PostA daily briefing for media sellers
No. 92 · 28 Sep 2026 · 12 stories
The double squeeze
New Jersey's data broker law and Morrissey's commoditization thesis are the same story from two ends of the supply chain. On one side, publishers that assumed their first-party reader relationship kept them exempt now face registration fees up to $1.5 million for selling data onward to brokers. On the other, AI developers value their content not as expertise but as interchangeable raw material to be mixed and resynthesised. What both share is the loss of control over the thing publishers thought they owned: the data, and the content, that defined their value. The only counter-leverage in either case is collective: en masse blocking of Google's scraper, or an industry-wide refusal to accept commodity pricing.
Engagement over volume
The Digiday summit made the response concrete. Reuters' dynamic paywall lifted ad revenue precisely because subscribers are authenticated and engaged, not because traffic grew. Future's GEO product is renewing clients while its own CRO admits no one has solved whether AI visibility becomes sales. The metric has shifted from monthly uniques to share of model, from traffic to engagement, and the publishers winning are the ones that built the direct relationship before the search referral disappeared. That is the same bet iHeart is making by selling broadcast programmatically and marketing its "guaranteed human" stance, and it is the same bet Channel 4's teen research validates: audiences still want ads, just relevant and trustworthy ones.
Automation's human cost
The two labour stories frame the risk. An agent can now write a brief, contact creators and negotiate rates, and agencies have quietly stopped hiring graduates because AI fills the gap. The counter-argument from inside the room was not "AI won't replace jobs" but "someone who knows AI and has your experience is coming to replace your job." Differential privacy is the privacy-tech answer to the trust problem automation creates, but its own caveat (noise degrades precision, clean rooms can enable privacy washing) is a reminder that every tool the industry is reaching for comes with an unresolved question.
What this means for publishers
Run the data-flow inventory now, not in 2027. If a publisher sells or licenses reader data onward to anyone who might qualify as a broker, they are likely a New Jersey "data collector," and the trigger is who you share with, not what you collect. Audit resolution, onboarding, enrichment and licensing before the April 1 registration window, and kill any arrangement whose value is outweighed by the fee and compliance burden.
Do not wait for a bilateral AI deal if you are mid-market or long-tail. The realistic on-ramps are the marketplaces (TollBit, Dappier, ProRata) and Google's pay-per-value pilot, and the research literature shows deals concentrate upmarket. Separate training-archive licensing from display/RAG rights and price scarcity (archive depth, verified data) rather than raw volume.
Sell authenticated audiences as the premium. Reuters' paywall lifting ad revenue is the cleanest proof point we have that engagement and logged-in identity beat raw reach, and it is the exact reframing publishers need against the "more impressions" pitch. GEO is the forward bet: 43% of publishers plan to invest in it this year, and the metric is share of model.
Treat the entry-level freeze as a talent supply-chain risk, not a cost saving. No junior pipeline now means no senior sellers and yield analysts in a decade. Redesign junior roles around AI oversight and measurement, because the competitors who skip this are quietly surrendering their future bench.
When you sell privacy tech, sell governance, not accuracy. Differential privacy and clean rooms inject noise and can enable "privacy washing" if sold as automatic compliance. Position clean, structured, governed data collaboration and keep your own first-party graph as the negotiating asset.
On the Radar
New Jersey data broker registration window opens April 1, 2027. Publishers should run a data-flow inventory well before, not after.
Google's pay-per-value AI licensing pilot. Watch whether it scales beyond the test and gives publishers a real price signal versus zero-click traffic loss.
Google ad-tech behavioural remedies. Digiday's Sept 23 scorecard found publishers see progress but "far from a clean win," so watch what the remedies actually change in auction mechanics.
EU Digital Services Act oversight of ChatGPT (designated a Very Large Online Search Engine in Aug 2026). Watch whether generative search gets search-engine-level transparency duties.
Programmatic I/O New York, Sept 28-29, 2026 (today). Expect agentic and AI-discovery chatter to dominate.