Autopilot publishing is the shorthand for a new era in which large AI models automatically produce answers and summaries drawing on newsrooms' reporting. The question on editors' and managers' minds is simple: will AI pay publishers enough to sustain journalism, or will it simply ride on the back of creators' work?
The record through 2025 shows payments are already happening , from bespoke, multi‑year deals to per‑use revenue shares , but the sums, structures and transparency vary widely. From OpenAI's landmark News Corp tie‑up to Perplexity's per‑article revenue shares and the Anthropic settlement precedent, the market is evolving under legal, commercial and regulatory pressure.
Deals on the books: who has signed and what they say
Major platform‑publisher agreements have moved from hypothetical to real. OpenAI inked a 'landmark, multi‑year' partnership with News Corp in May 2024 that gives access to current and archived content from titles including the Wall Street Journal, New York Post and The Sun, with News Corp saying the deal 'could be worth more than $250 million over five years.' The Financial Times struck its own April 2024 deal with OpenAI to let the company use FT content for model training and to surface attributed summaries in ChatGPT, with FT CEO John Ridding saying 'it’s right, of course, that AI platforms pay publishers for the use of their material.'
Other big players followed different paths: Microsoft said in October 2024 that its Copilot Daily product would pay participating publishers such as Reuters, Axel Springer and the FT for content surfaced by the product, while Meta reached a multi‑year deal with Reuters in October 2024 to provide real‑time reporting to its AI chatbot. The Associated Press was among the earliest news organizations to license archive material to OpenAI (January 2023), illustrating that this trend has been building for years.
Yet deals are not uniform. Some explicitly allow training on publisher content, some only allow surfacing and attribution, and many disclose little about payment mechanics. Even when companies tout partnerships as historic, the lack of publicly disclosed terms has left publishers and observers asking how much journalism is really worth in the AI economy.
How payments are structured: training rights, surfacing rights and revenue shares
Industry observers make an important distinction between at least three compensation models. First, licensing that explicitly permits use of content for model training; second, arrangements that allow AI platforms to quote, summarize and link to articles in answers (surfacing); and third, revenue‑share or ad‑share models tied to clicks or subscription conversions. Each creates different legal and commercial outcomes for publishers.
Startups and some vendors have leaned into per‑use mechanics: Perplexity launched a Publishers Program that shares ad and subscription revenue with publishers, citing double‑digit percentages per article in some reports and earmarking a $42.5 million pool for partner compensation. Per‑article payments can be calculated per query and can even count multiple articles used in a single answer, giving publishers a direct tie between their content being surfaced and compensation.
By contrast, many of the biggest platform deals are bespoke, multi‑year agreements with undisclosed terms. Reporting indicates some of OpenAI's smaller, non‑publicized licensing checks were in the low‑millions (reports suggest roughly $1M, $5M in several cases), sums that publishers have criticized as inadequate given the scale of value extracted from journalism in AI systems.
Publisher perspectives: what newsrooms want and fear
Publishers argue they need recurring, meaningful revenue streams from AI platforms to offset continued newsroom losses. The sector suffered thousands of job cuts across 2023, 2025, and newsroom leaders frequently say that licensing and revenue‑share arrangements could provide a lifeline if they are fair and predictable. FT CEO John Ridding and others have been vocal that AI platforms should pay for the use of journalism.
But publishers also fear that opaque deals and small one‑off checks will undermine subscription economics. Complaints include inconsistent attribution and the quality of linkbacks, the absence of long‑term protections for subscription models, and a lack of transparency about how platforms value and reuse reporting. Trade groups and unions have repeatedly called for clearer, enforceable standards.
Those concerns shape bargaining positions: high‑profile lawsuits (for example, the New York Times suit allowed to proceed in March 2025) and author class actions increase publishers' leverage and the legal risk faced by AI companies. Publishers are using a mix of litigation, negotiation and public advocacy to push for better terms and clearer practices.
Legal and regulatory pressure: fines, lawsuits and settlements
Courts and regulators have turned into powerful drivers of payment. A French regulator fined Google €250 million in March 2024 for breaching commitments around publishers' rights, signaling that authorities will enforce news rights in the AI context. In the U.S., the New York Times' copyright suit against OpenAI and Microsoft survived a motion to dismiss in March 2025, keeping a key test of scraping and training liability alive.
Class actions have also produced major settlements. In September 2025 Anthropic agreed to pay $1.5 billion to settle an authors' class action alleging mass copying of books for training Claude, a precedent that increases the stakes for publishers seeking compensation. These legal pressures create financial and reputational incentives for AI firms to negotiate licences or otherwise compensate content creators.
At the same time, legal uncertainty remains. Courts will gradually refine whether and how scraping archives for training violates copyright, and regulatory bodies in different jurisdictions will set varying standards , a patchwork that both incentivizes deals and complicates global licensing strategies for platforms and publishers alike.
Startups and experiments: Perplexity and the mechanics of per‑use pay
Not all compensation strategies come from the biggest players. Perplexity's Publishers Program is an example of an experimental marketplace approach: after plagiarism accusations in 2024, Perplexity signed deals with outlets such as Time and Fortune, introduced Comet Plus (a $5/month tier), and set aside a $42.5M pool to compensate publishers. Reporting has highlighted caps and percentage ranges , in some cases up to roughly 25% per query of ad revenue , and even models offering publishers back most subscription revenue tied to user conversions.
Per‑article or per‑answer accounting offers publishers a clearer line of sight between their work and the cash they receive, but it also raises practical questions about measurement, fraud prevention and the administrative cost of many small payments. Startups may be more willing to experiment with granular revenue shares, while large platforms tend toward larger, negotiated payments that are easier to administer at scale.
These marketplace approaches test whether direct compensation tied to usage can be sustainable. If per‑use payments meaningfully reward the reporting that improves AI answers, they could form a scalable complement to larger licensing deals , provided measurement and transparency improve.
What comes next: standards, transparency and sustainable models
By late 2025 the pattern is mixed but trending toward payments. Large platforms have made line deals with major publishers, startups have piloted per‑use revenue shares, and legal/regulatory pressure has increased firms' incentives to compensate creators. Still, the scale and fairness of payments remain unsettled: reported OpenAI payments range from token low‑millions to the seven‑figure estimates in some deals, while News Corp suggested a deal possibly exceeding $250 million over five years.
Publishers and policymakers now face choices. They can push for enforceable frameworks that distinguish training rights from surfacing rights, standardize attribution and linkbacks, and mandate disclosure of payment terms. They can also press for market mechanisms that scale fair compensation, such as standardized per‑article reporting and independent auditing of usage claims.
For 'autopilot publishing' to be sustainable, the ecosystem needs transparent deals, realistic measurement and a mix of upfront licensing, revenue share and product features that drive direct subscriptions. Absent those changes, payments will remain uneven and journalism's dependence on unpredictable checks and ad shares will persist.
Ultimately, will AI pay publishers? The short answer is yes , payments are happening and will increase under legal, commercial and regulatory pressure. The longer answer is that paying is not the same as paying fairly or predictably: achieving that will require clearer standards, better measurement and continued pressure from publishers, courts and regulators to turn line deals into a stable economic base for journalism.
If you want primary documents, I can fetch the News Corp/OpenAI press release, the FT announcement, the Anthropic settlement filings, Perplexity's publisher terms or the NYT complaint and court orders to ground these trends in primary sources.
