Skip to content

By selinarussell

Anthropic’s 2026 Usage Policy Update

Share

On 8 October, Anthropic published its 2026 Usage Policy. Among the rules on weapons, surveillance and elections is one aimed squarely at the race to be recommended by AI.

The policy bans using Claude to plant content that misleads on where it came from, who wrote it or whether it is independent (or not). Its target is the sources that AI and traditional search engines draw their answers from.


Anthropic’s own example is a network of websites that pretend to be unconnected while all repeating the same claims.

The policy takes effect on 12 November, and it shows what at least one AI company now refuses to treat as a legitimate signal.

The rule sits in a new section called “Do Not Engage in Deceptive Campaigns or Artificial Activity”, which Anthropic’s announcement says covers deception whether it is political or commercial. Its September threat intelligence report found state media outlets, government propaganda offices and commercial firms using Claude to run fake accounts and fake news sites.

The same section also bans selling any of these tactics as a service.

Fakers Counterfeit AI’s Trust Signal


New content is published every second of the day. In our experience, most of it is self-serving, says nothing new and does nothing to advance its field, so it is simply noise for AI and traditional search engines to sift through.

AI pays more to do the sifting. In 2023, Alphabet’s chairman John Hennessy estimated that an exchange with an AI model costs around 10 times more than a standard keyword search on Google.

At that cost, AI can’t afford to weigh every page equally, so one way it cuts through the noise is to look for third-party endorsements, since a business that others vouch for is more likely to deserve a recommendation.

That signal is exactly what fake source networks set out to counterfeit. Building one takes money, which gave businesses with bigger budgets a way to look more widely endorsed than they are, and anyone using Claude is now banned from doing it.

When AI looks for who vouches for your business, what will it find?

The Fakers Have It Back to Front


Fakers manufacture endorsements for content that has nothing to endorse, when genuine endorsements come naturally to content worth citing.

Publish something that tells your industry something it didn’t know, and digital PR outreach to journalists, bloggers, influencers and Reddit communities becomes far more effective, because you are bringing something to the table. Every mention and backlink then has a legitimate reason to exist.

Having built a large language model ourselves, we know that when AI follows those mentions back to your content and finds high information gain, it recognises them as genuine signal, and everything checks out.

The order matters: high information gain content first, then digital PR.

If your content tells journalists nothing they didn’t already know, what are they supposed to write about?

Information Gain Wins Tier 1 Coverage Without Paying

Our focus is information gain content first, the kind of content that makes AI smarter and tells the world something it didn’t already know.

That makes our digital PR far easier, which is how we’ve earned our clients backlinks and mentions from tier 1 publishers including Entrepreneur Magazine, Forbes and HubSpot, among many others, without paying the publishers a penny.

The same approach works in discussion threads like Reddit, because content that adds something new gives Redditors something to contribute, and they credit you for it.

We guarantee to get your business visible in AI.

Want to know where you stand within your industry? Reply to this email and we’ll show you whether AI would cite your content today, and what it would take to get chosen.


Yours,

Andreas

Founder & CEO, Artios.io

PS If you know a business who’d like to win the search game, please make an introduction or direct them to artios.io/winning-the-ai-search-game

Andreas Voniatis is founder of Artios.io, author of Generative Engine Optimization with Python (O’Reilly, 2026), and built his first Large Language Model in 2018 – four years before ChatGPT launched.

Artios helps challenger B2B businesses get chosen by AI when buyers search there.