reported checks therefore matter more than a detector’s output alone.

Simon said some reporters have told him they do not object to real experts using ChatGPT to polish their responses. The bigger risk comes when AI helps conceal a fabricated identity or invented credentials, allowing a bogus source to slip into print before anyone questions its legitimacy. Platforms can suspend accounts after the fact, but the final responsibility still rests with newsrooms.

The post Qwoted bans AI-generated expert profiles as PR spam floods reporters’ inboxes appeared first on The Media Copilot.

Substack readers can now run a check on any post longer than 100 words to see how much of it an AI model likely wrote. The company rolled out the feature on July 21 in partnership with the detection firm Pangram, with the scan available on web and iOS and Android support promised later.

The tool works only on content published starting on July 21. Alongside it, Substack is giving writers a dedicated space to state whether and how they used AI in a given piece. The idea, per the company, is to set reader expectations rather than police anyone.

No AI detector is reliable enough to call a piece machine-written with certainty. Pangram uses a deep learning classifier trained on around a million documents rather than the older perplexity-and-burstiness heuristics, and it reports strong internal accuracy numbers. But The Atlantic examined how well Pangram actually performs and found the picture more complicated than the marketing suggests. False positives are a known problem across the category, and lightly edited or paraphrased AI text is far harder to catch than raw model output.

Substack did acknowledge those limits in its announcement post. It also hinted at further features tied to AI content and reader preferences without saying what they are.

Its acknowlegdment matters because a wrong flag carries real cost. A human writer who gets tagged as an AI user has little recourse in the moment, and the accusation sticks. Non-native English writers have historically drawn more false positives from detection systems, which raises fairness questions for a platform with a global base of newsletter writers. Pairing the automated score with a voluntary disclosure field is Substack’s hedge against that, letting writers put their own account on the record next to the machine’s.

For newsrooms and independent publishers, the move is a preview of how disclosure norms may harden across platforms. If readers start expecting an AI provenance label on every post, editors will face pressure to document their own use of the technology and to decide where the line sits between an AI writing assistant and AI ghostwriting. Publishers who have built AI into their workflows, a shift The Media Copilot has tracked closely, will want a clear internal policy before a third-party score answers the question for them.

The real test isn’t whether the detector works. It’s whether readers trust a score that Substack itself acknowledges can be wrong. The company is betting users will treat AI detection as one signal among many, not a definitive verdict.

The post Substack adds AI detection to its posts through a new Pangram partnership appeared first on The Media Copilot.