As AI reshapes newsrooms, leading media outlets are charting different paths for its use

 By Erik P. Bucy & Milad Jalalian Ebrahimi

A labor dispute erupted at The Guardian’s London headquarters in December 2024, with nearly 500 journalists going on strike to protest the proposed sale of the Guardian-owned Observer.With the newsroom hobbled, the British publication’s management allegedly deployed generative AI tools such as ChatGPT and Claude to suggest headlines and write descriptions of photos for people using screen readers. The National Union of Journalists cried foul, calling this move a clear attempt to undermine collective bargaining.

As media and communication scholars, we saw the outrage over management’s alleged use of AI during the strike – the details of which the outlet’s management has disputed – as more than routine wrangling over adherence to labor law. The incident showed how AI is becoming integrated into daily news work through translation, interview transcription, research, writing and audience analytics. At the same time, the incorporation of AI in newsrooms has raised new worries about accuracy and bias.

In our recent research, we examined how three different media outlets – Reuters, the BBC and The Guardian – were deploying AI in their workflows.

We conducted a comparative case study, analyzing public documents published through 2025, including official AI policy statements, editorial guidelines, trial reports and industry presentations.

We found that there is no single story about AI’s integration into newsrooms. Instead, the way the technology is used appears to reflect the funding, values and mission of each outlet.

📊 Publisher Revenue Dinner Miami 2026

We are hosting a private dinner in Miami for senior revenue leaders across news, gaming, newsletter, specialist, and brand publishing. This dinner is for publishing professionals who are actively driving revenue strategy and want a candid, peer-level conversation on how to win the biggest commercial quarter of the year. If you are a Chief Revenue Officer, CEO, Programmatic Head, or Partnerships Director making decisions around yield, inventory, first-party data, and Q4 growth, this evening is designed for you.

Date: Wednesday, September 16, 2026
Time: 6:30 PM
Venue: Coral Room, Rusty Pelican Miami Restaurant, 3201 Rickenbacker Cwy, Key Biscayne, FL 33149, United States

What you will walk away with

• Practical yield, personalisation, and first-party data moves that are lifting revenue per visit right now

• A candid read on how your peers are pricing, packaging, and positioning inventory for the Q4 surge

• Insight into which new formats — video, newsletters, and creator partnerships — are actually paying in 2026

• An attendee-only summary of the evening's key insights and growth playbooks shared around the table

The dinner is high-trust, collaborative, and entirely off-the-record under Chatham House Rule. No panels, no pitches, no slides — just one focused evening on how to make every session, impression, and first-party signal work at full value in publishing's most important quarter.

Seats are strictly limited to 15 attendees.

A big thank you to Ezoic for co-hosting this evening with us.

Apply for your seat here ▸

🎯 LiveRamp × OpenAI: Measure Your ChatGPT Ad Performance

If you're running or planning ads on ChatGPT, you can now close the measurement gap. LiveRamp's Conversions API (CAPI) Hub connects directly to ChatGPT ad campaigns via secure server-to-server data connections, meaning you get reliable conversion tracking without depending on browser-based signals that often break or go missing.

In practice, this means you can:

• See which ChatGPT campaigns are actually driving conversions — not just clicks

• Optimize spend in real time based on accurate performance data

• Justify and grow your ChatGPT ad budget with measurement data you can trust

As consumer journeys shift toward AI-powered platforms, this gives marketers the infrastructure to keep up, connecting data, measuring impact, and proving ROI on one of the fastest-growing ad surfaces in the world.

Read more ▸

By Vahe Arabian

📊 Bot Traffic, Broken Promises: What Publishers Get Wrong About AI Licensing

Bot traffic is not a revenue line. It is a negotiating position.

AI crawler activity is accelerating. Publisher revenue from that activity is not.

DataDome recorded 17.7 billion AI-agent requests across its network in Q2 2026—a 45% increase from the previous quarter.

But the more important finding is who created that traffic and what publishers received in return.

Meta’s crawlers generated 9.1 billion requests, accounting for more than half of DataDome’s measured AI traffic. They returned almost no referral traffic.

ChatGPT moved in the opposite direction. ChatGPT-User crawled 6% less, while ChatGPT referral traffic increased 17%. It accounted for 80–88% of measured AI referrals each month.

Crawl volume and publisher value are not merely disconnected. In some cases, they are moving in opposite directions.

Analytics is not monetisation

This is why publishers should be careful about treating bot analytics as an emerging revenue line.

The data is valuable. It can reveal:

• Which companies crawl your content.

• Which URLs they request most frequently.

• How much referral traffic they return.

• Whether they are searching, training models or acting for a user.

• How often they re-fetch content that has not changed.

But a dashboard does not create a buyer.

The first commercial benefit is evidence, not income.

Crawl-to-referral ratios and most-requested URLs can turn a blind licensing conversation into an informed one. They can help publishers identify valuable parts of an archive, quantify infrastructure costs and decide which crawlers should be allowed, restricted or blocked.

That creates negotiating leverage.

It does not guarantee a licensing agreement.

Why vendors are shifting toward analytics

The earlier commercial model was brokerage.

A provider introduced a publisher to an AI company, negotiated a bespoke licence and collected a share of the agreement.

That approach was slow, difficult to scale and generally worked best for large publishers with distinctive archives and recognised brands.

Analytics is easier to productise. It can be deployed quickly, priced like software and renewed monthly.

A content marketplace requires both supply and paying demand. An analytics product only requires the publisher.

That does not make the analytics worthless. It means publishers must evaluate the product based on the value it provides now—not the revenue it might theoretically unlock later.

The latest crawler data strengthens that argument

Cloudflare reports that AI training accounted for 52% of classified crawler requests as of June 2026, up from 22% in spring 2025.

Mixed-use crawlers—which combine search, agent activity and training—represented more than 36% of crawler activity.

Cloudflare also estimates that more than half of AI crawler traffic is spent re-fetching pages that have not changed.

For publishers, that is an existing cost: bandwidth, CDN requests, server resources, log storage and security monitoring consumed without any guarantee that the content will generate a citation, referral or payment.

The data gives publishers a way to quantify that imbalance.

Three things the dashboard cannot decide

1. Whether blocking will cost more than it saves

Search, training and agent activity are different use cases.

However, some companies still combine multiple purposes within the same crawler. Blocking a mixed-use crawler may prevent training, but it could also affect search visibility, AI citations and referral traffic.

The commercial cost of lost visibility can arrive immediately. Licensing revenue may never arrive.

Publishers should therefore classify crawler activity before applying site-wide blocks.

2. Whether a buyer exists

More than 50 publisher–AI licensing agreements have been signed since 2023, according to Cloudflare. That confirms a market exists.

It does not confirm that every publisher has something buyers will license.

AI companies are more likely to pay for proprietary reporting, specialist archives, structured datasets, real-time information and content they cannot obtain elsewhere.

Commodity articles that repeat widely available information have much less scarcity.

Publishers need to price the differentiated archive—not assume that every crawled page has equal commercial value.

3. Whether a crawl created value

Per-crawl pricing rewards requests rather than meaningful use.

A crawler could request the same unchanged page repeatedly without citing it, displaying it or generating a single referral.

That is why commercial models are shifting from Pay Per Crawl toward Pay Per Use, where payment is associated with content appearing in an AI search result or agent response.

It is a more defensible model, but it still depends on reliable use reporting, willing buyers and sufficient demand.

Cloudflare’s 15 September change

From 15 September 2026, Cloudflare will introduce new crawler defaults.

For new domains, new sites added by existing customers and existing free-tier customers who have not changed their settings:

• Search crawlers will remain allowed.

• Training and Agent crawlers will be blocked by default on pages displaying ads.

• Mixed-purpose crawlers may also be blocked where one of their classified uses is restricted.

Publishers can override these settings.

But doing nothing is still a decision—particularly when mixed-purpose crawlers such as Googlebot, Bingbot and Applebot can be affected by the most restrictive applicable rule.

What publishers should do now

1. Treat bot analytics as evidence and cost visibility—not forecast revenue.

2. Separate Search, Training and Agent activity before blocking anything.

3. Review your Cloudflare crawler settings before 15 September.

4. Identify the proprietary content within your archive that a buyer cannot obtain elsewhere.

5. Compare crawler requests with citations, referrals and assisted conversions.

6. Measure the infrastructure cost of repeated and unnecessary crawling.

7. Continue building newsletter, membership and other first-party demand on the assumption that referral traffic will keep declining.

The mistake would be treating the dashboard as the deal.

Publishers could spend the next year measuring a potential revenue line that never arrives while overlooking the data’s immediate value: understanding demand, controlling access and negotiating from a stronger position.

Pull your last 90 days of crawler logs and separate the requests into Search, Training and Agent activity.

If the results surprise you, reply and tell me what you found. I’m comparing findings across publishers and will share the broader patterns back with readers.

Google tests replacing Sponsored Store with Featured Store in SERP | Search Engine Watch

Google is testing a new heading for its sponsored store section, swapping “Sponsored” for “Featured Store,” with a new subheading that reads “Shop these top products.” This section itself isn’t new Google was testing it in the SERP last year, but it now appears Google is changing the heading used for it. This looks similar to another test we’ve spotted in Shopping ad results in recent days, suggesting Google may be experimenting with several different labels and headings across its Shopping ad units at the same time.

Read more ▸

AI’s value in a news company should be measured in these 4 ways | INMA

A finance colleague showed me a report recently featuring 20 AI tools deployed across the group, several thousand employees trained, and more than 40,000 hours saved last quarter. She looked pleased. Then she asked the harder question: What have we actually done differently as a business because of this? We sat with that question for most of the afternoon. Neither of us had a clean answer. AI has quickly moved from curiosity to expectation. A few years ago, the debate was whether to experiment with it. Today the conversation is about scale: tools deployed, employees trained, hours saved, processes automated.

Read more ▸

We analyzed 6,400 New York Times stories to find out how comments change when you give readers more information | NiemanLab

Journalism exists to tell people what’s happening and give them what they need to understand it. That mission does not bend to budget or format. A story about a school board vote and a story about housing prices are both doing an essential job, just at different depths. The assumption underlying both types of stories is the same: that telling people more, and giving them precise information, helps them form their own judgments and take part in the conversation that follows. We wanted to know how this plays out in comments sections.

Read more ▸

Amazon DSP gains publisher-hosted creatives for GAM deals in 33 markets | PPC Land

The change is narrow in wording and wide in consequence. According to Amazon Ads, Amazon DSP now supports publisher-hosted creatives for Google Ad Manager programmatic guaranteed deals, allowing advertisers to bid on inventory where the creative assets sit inside the publisher's ad server rather than being delivered from Amazon DSP or a third-party ad server. Display and video formats are both covered. Amazon published the note in its product announcements feed on August 10, 2026. It carries no named executive, no partner list, and no performance claim. What it carries instead is a change to plumbing that decides which impressions a buyer can and cannot reach from a given seat.

Read more ▸

LinkedIn playbook offers guide to optimizing for AI | Social Media Today

LinkedIn published a new playbook on how to optimize content for artificial intelligence chatbot citations. The guide explains how to use LinkedIn to boost brand credibility, which is a key signal for AI chatbots. LinkedIn has been heavily invested in highlighting its value as a vehicle to help businesses improve their presence in AI replies. This has become more important as people increasingly turn to AI chatbots for answers, and as LinkedIn becomes a key resource for AI tools. Indeed, third-party research showed that LinkedIn is one of the most-cited references for AI chatbot answers.

Read more ▸

Anthropic says it will watermark text generated by its AI models | TechCrunch

Anthropic will watermark text generated by its models, including Claude, to comply with European regulations, the company now says. The AI model maker confirmed the watermarking in an updated support page. EU AI Act’s Transparency Code, which took effect on August 2, requires AI companies to mark AI-generated or edited content in a way other systems can identify them. Anthropic said that all models released after August 2 will automatically have tech to watermark both computer-generated text and files.

Read more ▸