The other way to monetize AI summaries
As AI search rises, Perplexity's Publisher Program is getting a second look. Plus: ChatGPT defeats CAPTCHA, Google Zero gets closer, and more.
With the issue of AI respecting copyright on the ropes from both the courts and the White House, this seems like a good time to revisit alternative ways for AI builders to compensate content creators. Case in point: the Perplexity Publishers' Program, which just got a boost by signing Gannett. Sharing revenue against advertising in AI summaries just might be a viable model—that is, assuming humans do their own searching in the future.
More on that in a minute, but I've got some news of my own: Since time slows down in summer, I want to give everyone the chance to save big on my fall AI courses, and I'm extending the 25% discount to August 7. I'm offering two entirely separate cohorts, one for journalists and one for PR professionals. Each one will have a maximum of 25 students (a big part of the value is personalized coaching, so there's limited bandwidth), so I encourage you to reserve your spot early. Classes start the week of Sept. 8.
Here's a little more detail on the courses, then let's talk about Perplexity's new deal.
A MESSAGE FROM THE MEDIA COPILOT AND THE UPGRADE
Most people are still playing with AI at the surface—tweets, outlines, press releases. But the real value comes when you go deeper.
We’ve built two hands-on, live courses—one for journalists, one for PR pros—to show you how. These six-week programs focus on strategy, workflows, and real productivity gains, with 1:1 coaching and a final project to cement your skills.
Kicking off in September, both courses are six weeks and 100% live. And both come with 1:1 coaching and a capstone project guaranteed to level-up your personal workflows.
💥 Act fast: Get 25% off either course if you enroll by August 7.
🔹 AI for Journalists: Use code AIJO25OFF-SEP (checkout link)
🔹 AI for PR & Comms: Use code AIPR25OFF-FL1 (checkout link)
This is your AI upgrade. Don’t miss it.
Perplexity's rev-share gamble gets real with Gannett
Remember the Perplexity Publishers' Program? Announced last summer, the program is Perplexity's answer to the question of how content creators get compensated by AI search engines. It amounts to a revenue-sharing agreement: instead of licensing content to begin with, Perplexity shares a portion of the revenue on any advertising sold against answers fueled by the publisher's content.
After an initial wave of agreements with media companies, including Time, Fortune, and The Texas Tribune, the program has been making slow progress over the past year, signing The Los Angeles Times, Adweek, and a few international publications in the winter. Since then, there's been little news of more media outlets joining up, even while OpenAI, Amazon, and others continued to make licensing deals.
So you'd be forgiven for questioning Perplexity's commitment to the program before it announced this week that it had signed Gannett—publisher of USA Today and hundreds of local newspapers—to come on board. Apparently a year in the works, the deal is the first AI agreement that Gannett has signed with anyone, earning it more pomp and circumstance than your average press release, and an official post on the Perplexity blog.
That may signal a redoubling of efforts to make the program work, which has earned its share of skepticism. When Perplexity launched the program last year, it didn't yet have any ads on its platform and the person in charge had more experience scaling tech businesses than selling advertising. Advertisers reportedly showed some skepticism for the types of ads on offer—typically sponsored questions tacked onto a search—and some doubted the program could succeed.
Really, the most notable thing to happen to the Perplexity Publisher Program in the last year is that it was the target of a major lawsuit from News Corp, which accused Perplexity of trying to backdoor a form of licensing through a rev-share program. As publisher after publisher added to the stack of partnerships with OpenAI, it started to feel like the media wasn't just following the money, but actively shunning Perplexity's program.
Now, with new advertising hires, the Gannett deal, and legal winds that appear to be shifting in favor of AI companies, Perplexity's program is starting to look relevant again. And all the industry consternation over search referrals, which are rapidly evaporating, mean publishers are looking even harder for new line items on the balance sheet that make the total go up, not down.
The wild card is Perplexity's Comet browser, which signals a shift into an agentic web browsing experience. The blog post about the Gannett deal calls it out specifically, alluding to all the agent activity—when the Comet Assistant performs browser tasks on behalf of users—that it will enable, and all the content it'll need to lean on for informational context (some of which may even come from Gannett).
The thing is, agents don't look at ads. As the Comet Assistant goes to get information to do whatever odd jobs the user wants, almost all of that activity will happen in virtual windows or thinking steps the user will never see. There's a strange irony in Gannett boasting about a deal where the revenue is dependent on humans using Perplexity, and in the same announcement touting a web browser whose primary feature is a robot doing the browning for you.
Maybe I'm wrong, and the surge in bot activity will create more opportunities for ad revenue. Maybe "ads for bots" will be a thing. But as we continue further into the great AI unknown—where business models, content formats, and even the nature of web traffic seem up for grabs—I can see why most publishers would prefer the predictability of a licensing agreement, where you don't have to wonder if the audience you're attracting even has eyeballs.
The Chatbox
All the AI news that matters to media*
The zero-click apocalypse continues
Research that shows how AI search devastates traffic continues to mount, Press Gazette reports. This time it's from Authoritas, which reveals publishers lose nearly 50% of their traffic on searches with Google's AI Overviews. The study, submitted as part of a legal complaint to the UK's competition watchdog, found desktop clickthrough rates dropped 47.5% and mobile rates fell 37.7% when AI Overviews kicked in. While Google predictably dismissed the findings as "flawed," the research certainly confirms that many, if not most, users are satisfied enough by AI summaries to skip visiting the original sites entirely. Even more concerning: the report's conclusion that Google could feasibly deploy AI Overviews for over 90% of queries if it chose to. The zero-click future that publishers have long feared is rapidly becoming the present.
A twist in the bot masquerade
ChatGPT Agent, which can browse the web and perform actions on behalf of the user, can bypass Cloudflare's "I'm not a robot" checks—those web pages that make you click a box before proceeding. It's a minor act, but a major turning point: bots can now defeat the very security measures designed to keep them out. If AI can convincingly masquerade as human users, it fundamentally changes how publishers think about audience analytics, subscription walls, and even basic web traffic data. More advanced CAPTCHA systems (where you pick the traffic lights or somesuch) should still work, but it's just a matter of time before those fall, too. As Toshit Panigrahi of TollBit told me the other week, this may be an area where regulation is needed: a bot should be forced to say what it is, if only so we know what content to show it. (AI-assisted)
Out: disclosing AI. In: explaining AI.
Research from the Center for News, Technology & Innovation reveals a troubling catch-22 for newsrooms trying to be transparent about AI use. While audiences consistently say they want disclosure about AI-assisted content, studies show that labeling articles as AI-generated actually reduces trust and perceived accuracy—even when the content quality remains identical. The effect varies wildly based on tiny differences in label text, and the research suggests people with deeper journalism knowledge are often more concerned about AI use, not less. This puts media organizations in an impossible position: be transparent and potentially damage credibility, or stay quiet and risk appearing deceptive. The findings suggest that simply slapping "AI-assisted" labels on content isn't the solution—newsrooms need to focus first on better communicating their editorial processes and verification standards before audiences can properly evaluate the role of technology in their work. (AI-assisted—here's what that means 😉)
Bot blocking evolves into bot billing
DataDome, a company that specializes in blocking bots for publishers, announced a partnership with TollBit, adding the ability to monetize bot traffic rather than blocking it entirely. The partnership allows websites to identify "compliant" AI agents and charge them for access through automated paywalls. Both companies have seen big increases in AI scraping, which represents both a threat to old models and an opportunity to create new ones. It takes clout and cooperation to do that, though, and the alliance, along with developments like Cloudflare's Consent to Crawl standard, may give the industry the push it needs to build an AI ecosystem that works for everyone. (AI-assisted)
The AI copyright fight gets reinforcements
Mediavine, an ad-management company representing 17,000 websites, is petitioning the U.S. Copyright Office for immediate action on fair use exemptions, mandatory licensing frameworks, and transparent attribution requirements, and it's launched a Chage.org petition for creators to voice their support. The company's five-point platform essentially argues that AI training on copyrighted content without permission shouldn't qualify as fair use, and it provides a roadmap for fighting AI substitution through policy rather than just business deals. However, given that the current administration has recently made clear it does not agree that AI companies should be required to respect copyright, the whole exercise may end up being academic.
*Some items are AI-assisted. For more on what this means, see this note.