

Content licensing arrives: your archive is a product, not a traffic trap
Axios hit its 2026 revenue target in early September, almost two months ahead of last year, while Reuters saw site traffic fall 19 percent and still called its advertising healthier. Attention is no longer the only thing a library of words can earn.
By Katie Delaney / 2026-09-21 / 15 min read

What Axios Direct actually sells
Reuters site traffic fell this much year on year in August 2026, per Similarweb figures Digiday cites, while its digital chief says advertising is doing better than before the paywall
Start with the number that should make every content lead sit up straighter. Reuters' site traffic dropped 19 percent year on year in August, according to Digiday's report citing Similarweb, yet Phil Andraos, Reuters' general manager for digital, says the advertising business is doing better than it was before the paywall arrived. The hedgerow thinned and the harvest held. That is the first lesson of content licensing: traffic is a symptom, and the real question is what the library itself is worth to somebody who will never click a link.
Now for the second number. Digiday reported on 18 September that Axios hit its 2026 revenue target in early September, against the end of October in 2025. Its chief revenue officer, Jacquelyn Cameron, put the direct-from-client share of this year's revenue at about 52 to 54 percent, and Axios is readying Axios Direct: content feeds built for AI models and agents rather than for people browsing a homepage.
| Item | Value |
|---|---|
| Day target was hit | 304 to 250 |
Read that chart with a squint. Early September and end of October are Digiday's words, not dates, so the day counts are ours and approximate. What is not approximate is the direction: a publisher that treats content licensing as a revenue line, not only attention to its pages, is getting to its number sooner.
Three feeds for three kinds of buyer
Axios Direct will arrive in three tiers, per Digiday. The first, an Enterprise feed, launches this month for investment firms, sold on an annual fee that scales with the company's size and its assets under management, on two-year contracts, and pitched as an upsell to existing Axios Pro Deals subscribers. The second, a Corporate AI feed for companies with their own models, has no launch date. The third, a feed for individuals' personal AI agents, is described as an RSS feed for ChatGPT, Claude and their kin, and gets a deeper look in 2027.
Launching this month. Investment firms pay an annual fee set by their size and assets under management, on a two-year contract.
In development, no launch date. Built for companies whose own AI models ingest the content directly.
Earliest stage. An RSS-style feed for individual subscribers' agents, explored more deeply in 2027.
Notice what is missing from that list: a per-crawl toll, a training-data cheque, a traffic promise. Axios is pricing by the buyer's balance sheet, the way a data terminal does, so the fee follows the value to the buyer and not the volume consumed. Axios's stated reason for two-year terms is room to experiment and fine-tune. That is the move worth stealing for any content licensing plan: a fox does not sell every rabbit at the same price.
How the Reuters paywall decides who pays
US dollars a month, flat
One price, no introductory offers, per Digiday.
Countries, same offer
Available in 80 countries at the same flat price.
Reuters segment growth, Q2 2026
Thomson Reuters reported $229 million against $218 million, 4 percent organic.
The Reuters paywall is a masterclass in refusing to price by guesswork. Reuters charges a flat $4 a month in 80 countries and offers no introductory discounts, per Digiday. The clever part is the gate itself, a dynamic paywall run by machine learning that weighs each visitor's behaviour and how a piece of content has performed, estimates how much advertising revenue that visitor could earn, and only then decides whether to show the wall at all.
Say that again slowly, because it is the opposite of a blunt meter. A dynamic paywall treats every page view as a small auction between two revenue lines: the ad the visitor would earn and the subscription the visitor might buy. Andraos told Digiday that the two lines reinforce each other and that subscribers are highly engaged. Reuters also began blocking bots in May, and says bot traffic has fallen while monetisable traffic has stayed unchanged, though the article does not tie the 19 percent drop to any single cause.
Traffic is a vanity number until somebody decides what a visit is worth.
The revenue underneath
Thomson Reuters' second-quarter 2026 results put the Reuters segment at $229 million against $218 million a year earlier, up 5 percent, or 4 percent organically. The filing itself credits higher Agency revenues and a contractual price increase under its news agreement with LSEG, and the segment discussion we read does not mention AI licensing at all. So treat Digiday's account of paywall success as management's story, not audited fact, and enjoy the fox's habit of checking the tracks twice.
Then there is the archive. In May, Thomson Reuters chief executive Steve Hasker told Press Gazette that Reuters' content licensing for AI covers the text archive only, not the news feed, images, video or audio, that the company aims for the highest price it can get, and that it has tried to keep the deals short so they must be renegotiated. Read together, the picture is precise: the freshest work is protected by a paywall and a wholesale line, and the settled archive is what the AI shops rent, briefly, at a price set by scarcity.
If your own content has a shelf life, the same content licensing logic applies to it. We wrote about the squeeze on publisher traffic in our playbook for B2B content when Google sends less, and the Reuters paywall is the other half of that story: when clicks thin, decide what a visit, or a feed, is worth.

What AI licensing deals pay, and who holds the tollgate
| Item | Value |
|---|---|
| Axios, 2026, Direct from clients | 53 |
| Axios, 2026, Everything else | 47 |
In content licensing, direct is not a religion. Cameron told Digiday that Axios has not ruled out third-party AI marketplaces, that she expects it to appear in one at some point, and that it is talking to Snowflake about the Snowflake Marketplace. That is a sensible mix for a publisher with a name: sell direct where the buyer is known, and let marketplaces sweep the loose grain.
Wiley shows the other end of the field. Its fiscal 2026 results, released on 16 June 2026, report $49 million of AI revenue, up 23 percent, with lifetime AI revenue above $110 million, and name partnerships with IQVIA and OpenEvidence. Those are specialist buyers paying for specialist knowledge, which is exactly the shape a regulated B2B archive might take in content licensing terms.
Google is building a third lane. Digiday reports that Google's AI contribution pilot pays publishers a monthly figure shown in Search Console, based on value rather than raw usage, and only when Google judges that content meaningfully contributed to an AI answer across Gemini, AI Overviews and AI Mode. Publishers can opt out. Early participants call the calculation a black box, and some describe payouts as small against advertising revenue. In other words, this flavour of AI licensing is a payment set by the platform's own judgement.
The tollbooth problem
The most useful sceptic of content licensing is the Open Markets Institute. Its April 2026 report, Same Gatekeepers, New Tollbooths, argues that publishers are in a double bind, because the same Big Tech firms whose AI products erode website traffic are building and controlling the licensing infrastructure. It describes a three-tier market: big bilateral deals, an intermediary layer of startups and platforms, and an uncompensated majority outside any framework. A Brookings analysis by the report's lead author, published on 9 June 2026, adds that current AI licensing deals price retrieval-time use while treating model training as a donated cost.
| Item | Value |
|---|---|
| Bots bypassing limits | 3.3 to 12.9 |
That is why the plumbing matters. Cloudflare's pay per crawl lets a site owner choose, per crawler, to allow, charge or block, using the long-dormant HTTP 402 Payment Required status. RSL, Really Simple Licensing, is an open standard for machine-readable licensing terms covering attribution, pay-per-crawl and pay-per-inference, backed by names including Cloudflare, Reddit, Vox Media and Yahoo. Neither guarantees a content licensing buyer. Both put your terms where a machine can read them. The flip side of all this, refusing AI training outright, is covered in our guide to AI crawl control.
Regulated brands and content licensing: sort the archive
| Asset in your archive | Best home | Why |
|---|---|---|
| Original datasets, indices and benchmarks | Licence | Structured, timely, hard to copy: the Axios Enterprise logic |
| Analysis written for named professionals | Gate | Worth paying for once the reader is known |
| Licence status, compliance and regulatory explainers | Keep open | Correct citation beats exclusivity; a wrong AI summary is a compliance risk |
| Product, pricing and eligibility pages | Keep open | Accuracy beats scarcity on the pages buyers check last |
| Client case studies with figures | Gate or hold back | Contracts and privacy limits usually decide this one, not marketing |
| Old news posts | Keep open, tidy | Cheap to serve and rarely worth a licence |
- Original datasets, indices and benchmarksLicenceStructured, timely, hard to copy: the Axios Enterprise logic
- Analysis written for named professionalsGateWorth paying for once the reader is known
- Licence status, compliance and regulatory explainersKeep openCorrect citation beats exclusivity; a wrong AI summary is a compliance risk
- Product, pricing and eligibility pagesKeep openAccuracy beats scarcity on the pages buyers check last
- Client case studies with figuresGate or hold backContracts and privacy limits usually decide this one, not marketing
- Old news postsKeep open, tidyCheap to serve and rarely worth a licence
Every regulated B2B brand is sitting on a den of assets it has never priced. The honest first step in content licensing is a triage, not a contract. Ask of each asset three questions: would anyone pay for it in a structured form, would giving it away make you more citable, and could either choice put you offside with a regulator or a client? Our content marketing services start from exactly that inventory, because you cannot license what you have not counted.
In fintech, the closest match to Axios is proprietary market data, rate commentary and product benchmarks, the material a firm's own clients would pay to have arrive in structured form. If that describes your archive, read our fintech marketing work with the licensing question in mind. In health, Wiley's IQVIA and OpenEvidence partnerships show specialist knowledge finding specialist buyers, but patient-level material is never on the table, and healthcare marketing lives or dies on that line.

In cybersecurity, the temptation is to license threat research, and the trap is that research describing exploitable weaknesses in detail can do harm in the wrong hands; treat that as a legal and ethical question first, and see how cybersecurity marketing earns trust without over-sharing. In iGaming, the licence status, responsible gambling and player-protection pages are the ones a summary must get right, so iGaming marketing should default them to open, dated and quotable.
The prize on the open pages is being the source the machine quotes correctly. Google's own AI features guidance says there are no additional requirements to appear in AI Overviews or AI Mode and no special optimisation is needed, and that snippet controls such as nosnippet and max-snippet sit with the site owner. Good pages, honestly dated, are the whole trick, and our SEO and GEO services are built around making them quotable.
There is also the wider mood in the trade. Search Engine Journal's State of Search 2027 survey found 43 percent of respondents plan to invest in or prioritise GEO, yet only 9 percent feel very confident measuring AI visibility. Half the pack is heading up the trail with no compass, and the licensing question is one the compass would answer.

Monday morning, before you sign or block anything
Judge content by visits
Every page is scored by sessions, so the archive is only as valuable as last month's click count, and a fall in clicks reads as a fall in worth.
Judge content by buyers
Every page is scored by who would pay to use it, in what form and under which terms, with visits as one signal among several.
You do not need a Reuters-scale paywall or an Axios-scale sales team to think about content licensing by Monday. You need a list. Sort your archive into what you would licence, what you would gate and what you would keep open, put a name against each pile, and book the conversation with compliance before you book the one with a buyer.
Keep content licensing contracts short and reversible. the Thomson Reuters chief executive says he keeps AI deals short so they can be renegotiated, and Axios is using two years to experiment. The legal ground is still being surveyed: the US Copyright Office published a pre-publication Part 3 of its AI report on generative AI training in May 2025 and says a final version will follow. A clause you cannot revisit is a hostage to a ruling nobody has written yet.
Measure what you decide. Track citations of your open pages, enquiries that mention an AI answer, and any machine-readable terms you publish, and review the split every quarter. Our reading of the wider market is in our piece on the content results gap, and on why expertise is the one thing an AI cannot be handed, in Content Marketing Expertise Cannot Be Downloaded. Rates for the work are on our pricing page.
The brand that does this in September is not chasing a trend. It is deciding, calmly and early, which of its drawers stays open to every passer-by, which opens only for a key, and which is worth a price. That is a quiet, moonlit kind of advantage, and it belongs to whoever counts first.
Frequently asked questions
What is content licensing?
Content licensing means giving another party permission to use your content, under agreed terms and usually for a fee. In the AI era that can mean a feed an AI model ingests directly, an archive used to answer questions, or machine-readable terms on your site.
How much do AI licensing deals actually pay?
It varies enormously and most figures are private. Wiley reported $49 million of AI revenue in its 2026 fiscal year, while early Google pilot participants describe small payouts against advertising. The Open Markets Institute says the long tail sits outside any compensation framework.
What is a dynamic paywall?
A dynamic paywall decides per visitor whether to show the wall, instead of metering everyone the same way. Reuters' version uses machine learning on user behaviour and content performance to estimate advertising value before deciding.
How does the Reuters paywall work?
Reuters charges a flat $4 a month in 80 countries with no introductory offers. Its machine-learning system estimates how much ad revenue a visitor could earn and then decides whether to show that visitor the paywall.
What is AI licensing and does a regulated brand need it?
AI licensing is the sale or grant of rights for AI systems to use your content. A regulated brand does not need it, but does need to decide deliberately which assets to licence, gate or keep open, with compliance involved.
What is Axios Direct?
Axios Direct is a set of content feeds for AI models and agents. Digiday reports three tiers: an Enterprise feed for investment firms launching this month, a Corporate AI feed in development, and a personal agent feed being explored in 2027.
B2B content marketing when Google is extra : the publishers’ playbook
The traffic side of the story: what B2B content teams do when search sends fewer clicks.
Read the pieceContent marketingAn AI marketing agent raised $180M. Content results just hit a 12-year low
Where the gap between AI content promises and content results actually sits.
Read the pieceContent marketingContent Marketing Expertise Cannot Be Downloaded
Why expertise, the thing worth licensing, is the part no model can be handed.
Read the pieceSEO AND GEOWhen seo reporting tools go dark, the fox reports the gap
How to measure visibility when the tools you rely on lose their view.
Read the pieceReady to find out what your archive is worth?
Regulated B2B brands work with folkfox to triage their content, publish machine-readable terms, and stay quotable where it matters.
Want folkfox in your Google results and AI answers? Set folkfox as a preferred source.

