Content marketing and publishing news September 2026: AI answers, traffic shifts and licensing fights

The five weeks to 7 September 2026 turned the argument about AI and content into a set of filings and numbers. OpenAI and Microsoft asked a New York judge on 4 September to end the 10.8 million article copyright case, Brussels closed a publisher questionnaire on Google’s AI opt-out on 28 August, and crawl data showed AI bots taking 227 pages for every visitor they sent back.

The case that could set the price of published content

4 September 2026 · Source: PPC Land

What happened. OpenAI, Microsoft and the news plaintiffs all filed cross-motions for summary judgment on the same day in the consolidated action (25-md-3143) before Judge Sidney H. Stein in the Southern District of New York. The plaintiff side is not just The New York Times: it includes the Daily News group, Ziff Davis, the Center for Investigative Reporting and The Intercept. Publishers say OpenAI copied 10.8 million articles across five stages of its pipeline. OpenAI counters that verbatim reproduction appears in 24 instances across the 20 million anonymised chat logs the court ordered produced in November 2025, or 0.00012% of them, and that publisher revenues have gone up since ChatGPT launched.

Why it matters. Both sides are asking a judge to answer the fair use question before a jury ever sees it. A ruling either way changes what a licensing deal is worth, and every brand publisher sits downstream of that price. There is a second argument buried in OpenAI’s filing that deserves attention: it claims copies made before publishers blocked its crawlers were impliedly licensed under robots.txt precedent. If that holds, the date you added a disallow line becomes legally meaningful.

What to do. Check when your robots.txt and terms of use last changed, and keep the evidence. If you licence content to anyone, do not sign a multi-year AI rights clause until this motion is decided.

Brussels asks whether Google’s AI opt-out is worth anything

1 September 2026 · Source: PYMNTS / CPI

What happened. The European Commission sent publishers a questionnaire about the AI search opt-out Google proposed in June, which would let a site keep its position in conventional results while excluding its content from AI-powered search. Responses were due by 28 August 2026, and the exercise was reported on 1 September. It feeds an antitrust review of whether AI Overviews cut referral traffic and revenue in a way that breaks competition rules. Britain’s competition regulator has separately told Google to give publishers a content exclusion option for AI features.

Why it matters. The practical question the Commission is really asking is whether an opt-out that keeps blue links but drops you from the answer is a choice at all. On the evidence so far it is a menu with one edible item. And Google has already collected more than 10 billion euros in EU antitrust penalties over two decades, so this is not an idle inquiry.

What to do. Nothing yet, but decide now what you would do if the opt-out arrived in your market tomorrow. Our September SEO news report covers the Search Console side of the same shift, including the generative AI reports that landed on 31 August.

AI crawlers still take far more than they send back

14 August 2026 · Source: Digiday

What happened. TollBit’s report across 3,906 publishers, 456 of them European, put the scrape-to-referral ratio at 150 to 1 in Q1 2026 and 227 to 1 in Q2. European sites saw roughly four times the median AI scrapes of North American sites, scraping on them rose almost 20% between January and June, and their robots.txt instructions were ignored nearly three times as often. European publishers took one human referral per 179 AI bot visits. AI apps account for 0.05% of external referrals to European sites against 0.16% in North America.

A day later, PPC Land reported that about 15% of identified AI page fetchers hit URLs European publishers had explicitly blocked, and that ChatGPT-User reached disallowed pages on nearly half of the European sites that named it in robots.txt. OpenAI’s position is that user-triggered fetchers may not apply robots.txt rules at all.

What to do. Pull your server logs for August and work out your own crawl-to-referral ratio per bot. It is the single most useful number a content team can add to its dashboard this quarter, and the marketing analytics guide shows where it fits alongside GA4.

The traffic numbers publishers actually reported

6 to 13 August 2026 · Source: PPC Land

What happened. USA TODAY Co. reported average monthly unique visitors of 158 million for the quarter ended 30 June 2026, down from 180 million in Q1. Digital advertising fell 9.2% to $79.8 million while digital-only subscriptions rose 6.8% to $45.6 million. The company named three roughly equal causes: search traffic decay from algorithm changes, the loss of a programmatic partner, and a platform policy change affecting a sponsored links partner. A week later Mediavine, which represents independent publishers, said Google Search page views across its network fell 34% and Google Discover page views fell 15% between December 2024 and December 2025, with small publishers losing 60% of search referrals over two years and chatbot referrals still under 1% of page views.

“What really matters is ARPU per unique.” Michael Reed, chief executive, USA TODAY Co., 6 August 2026

Why it matters. Two very different businesses, one shape. Subscriptions and owned audience grew while everything that depended on a search click shrank. That is the strategic read of the whole quarter, and it argues for the email and community half of your plan rather than the twelfth blog post. The email and SMS marketing guide is the practical follow-on.

Listicles kept their rankings, and it barely helped

24 August 2026 · Source: PPC Land

What happened. Kevin Indig crawled 5.32 million organic result rows across 60,000 US English desktop queries in 15 verticals over 47 hours in August 2026. Listicles still appear in the top ten for 55.1% of queries and in the top three for 32.3%. The unannounced mid-January Google update that was assumed to have hit self-promoting vendor listicles did the opposite: 62% of them gained traffic, with a median gain of 38%. Position one share for listicles slipped from roughly 19 to 20% in January to about 15% in August.

Why it matters. Here is the sting. On 92% of pages carrying a top-ten listicle, Reddit or YouTube also appears, and AI Overviews sit on 83.7% of results pages in the sample. So the format survived the update and lost the click anyway. Ranking and traffic have come apart, and any content plan still measured on positions is measuring the wrong thing.

What to do. If you run comparison or “best tools” pages, keep them, but rebuild them to be quotable in an answer: a clear verdict near the top, a real comparison table, named prices, a date. Our content marketing guide sets out the format in full, and the 2026 SEO guide covers the citation mechanics.

Content budgets are quietly being rerouted

7 August 2026 · Source: Digiday

What happened. Fractl found 82% of marketers have put at least some search or content budget into AI visibility, with an average of about 24% and 43% spending more than a fifth of the budget there. Eighteen per cent have allocated nothing. The split by team is the interesting part: performance marketing routes 32% and SEO 31%, while brand sits at 21% and content at 20%. Semrush’s index put the average number of sources behind an answer at 15.4 for ChatGPT, 11.4 for Google AI Mode and 3.3 for Gemini. A 5W analysis found at least 27% of AI citations come from journalism, rising to 49% on time-sensitive queries. Separately, Scribewise reported that 67% of marketers appear in AI answers less often than they want, 71% are not tracking share of voice, and 67% are not looking at AI bot traffic at all.

Why it matters. The team whose output gets cited is the team funding the work least. That gap is the one thing on this page I would act on first, because it is entirely within your control and costs nothing but a reallocation. If journalism supplies a quarter of citations and half on fresh topics, publishing something genuinely new and dated is a citation strategy, not a nice-to-have. Read it next to our AI marketing news for September.

A Twitch class action prices creator work as training data

20 August 2026 · Source: PPC Land

What happened. Warren Pandiscia, a Connecticut streamer, filed a putative class action against Twitch Interactive and Amazon in the Northern District of California (case 3:26-cv-08721), claiming content was used to train Amazon’s Nova Reel video model without consent or payment. The claims include breach of implied contract, unjust enrichment and California unfair competition, with the amount in controversy above $5 million. The complaint leans on the fact that Twitch amended its policies on 12 August 2026 to add AI training language, arguing that the change is an admission the earlier terms did not cover it. Twitch has 240 million monthly users and between 3.2 and 6.9 million creators active in a month. For comparison, the filing cites the roughly $60 million a year Google pays Reddit and the $200 million-plus Reddit has disclosed in total AI data licensing revenue.

Why it matters. Platform terms are being rewritten under everyone, and the rewrite date is now evidence. Anyone running creator campaigns or holding a library of user-generated content should assume the same question lands on their contracts next.

What to do. Diarise a review of every platform’s terms you rely on, and make sure your own creator agreements say plainly whether AI training is in or out.

The IAB says when you must label AI creative, and when you need not

18 August 2026 · Source: PPC Land

What happened. The Interactive Advertising Bureau released version 2 of its AI transparency and disclosure framework, specifying disclosure triggers down to Unicode characters and metadata fields. Labels are expected for AI-generated images and video, synthetic voices of the dead, synthetic voices of living people in fabricated scenarios, photorealistic AI influencers and digital twins. Colour correction, upscaling, stylised imagery, authorised voice clones, generic voiceovers, background music and, notably, text and copy generation are exempt. The rollout runs over 24 months, with 60 days to appoint an AI disclosure lead. The framework cites NYU Stern research finding that adding an AI disclosure to an ad cut click-through rate by 31.5%. It also notes a 37-point gap between the 82% of advertising executives who think consumers feel positive about AI creative and the 45% of Gen Z and millennials who actually do.

What to do. Appoint the disclosure lead now rather than in month 23, and treat the 31.5% figure as a design problem: where a label is required, test placement rather than assuming the loss is fixed. The video marketing guide covers the AI video tools this touches most.

What to do this month

Four things, in order. First, calculate your crawl-to-referral ratio per bot from August logs, because you cannot argue about AI traffic without your own number. Second, move budget: if your content function is at the 20% Fractl average, that is a decision you made by not making it. Third, rebuild your two highest-value comparison pages so an answer engine can lift a verdict from them, with a date and a table. Fourth, read your platform terms of service, all of them, and note the last amendment date.

And one thing not to do. Do not sign anything granting AI training rights over your archive until the New York ruling lands. The weekly roundup for 7 September puts these stories next to the ad tech and social developments from the same week.

One gap worth naming: newsletter platform news was thin in this window. There was product movement around AI bot blocking on the major creator platforms in late August, but nothing I could confirm to a primary source in time, so I have left it out rather than repeat a secondhand summary.

Last researched and updated: 7 September 2026.