How Publishers Make Money From AI in 2026: The New Toll Booths, Google's Payments, and the Path That Actually Pays

Open Google Search Console this month and a new box may be waiting for a small group of publishers. It reports a monthly earnings number for AI, money Google says it will pay when your content shapes an answer inside Gemini, AI Overviews, or AI Mode. For two years the story ran one way, AI reads your work and gives nothing back. Now a search engine is quietly cutting checks, a Cloudflare tool is charging bots a toll at the door, and a licensing standard backed by Reddit and Yahoo is trying to price your archive. If you run a Facebook page and a website, the question changes from whether AI will ever pay to which of these actually reaches you.

The honest answer is that the money is real, the money is early, and most of it is still small. There are now four ways publishers earn from AI, and they do not reach the same operators or pay the same amounts. Two of them are barely a year old and one only reached publisher dashboards this September, and these newer ones do trickle down further than the nine-figure licensing deals ever did. But the durable path, the one that turns AI from a channel that quotes you for free into a channel that pays you, is the same one it has always been, and every new payment mechanism quietly depends on it. This guide names all four, shows the current numbers, and then shows the foundation that makes any of them work for a Facebook-first publisher.

What actually changed: AI started paying, in small ways

For most publishers the AI money has been a spectator sport. Reddit and News Corp signed deals, everyone else watched. Three things have started to move that picture since the middle of 2025.

Cloudflare turned crawler blocking into crawler billing in July 2025. Google began testing direct payments to publishers whose content feeds its AI answers. And in September 2025 a coalition of publishers launched a shared licensing standard so a site does not have to be Reddit to put a price on its content. None of these replaces your ad revenue. Taken together they mean the question is no longer if you can charge for AI access, but whether the amount is worth the setup and whether you are the kind of source AI wants badly enough to pay for.

The money is real, the money is early, and most of it is still small. The path that turns AI into a channel that pays you is the same one it has always been.

The four ways publishers earn from AI right now

1. Licensing your archive to AI companies

This is the path with the biggest headline numbers and the one an everyday publisher is least likely to touch. AI companies pay some publishers for the right to train on and ground their models in a publisher's archive, and the reported deals are large.

Reddit pulls in around $130 million a year in combined licensing from Google and OpenAI. News Corp's OpenAI deal was reported at $250 million over five years, about $50 million a year. Wiley booked $49 million in AI licensing in its 2026 fiscal year. Taylor and Francis signed with Microsoft for roughly $10 million in the first year.

The part the headlines bury has not changed. This money flows to the brand-name archive with the scale to negotiate, and the long tail of small and mid-size publishers sees no meaningful revenue from a direct deal. AI companies want the firehose, not a network of niche Facebook pages. If you are reading this to size up a licensing check, that door still does not open for you. What has changed since mid-2025 is that two new mechanisms are trying to open a smaller door for everyone else.

2. Charging crawlers a toll: Pay Per Crawl and RSL

The first of those smaller doors is a toll booth. Instead of negotiating a custom contract, you set a price and let the infrastructure collect it.

Cloudflare launched Pay Per Crawl in July 2025 to do exactly that. When an AI crawler requests a paid page without offering to pay, Cloudflare's edge answers with an HTTP 402 response, the web's rarely used Payment Required status, along with your price, according to Cloudflare's own announcement. The AI company either pays the toll or walks away, and Cloudflare sits in the middle, charging the AI company and passing the earnings to the publisher, as TechCrunch reported at launch. Because it does not rely on voluntary compliance, it can actually block a crawler that refuses to pay.

Alongside it, in September 2025, publishers including Reddit, Yahoo, Medium, O'Reilly Media, Ziff Davis, and People Inc. backed Really Simple Licensing, or RSL. RSL takes the idea behind robots.txt further. Instead of a plain yes or no to a crawler, a site publishes licensing terms, and the standard supports free, attribution, subscription, pay-per-crawl, and pay-per-inference models. The RSL Collective is described as a rights collective similar to ASCAP and BMI in music, pooling licensing across publishers so AI companies can license at scale. The catch is that RSL still depends heavily on AI companies cooperating, which is why its backers describe it as working alongside Cloudflare's Pay Per Crawl blocking rather than replacing it.

Both are real, and both are more democratic than a $250 million licensing deal, because any site can set a price without a boardroom negotiation. The honest limit is demand. A toll only earns when a crawler wants your specific pages enough to pay, and that want tracks the same thing that earns you a citation, being a distinct, trusted source on a subject. A toll booth on a road nobody drives collects nothing.

3. Google's AI contribution pilot: getting paid inside Search Console

The second smaller door is Google's, and it is the one showing up in publisher dashboards right now. Google is testing what it calls the AI contribution pilot, described on its public policy blog in June and reported operating through Search Console in September 2026. Publishers in the pilot see an AI earnings widget that reports a monthly number, and payment is usage-based with no upfront fee.

The mechanics matter, because they tell you what Google is actually buying. Your content earns only when it contributes significantly while an answer is being generated across Gemini, AI Overviews, and AI Mode. Google's own help text is blunt about the boundary. Content that merely confirms a fact or gets linked after the answer is written does not qualify. You get paid for shaping the answer, not for being a footnote next to it.

Read that boundary twice, because it is the thesis of this entire article. Google will only pay you if you are the source its AI leans on to answer a question. That is a payment for AI Citation Presence, delivered as a check instead of a compliment.

4. Getting cited by AI, then owning where that traffic lands

The fourth path is the oldest one, and it is the reason the other three now work at all. When someone asks ChatGPT, Google AI Overviews, or Perplexity a question in your niche and the answer names your site and links to it, you gain two things. A slice of high-intent traffic, and a citation that compounds your authority as the answer. That traffic does not land on a licensing spreadsheet. It lands on your website and your Facebook page, where you already know how to turn attention into money.

That roughly 1 percent figure cuts both ways. It means AI is not yet sending publishers a flood of traffic, so anyone promising an AI windfall this quarter is selling something. It also means the field is wide open and the cost of becoming the cited source is low right now, before every competitor in your niche has worked it out. The publishers who win the next two years are the ones building AI Citation Presence while it is cheap, not the ones waiting for the flood and then trying to catch up.

Why the new toll booths still are not your business model

Two of these paths are far more reachable than licensing ever was, so it is worth being precise about their limits before you build a plan on them.

Google's pilot is early and, by the account of people inside it, small. Executives have described the calculation as quite black box, the widget reports earnings with no breakdown of how they were reached, and early returns have been called peanuts next to ad revenue. At least dozens of publishers have been approached, participation is voluntary, and you can turn it off inside Search Console at any time. The first numbers have since landed. The Information reported that Google is paying roughly 100 digital publishers through the pilot, and that for several small and midsize sites the payments amount to less than one-tenth of 1 percent of their advertising revenue. One publisher that had been in the program for a few months had earned about $50,000 to $60,000 so far. It is a real signal, not a paycheck to plan around.

The tolls have their own ceiling. Pay Per Crawl needs an AI company to want your pages badly enough to pay at the door, and RSL needs those companies to cooperate with the terms you publish. Cloudflare itself has started to move past the per-crawl model. On September 30, 2026 it launched Pay Per Use in beta, where AI companies report when they actually use a publisher's content and Cloudflare handles the billing and payouts, and it explained the shift in plain terms: charging for every crawl makes the buyer pay before it knows what it needs, and many buyers won't. Both reward sites that AI systems already treat as necessary sources, which is a position you earn, not a switch you flip. And setting a toll too aggressively can wall you out of the citations that send real readers, so the setup is a judgment call, not a default.

Every new AI payment rewards the same thing. Google pays you for shaping the answer. A crawler pays a toll for pages it needs. Both require you to be the source, not a footnote.

Here is the through line under all of it. Licensing pays the archives big enough to negotiate. Tolls pay the sites whose pages AI cannot skip. Google's pilot pays the content that shapes the answer. Every single one rewards being a necessary, distinct, trusted source, which is the fourth path. So the work does not split four ways. It concentrates on one thing that opens all of them.

This is Generative Engine Optimization, GEO, often referred to as AI search optimization or AEO, the practice of making your content the source AI answer engines quote as discovery moves from ten blue links to a single generated answer. For a Facebook-first publisher, GEO is the next layer of diversification past one platform, and it pairs with the owned website and Google Discover you should already be building. We have written the full playbook in GEO for publishers in 2026 and the case for the owned-site revenue stack that catches the traffic. If your first worry is bots reading your pages faster than humans, start with what AI crawler traffic actually is and which bots to let in. This piece is the money question underneath all of them, and the answer is to become the source worth paying for.

Become the source AI cites: what actually moves the needle

Getting cited is not a trick, it is an authority position, and PIB measures it directly. Your AI Citation Presence is how often and how prominently AI systems name you when they answer questions in your category. Your GEO Readiness Score is the composite of the signals that drive it. The work of raising them is concrete.

Be the clear entity, not a vague page. AI systems reward content that is unambiguous about who is answering and what the answer is. That means a real named author or brand with a track record, a page that answers one question completely rather than ten questions halfway, and Topical Authority built by covering a subject in depth across many linked pieces. This is Entity Positioning, and it is the same reason Google spent a decade rewarding depth. AI raised the stakes, because now the reward is not a ranking, it is being the answer itself, and increasingly a payment for being it.

Answer the question a human actually asks, in the first two lines. Answer engines lift the passage that most directly resolves the query. A page that opens by restating the reader's real question and then answers it cleanly is easy to quote. A page that opens with a scene and buries the answer in paragraph nine is not. Structure for the pull.

Make yourself easy for the machine to read. This is Technical Retrievability, and it is where the plumbing earns its keep. You need clean HTML, real structured data with FAQ and Article schema so the engine understands the shape of your content, accurate published and updated dates because freshness is a retrieval signal, and internal links that let a crawler follow your Topical Authority. Here is the tease for the operators reading this. You can wire the freshness and schema layer as a repeatable job, in n8n against your CMS, in a Make scenario, or in a scheduled build step, so every new piece ships retrievable by default instead of being fixed one at a time. The point is not the tool. The point is that retrievability is a system you run, not a page you polish once.

The control that protects your content without killing your traffic

For two years publishers faced an ugly choice. Block AI crawlers and you risked blocking the search crawlers that send readers. Allow everything and you handed over training data for free. In September 2026 that trade-off got a cleaner answer, and it is worth understanding before you touch a setting.

Cloudflare added a Disallow AI Training option, reported on September 15, 2026, that separates the two. It writes no-training directives for the crawlers that support them, a Disallow for Google-Extended and for Applebot-Extended, while keeping those same crawlers active for search indexing. Bing support is still pending because Microsoft has not added the robots.txt token yet. The older Block setting, by contrast, now stops those crawlers entirely, including for search, so migrating to the training-only control is the move if you want to keep your readers.

This also clears up the most expensive mistake publishers make. Google's own documentation says Google-Extended does not affect a site's inclusion in Google Search or its ranking, and a separate Search Console setting governs appearance in AI Overviews and AI Mode, so blocking Google-Extended on its own does not remove you from those answers. If your goal is to keep being cited while opting out of training, the training-only control does that. If your goal is to get paid for access, that is where Pay Per Crawl and RSL come in, as a priced gate rather than a closed door. What you never want to do is flip a broad block in the name of protecting your content and quietly cut your own traffic while leaving readers with nothing to find. Read your own settings before you change them.

The foundation that turns citations into revenue

A citation is not a deposit. It is a visitor, and now sometimes a small usage payment on top. The real money still happens on the property that visitor lands on, and this is where the Facebook-first publisher has an edge the pure-web publisher does not. You already run a distribution engine and a monetization surface. AI citations, GEO, and Google Discover become new mouths of the same funnel that feeds your Facebook Content Monetization and your owned-site ad revenue. That stack, the platform payout plus the owned site plus the search and AI traffic on top, is the Publisher Revenue Stack, and no single AI headline replaces it. It compounds it.

The through line under all of it is the same one that runs every real PIB engagement. Not a one-time setup, but a loop of analysis and optimization. You read your own data to see which pieces AI is already citing and which pages already earn, and you push more of what works. You lean into the authenticity a machine cannot fake, the real operator voice and lived track record, because when content becomes infinite that authenticity is the thing that gets cited and the thing that converts. You make the small deliberate moves, longer and more useful answers, tighter internal links to build Topical Authority, careful by-hand distribution of your best-earning content into a few places where it actually belongs, never anything coordinated or spammy. The automation race is on, and most publishers are running it wrong by bolting AI on top and shipping generic slop that AI itself then declines to cite, let alone pay for. The publishers who win combine proven systems, human authenticity, and the right technology at the same time.

Where to go from here

If your question is AI visibility specifically, which is the thing every one of these payment mechanisms rewards, the GEO Authority System at $499 runs the LLM Visibility Evaluation and hands you the playbook to raise your AI Citation Presence, so you become the source Google's pilot pays and a crawler will toll for. If you want the technology on-ramp, the Facebook Automation Machine at $397 is the n8n build that automates the repetitive publishing work, including the retrievability and freshness layer above, so your human judgment stays where the authenticity lives, and the $10K/Mo Profit Playbook at $197 is the strategy layer on top. And if you would rather have it run for you, PIB Consulting trains your team while you keep 100 percent of the upside, and Turnkey Management runs the pages on a revenue share with no money down. The move is not to chase a licensing headline that was never written for you. It is to build the owned foundation that makes AI a channel that pays you instead of a channel that quotes you for free.

Frequently asked questions

Can a small publisher actually get paid by AI in 2026?

Yes, in small and early ways. Google's AI contribution pilot pays some publishers through Search Console when their content shapes an AI answer, and Cloudflare Pay Per Crawl plus the RSL standard let any site put a price on crawler access. The amounts are modest so far. The Information reported that Google's roughly 100 pilot publishers are being paid less than one-tenth of 1 percent of their ad revenue in several cases, and all of these programs reward being a source AI systems actually want. A direct licensing deal like Reddit's remains a giants-only game.

How does Google's AI contribution pilot pay publishers?

It pays usage-based amounts, with no upfront fee, when your content contributes significantly while Gemini, AI Overviews, or AI Mode generate an answer. Content that only confirms a fact or gets linked after the answer is written does not qualify. Publishers in the pilot see a monthly earnings widget in Search Console with no breakdown of the calculation, can opt out anytime, and executives have called the math quite black box.

What is Pay Per Crawl, and is it different from RSL?

Pay Per Crawl is Cloudflare's system that answers an AI crawler with an HTTP 402 Payment Required response and your price, enforced at the network edge, so a crawler that refuses to pay is blocked. RSL, Really Simple Licensing, lets a site publish machine-readable licensing terms and supports models from free to pay-per-crawl to pay-per-inference, with the RSL Collective pooling licensing across publishers. The key difference is enforcement. Pay Per Crawl blocks non-payers at Cloudflare's edge, while RSL depends more on AI companies cooperating, and the two are designed to work together.

Should I block AI crawlers to protect my content?

Use the training-only control, not a broad block. Cloudflare's Disallow AI Training setting, reported September 15, 2026, tells crawlers not to train on your content while keeping them active for search, using Disallow rules for Google-Extended and Applebot-Extended. A full block now stops search crawlers too, which can cut your reader traffic. If you want payment rather than a closed door, layer Pay Per Crawl or RSL on top.

How do I actually get cited by ChatGPT or Google AI Overviews?

Answer one question completely and early on the page, build Topical Authority with depth across many linked pieces, be a clear named entity with a track record, and keep your content technically retrievable with accurate dates and structured data. Then track your AI Citation Presence over time and push more of what is already getting quoted. It is an authority position you build, and it is now the qualification for getting paid, not just cited.

Is AI sending publishers meaningful traffic yet?

Not much yet. AI platforms drive an average of only about 1 percent of overall web traffic across ten major industries. That is exactly why positioning now is cheap. The publishers building AI Citation Presence before the channel grows are the ones who will own it when it does, the same way early Google Discover publishers pulled ahead.

Key takeaways

• There are now four ways publishers earn from AI in 2026: licensing archives, charging crawlers a toll with Pay Per Crawl or RSL, Google's usage-based AI contribution pilot, and being the cited source that owns where the traffic lands.
• Direct licensing still pays only the giants. Reddit near $130 million a year, News Corp around $50 million, Wiley $49 million. A niche publisher earns effectively nothing from a custom deal.
• Google's AI contribution pilot pays through Search Console only when your content shapes an answer as it is generated. About 100 publishers are in it, several are earning under 0.1 percent of their ad revenue, and the calculation is opaque, but the qualification is telling, you must be the cited source.
• Pay Per Crawl blocks crawlers that will not pay your HTTP 402 toll, but many AI buyers will not pay per crawl, which is why Cloudflare launched a Pay Per Use beta in September 2026. RSL publishes licensing terms but leans on AI companies cooperating. All of them reward pages AI cannot skip.
• Use Cloudflare's Disallow AI Training control to opt out of training without blocking search. A broad block now stops search crawlers too and can cut your traffic.
• Every new payment rewards AI Citation Presence, so the durable path for a Facebook-first publisher is still get cited, then catch that traffic on the site and page where you already monetize.

Sources

• Search Engine Journal, Google tests paying publishers for AI answers via Search Console, September 2026: https://www.searchenginejournal.com/google-tests-paying-publishers-for-ai-answers-via-search-console/589414/
• Search Engine Land, Google tests paying publishers for using its content in AI Mode, AI Overviews and Gemini, September 2026: https://searchengineland.com/google-tests-paying-publishers-for-using-its-content-in-ai-mode-ai-overviews-and-gemini-488382
• Digiday, Google rolls out pay-per-use AI licensing program to publishers, 2026: https://digiday.com/media/google-rolls-out-pay-per-use-ai-licensing-program-to-publishers/
• Search Engine Journal, Cloudflare lets sites disallow AI training without blocking Googlebot, September 2026: https://www.searchenginejournal.com/cloudflare-lets-sites-disallow-ai-training-without-blocking-googlebot/589559/
• TechCrunch, Cloudflare launches a marketplace that lets websites charge AI bots for scraping, and Cloudflare's policy pushing AI companies to pay for content: https://techcrunch.com/2025/07/01/cloudflare-launches-a-marketplace-that-lets-websites-charge-ai-bots-for-scraping/
• The Register, New RSL spec wants AI crawlers to show a license or pay, September 2025: https://www.theregister.com/software/2025/09/11/new-rsl-spec-wants-ai-crawlers-to-show-a-license-or-pay/1292571
• Cloudflare, Introducing pay per crawl (HTTP 402 Payment Required), July 2025: https://blog.cloudflare.com/introducing-pay-per-crawl/
• Cloudflare, Pay Per Use: when AI uses your work, you should get paid, September 30, 2026: https://blog.cloudflare.com/pay-per-use/
• Search Engine Journal, Google AI pilot pays some sites under 0.1% of ad revenue (reporting The Information), October 1, 2026: https://www.searchenginejournal.com/google-ai-pilot-pays-some-sites-under-0-1-of-ad-revenue/591717/
• Publishers Weekly, AI and research drive gains at Wiley in fiscal 2026 ($49 million AI licensing): https://www.publishersweekly.com/pw/by-topic/industry-news/financial-reporting/article/100650-ai-research-drive-gains-at-wiley-in-fiscal-2026.html
• Inside Higher Ed, Taylor & Francis AI deal ($10 million initial Microsoft access fee), July 2024: https://www.insidehighered.com/news/faculty/research/2024/07/29/taylor-francis-ai-deal-sets-worrying-precedent
• Plagiarism Today, AI licensing comparison, RSL vs Pay-Per-Crawl, September 2025: https://www.plagiarismtoday.com/2025/09/11/ai-licensing-comparison-rsl-vs-pay-per-carwl/
• Digiday, Publishers cautiously count AI licensing as notable revenue, and In Graphic Detail, the state of AI referral traffic, citing Conductor data that AI platforms drive an average of 1 percent of overall web traffic across 10 major industries: https://digiday.com/media/in-graphic-detail-the-state-of-ai-referral-traffic-in-2025/
• LLM Pulse, Every AI content licensing deal mapped, 2023 to 2026: https://llmpulse.ai/blog/ai-content-licensing-deals/
• Google Search Central and Google public policy blog, guidance on AI features, Google-Extended, and the AI contribution pilot: https://developers.google.com/search