Recommended Articles
Deals & Exits

The Publisher–AI Licensing Wave Has a Template Now: Inside the Reach–Google Deal

By Angeled · October 16, 2026 · Deals & Exits
The Publisher–AI Licensing Wave Has a Template Now: Inside the Reach–Google Deal
The Publisher–AI Licensing Wave Has a Template Now: Inside the Reach–Google Deal

On October 16, 2026, Reach — the UK publisher behind the Daily Express, Daily Star, and OK! Magazine — agreed a two-year paid deal with Google granting expanded rights to use its UK and Ireland news content for Google's news AI pilot. More than 200 publications are now in that pilot. In the same breath, Reach sent OpenAI a cease-and-desist, citing more than 100 million scrapes of its content in six months. Pay one lab, sue the other: this is the new publisher playbook, and it now has a template.

What the deal actually covers

The agreement gives Google expanded rights to Reach's news content specifically for a news AI pilot — meaning the content can be used to ground, train, or retrieve against in AI-generated news experiences. It's a two-year, paid arrangement: finite term, defined scope, real money changing hands. That structure matters more than the headline, because it establishes the commercial shape future deals will copy: term-limited licenses for defined AI use cases, not blanket buyouts of a publisher's archive.

The emerging licensing stack

Three models are crystallizing across the publisher–AI market:

1. Lump-sum licensing deals. A publisher grants defined AI rights for a fixed fee and term — the Reach–Google shape. Clean, predictable, and increasingly standard for large content owners with leverage.

2. Per-use or "pay-per-value" pilots. Instead of a flat fee, compensation tracks actual usage — how often content is retrieved, cited, or surfaced in AI answers. Google's pilot framing points this way: pay tied to measurable value extraction, which appeals to smaller publishers who can't command lump sums.

3. Opt-out and conduct regimes. Regulation is becoming the backstop. The UK Competition and Markets Authority's June 2026 search conduct requirements for Google created a formal framework around how dominant platforms use publisher content. Opt-out mechanisms — where publishers can withhold content from AI training or AI Overviews — are shifting from voluntary courtesy to enforceable right.

How publishers should value their archives

Publishers historically valued content as advertising inventory. AI licensing demands a new valuation lens with two distinct components:

Training-data value is about corpus scale and quality — the raw material for model improvement. It's a bulk commodity where leverage comes from uniqueness (archives competitors can't replicate) and cleanliness (rights-cleared, structured content).

RAG/grounding value — retrieval-augmented generation — is about freshness and authority. When an AI answers "what happened today," it needs today's news from a trusted source. That's recurring, high-margin value, because it's consumed continuously, not once. Reach's 100-million-scrapes figure is effectively an invoice: it quantifies the uncompensated grounding value OpenAI extracted.

Smart publishers will price these separately and negotiate them separately.

What this means for AI startups building on web content

The era of "scrape first, apologize later" is ending. For AI startups, licensing is becoming a real cost line in the P&L — alongside compute and talent. Three implications:

  • Budget for content like you budget for GPUs. If your product retrieves or trains on publisher content, model licensing costs into unit economics from day one.
  • License-first is a moat. Startups that secure clean, licensed content pipelines will have a defensibility advantage over competitors built on scraped data facing cease-and-desists.
  • Watch the regulatory floor. CMA-style conduct requirements are spreading. Building on the assumption that today's scraping norms persist is a legal risk, not a strategy.

Why it matters for founders and investors

Content licensing is becoming a genuine market — with templates, price discovery, and regulatory scaffolding. For founders, that means two opportunities: build the licensing infrastructure (rights management, usage metering, payment rails for the pay-per-value era), or build AI products with licensed data as a competitive moat. For investors, publisher–AI deals are a new diligence item: ask every AI startup where its training and grounding data comes from, and what it costs.

Angeled.com connects founders with investors.

Sources: Digiday, Oct 16, 2026.

Read Next