How Will Publishers Generate Revenue in the Age of AI Agents? 

As AI agents become the new gateways to information, publishers face a pressing dilemma: how do you make money when readers no longer land on your site? 

In this piece, we look at the three main revenue paths emerging in the age of AI: subscriptions, licensing deals, and pay-per-scrape, and what they could mean for publishers of different sizes. 

Subscriptions 

According to the 2025 Reuters Institute Digital News Report, only 18% of people pay for online news. Yet subscriptions remain the most durable source of direct revenue for publishers.  

The logic behind encouraging subscriptions is simple: AI can provide quick answers, but readers still turn to trusted brands for depth, expertise, and identity. Thomas Baekdal calls actively seeking out a publisher rather than relying on intermediaries “direct intent”.  

For the subscription model to work, Alan Hunter, Co-Founder of media consultancy group HBM Advisory, argues that publishers will need to grow increasingly niche. They’ll need to build authority and expertise that machines can’t replicate.  

Similarly, as Tom McCave, past VP of Performance Marketing at The Economist, presented at last year’s Digital Growth Summit, the role of branding for publishers is more important than ever, as it gives a “competitive advantage”. A strong brand gives audiences a reason to connect with a news outlet. When the brand clearly reflects their expectations and values, casual readers can turn into loyal supporters—and even advocates. As a recent 404 Media article highlighted, with AI, we’re witnessing a proliferation of low-quality, rehashed content. Building a sense of community, loyalty, and trust through branding is a key solution for those looking for someone to help curate the deluge of content. Building brand trust here is therefore key, and something that users would pay for.  

Still, subscriptions are no cure-all. Subscription fatigue means most readers will only pay for one or two outlets, leaving smaller or generalist publishers exposed. And as AI agents deliver ever more complete answers, publishers without strong brand equity may struggle to convert readers. As Hunter puts it: “Why go to a publication that can give you a surface-level view on a topic when that news is much easier to access via an AI interface—where you can also ask follow-up questions?” 

Licensing Deals 

Licensing is also a way for publishers to monetize in the AI era. AI platforms need high-quality, trustworthy data to answer queries, so instead of scraping, they are paying for access. Examples include:  

• The Associated Press, Hearst, Le Monde, The Financial Time’s deals with OpenAI 

• The New York Times licensing to Amazon AI 

• FT, Reuters, Axel Springer, Hearst, USA Today Network’s deals with Microsoft 

• Time, Der Spiegel, Fortune, Entrepreneur, The Texas Tribune’s deals with Perplexity 

Who this model can work for is up for debate. In principle, this model works best for large, established publishers with strong brands and archives of authoritative content. They can get paid for journalism already produced, without added distribution costs.   

Smaller and local outlets risk exclusion. They often lack leverage to secure deals, even while their content is still used. As a 404 Media article reported:  

“Upneet Grover, founder of LH2 Holdings, which owns several smaller blogs, suggested that ‘a lot of these licensing revenues are not moving the needle, at least from the deals we’ve seen, but there’s this larger threat of more referral traffic being taken away from news publishers [by AI].’”  

Licensing also carries the risk of disintermediation: if AI agents answer questions directly, publishers lose audience relationships, weakening brand equity over time. 

Pay-per-Scrape 

Pay-per-scrape (or pay-per-crawl) charges AI platforms each time their bots access content. Instead of blanket permission, publishers can block, allow, or set a price per request. The idea is appealing: if journalism powers AI, publishers should be compensated for every use. 

Cloudflare recently launched this at scale, blocking AI crawlers by default and letting publishers charge for access. Tollbit also offers a similar service. In theory, even smaller sites gain leverage, with billing and enforcement handled by the intermediary servicer. 

This could work for major newswires, national dailies, and specialized outlets whose content AI frequently seeks out. The idea is that even modest per-query fees could add up at scale. It may also benefit mid-tier publishers with valuable archives but less bargaining power for big licensing deals. 

Yet challenges remain. As a recent Tech Radar post highlighted, pricing is tricky—if you pay per scrape, no matter the content, a flat fee undervalues investigative work while overvaluing short updates. Also, enforcement is imperfect: bad actors can spoof bots or scrape indirectly from public datasets. And if AI companies can avoid scraping by relying on licensed datasets from large publishers, smaller outlets again lose leverage.  

So where does that leave us? 

What’s clear is that publishers cannot simply wait for AI platforms to decide their fate. No single revenue model will be enough on its own. They will need to experiment across multiple models, push collectively for fairer terms, and double down on the one advantage machines cannot replicate: trust, identity, and human expertise. In the age of AI agents, revenue may come through new channels, but the underlying rule stays the same: publishers that can build distinct value for their audiences will be the ones that survive.