The agent web has no sheriff
The agent web has no sheriff
Nobody set the rules for AI agents, so Meta did. Plus Axios feeds the bots, Google's AI ads go global, and a free scraper hunt.
Meta spent the past two weeks fighting with publishers over a terms-of-service change that, read one way, gave it permission to crawl the website of any business with a Meta account. The company has since trimmed the language twice and pushed the start date indefinitely.
The whole incident is instructive, and it highlights that nobody’s written the rules for how AI agents use the open web. Not lawmakers, not the courts, not any standards body. Into that vacuum, the companies building the agents keep writing the rules themselves. Meta is just the most recent example, and the one with the least goodwill to spend.
I dive into what it means this week, but before that a quick heads-up that I’ll be at Adobe MAX in Miami Beach next week. If you see me, come say hi.
Train the whole team, not just the early adopters
Most of what I do outside this newsletter is custom AI training for media, PR and communications teams. Sessions run weekly, biweekly or monthly, built around your tools, your workflows and the work your people are already doing. Recent sessions include:
AI Agents: From Answers to Action. Where each level of autonomy fits, what oversight looks like, and how to move from asking AI a question to handing it a task.
Generative Engine Optimization. How to position your organization or your clients to show up in AI-generated answers.
Custom AI assistants. Build GPTs, Gems and Copilot agents that know your clients and write in your voice.
AI governance and policy. Data handling, guardrails and rules a team will actually follow.
Vibe coding. Turn flat spreadsheets into interactive dashboards with no programming background.
Every session comes with recordings, prompt libraries and takeaway materials, plus a monthly briefing on what changed. We take on a limited number of partnerships each quarter.
Meta's scraping fight shows nobody has set the rules for AI agents
I wrote this week about who actually gets to set the rules for AI agents, and what the Meta fight says about the answer. It turned out to be less a story about scraping than about who holds the leverage when there is no rulebook.
Start with what Meta was probably doing. I doubt the company reversed its position on the biggest legal question in AI and media and then buried it in a terms update. Meta says the crawling language came from folding its Ad Creative Generative AI Terms into its main commercial terms, and tools that build ads for a business do need to read that business’s website. As Meta leans harder into agents like Muse, its bots will touch the open web more, not less.
That explains the origin. It does not explain all of the wording. One provision lets Meta create “test credentials,” which Cornell law professor James Grimmelmann said would let it sign up as a fake user to get past registration walls. There is no opt-out short of abandoning Meta’s business tools. And the change landed just as Meta started charging some accounts, through its new Meta One subscription, to post links to the same content its bots could crawl for free.
The same dynamic showed up somewhere else recently. Axios reported that banks were seeing a rise in “unauthorized” AI agents reaching customer accounts, which sounds alarming until you notice that most of it is people using AI to get into their own accounts. The banks still have a real problem, mostly liability when something inevitably goes wrong. JPMorgan Chase CEO Jamie Dimon told analysts that letting in outside agents is “just too risky for clients and for JPMorgan,” and the bank now blocks them at the login page. Schwab is doing the same while steering clients toward its own assistant, which also happens to protect its cash business. Citi blocks third-party agents too. Among the agents getting shut out is Meta’s own Muse.
Put the two side by side and the shape is identical. A platform spotted a gap in how AI was being used and, because it held the leverage, it set the rules. That is what happens when nobody else has. Whoever carries the biggest gun writes them. With agents, we are all living in Tombstone.
Perplexity and Cloudflare ran the same script last year. Cloudflare accused Perplexity of using stealth crawlers to slip past sites’ bot-blocking preferences, and Perplexity argued its agents were legitimate proxies for real users, entitled to the same access. Dealing with sites one at a time, Perplexity had the upper hand. Once Cloudflare got involved, as the gatekeeper for a large share of the web, the math changed. It delisted Perplexity as a verified bot and handed every customer a way to block its stealth crawlers. Perplexity has since appointed itself enforcer elsewhere, blocking ads that Time served to AI agents as “deceptive” without saying how it defines the word.
The townspeople are starting to band together. SPUR, the Standards for Publisher Usage Rights coalition, is now more than 80 publishers working out how content gets tracked and valued when AI systems use it. In Congress, the bipartisan Stealth Bot Prohibition Act would require crawlers to identify themselves and their purpose, and several states including New York have passed their own versions. Courts and regulators are moving too. A Norwegian court has temporarily blocked Meta from enforcing the terms there, and the U.K.’s Competition and Markets Authority has opened a formal investigation into whether they are unfair.
My own read is that the lesson for Meta, and for anyone else using leverage to set rules, is that overreach costs you. Legal departments will always grab the widest set of rights available, but with no rules governing those rights, that breeds suspicion. The deeper problem is that the rules are Meta’s to write at all. Leverage decides what an agent may do with what it collects: Meta through its terms, JPMorgan through its login page, Perplexity through its index. I get into what WAN-IFRA wants Meta to put in writing, why an open-ended pause changes nothing, and what all of this means for anyone running a newsroom or a brand site, in the full column. Until somebody builds a framework everyone has to follow, you cannot blame publishers for circling the wagons every time a company pins on a badge and declares itself the new sheriff in town.
The Chatbox
All the AI news that matters to media*
90% of your page is junk to a bot
An AI agent reading a web page burns most of its effort on markup meant for browsers. “If you are an agent and you’re reading an HTML page, 90% of the content is junk,” People Inc. chief innovation officer Jon Roberts said at Digiday’s Media Strategies London event, as Digiday reported. So the publisher is building custom APIs that serve article text in Markdown with the author and publication date attached, what Roberts called trust signals that matter. People Inc. started by letting partners crawl its pages, moved to Amazon S3 buckets for more control, and is now dating the content its own AI tools consume before extending the method to outside partners. Roberts said the company is not trying to sell ads to agents. Digiday did not name the AI companies involved or report pricing. (AI-assisted)
Your next subscriber might be an agent
The reader a publisher optimizes for may soon be software. WIRED reports that Axios and Every are both shipping products aimed at AI agents rather than people. Axios Direct, announced at the Digiday Publishing Summit in September, is a set of standardized structured feeds that chief revenue officer Jacquelyn Cameron compares to a Bloomberg terminal rather than a collection of stories. The first feed launched this fall for Axios Pro Deals subscribers, who pay $2,499 a year, and gives their agents access to the company’s M&A reporting. Every went the other direction, embedding prompts and “Read With Claude” buttons in articles so a reader’s own agent can rebuild the reporting as a summary, podcast or video. Neither publisher has disclosed what any of it earns. (AI-assisted)
One bot looks normal until you compare notes
A visitor who reads 400 Telegraph articles could be an unusually devoted subscriber. The same visitor working through hundreds more at Newsquest and The Guardian “almost certainly is a bot,” NLA Media Access chief executive Henry Jones told Press Gazette. That cross-publisher pattern is the premise of NLA’s Collective Intelligence Advisory Service, a free offering that analyzes participating publishers’ website data to flag likely unauthorized scraping. Traffic that passes as ordinary on one site stands out once several outlets compare notes, which is the gap single-site detection leaves open. A pilot with Newsquest and three national publishers, including The Telegraph, surfaced companies in Eastern Europe and Asia-Pacific taking content. Some bought licenses once NLA got in touch. (AI-assisted)
Google takes AI Mode ads worldwide
Ads inside Google’s AI Mode are heading overseas for the first time, chief business officer Philipp Schindler told investors on Alphabet’s third-quarter earnings call, describing it as taking the monetization footprint global. TechCrunch reported the expansion next to the figure that explains the urgency. Search revenue grew 12% year over year, down from 17% in the second quarter and 19% in the first. Google is also widening multi-turn ads and other AI Mode formats in the U.S. Schindler said sponsored product listings inside AI Mode draw click-through rates comparable with organic listings, and that research shows users find them relevant, though TechCrunch noted Google supplied no numbers behind either claim. The ad inventory is moving to where the answers already are. (AI-assisted)
Queries are up and referrals are down
How do search queries hit a five-year high while publisher referrals keep sliding? That is the question hanging over Alphabet’s third-quarter results, reported by The Verge, in which CEO Sundar Pichai said the company’s AI investments drove the fastest year-over-year growth in Search queries it has recorded in five years, commercial queries included. Pichai gave no growth rate and no total query volume. Alphabet reported $127.6 billion in revenue, up 25%, with Cloud growing far faster than Search. The earnings materials do not break out which products or AI features produced the extra queries, and they do not address whether AI answers are prompting follow-up searches or simply absorbing the clicks that used to leave Google. (AI-assisted)
*AI-assisted news items are drafted with AI and then carefully edited by Media Copilot editors.
The Media Copilot is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Who We Are
The Media Copilot covers how AI is changing the media. We also train journalists and PR professionals on how to use the latest AI tools in their work. Check out our courses:
AI for PR and Communications Professionals: From deep research for list creation to vibecoding with campaign data, we teach the new PR tool stack. Offered in partnership with The Upgrade.
AI for Journalists: Master the AI skills that let you report faster, write smarter, and protect credibility. Offered in partnership with The Upgrade.
AI Quick Start: Just getting started on your AI journey? We’ll take you from trial and error in ChatGPT to deliberate prompting for specific workflows in media and PR. We’ll also walk you through building your first AI assistant.
In charge of a larger team? We offer customized training and consulting packages as well. Email us to set up a call.
Podcast and YouTube Channel
Every week, we have in-depth conversations about AI in the media with industry thought leaders, executives, innovators, and fellow journalists. You can subscribe on Apple Podcasts or on your favorite podcast platform.
The podcast is also available at The Media Copilot YouTube Channel.
If you have feedback on any of our content or services, don’t hesitate to reach out. You can simply reply to this email or reach out at [email protected].
You’re currently a free subscriber to The Media Copilot. For the full experience and access to more (human-authored) content, upgrade your subscription.
© 2026 AnyWho Media LLC