Your paywall strategy used to be a trade between subscription revenue and search traffic. In 2026 it has a third variable: whether AI assistants can read, cite and send readers to your work at all. A July 2026 test by 5W and Everything-PR found that hard-paywall and metered publishers earned 0% of the AI citations in its sample, while open-web sources took 91.3%. A much larger study by Goodie, covering 31 million AI citations, shows the picture is more complicated, but it points the same way: content AI cannot reach rarely gets credit.
At the same time, publishers are leaning harder on direct revenue. Forbes told a Digiday event this month that its referrals were down about 45% year over year, and AdExchanger noted on October 22 that publishers hit by falling AI search traffic are turning to paid newsletters. So the question is no longer paywall or no paywall. It is which content you gate, which content you leave open to earn visibility, and how you make both pay.
Why Paywall Strategy Changed in 2026
Search engines learned to work with paywalls years ago. Google's structured data for paywalled content lets a publisher mark which sections are gated, which helps Google tell a legitimate paywall apart from cloaking, the practice of showing crawlers different content than readers. A subscriber-only article can still rank if it is marked up correctly.
AI answer engines do not work the same way. Most retrieve pages in real time, summarize them, and cite a handful of sources. If the retrieval step hits a login wall or an exhausted meter, the page is effectively invisible for that answer. The 5W study authors put it simply: an engine can only cite what it can read. That claim is their explanation rather than something their test measured, but it matches how retrieval systems behave.
What the Data Says About Paywalls and AI Citations
The 5W and Everything-PR index, released July 20, 2026, ran 40 queries through Claude with real-time web search and classified every cited source by access model.
- Hard-paywall publishers such as The Wall Street Journal, the Financial Times and Bloomberg: 0% of citations.
- Metered publishers such as The New York Times, The Washington Post and The Economist: 0%.
- Freemium publishers such as Forbes, CNBC and Fast Company: 8.7%, with 58% of that from Forbes alone.
- Open-web sources such as Reuters, AP, Wikipedia and trade publications: 91.3%.
Treat that zero carefully. An independent review published by Liamvi points out that the test used a single assistant, did not state how many citations it counted, and comes from a PR firm that sells AI visibility services. 5W itself says a second wave of 500 queries across five engines is still in development. The result is a strong signal, not a law.
The larger study: citations are scarce and concentrated
Goodie's study, which tracked 31 million AI citations from October 2025 to July 2026 across engines including ChatGPT, Gemini, Claude, Perplexity, Google AI Overviews and AI Mode, adds the context publishers need:
- News domains earned only 1.3% to 1.9% of all AI citations in any month.
- The top five publishers took 66% of news citations and the top ten took 84%. Forbes, which runs a freemium model, held 33% on its own.
- Traffic rank did not predict citation rank. The New York Times is first in visits but sixth in citations.
- AI models favored service content such as rankings, comparisons and explainers over straight news reporting.
Blocking, Licensing and Leakage: The Access Paradox
The Goodie data also shows why a paywall strategy cannot be designed around one AI company. The New York Times blocks every major AI user agent. Across the sample, ChatGPT cited nytimes.com zero times, Gemini zero times and Claude once. Yet the Times still collected 37,642 citations from Grok, 8,007 from Google AI Overviews and 4,303 from Perplexity, engines that, according to Goodie, offer publishers no practical way to opt out.
Licensing changes the pattern inside one lab. AP blocks every OpenAI crawler, yet 80% of AP's citations come from ChatGPT, because licensed content travels through the contract rather than the crawl. Google is running its own version: its paid News AI pilot now includes more than 400 publications, and Forbes joined this autumn in exchange for payment and performance insights.
There is a cost to locking everything down, too. Goodie cites Wharton and Rutgers researchers who found roughly a 7% decline in weekly visits within six weeks of publishers adding robots.txt blocks, measured in human browsing data rather than bot logs. Visibility and traffic are connected even when the link is not obvious.
A Paywall Strategy That Works With AI
The publishers that come out ahead are not choosing between open and closed. They are deciding what each piece of content is for. Here is a practical framework for an independent site or a niche publication.
1. Sort content by job: reach, trust or revenue
Explainers, definitions, comparisons and how-to guides are what AI engines cite most. Keep these open so they earn citations and new readers. Gate the content readers will pay for: original data, deep analysis, tools, templates, archives and timely insider coverage.
2. Prefer freemium or registration walls over hard walls
In the 5W sample, freemium was the only gated model that earned any citations, and Forbes dominates both studies with a model that leaves much of its content readable. A free registration wall also builds the email list you need for newsletters, which is where many publishers are now finding paying readers.
3. Mark up gated content correctly
Use Google's paywalled content structured data with the isAccessibleForFree property and a cssSelector that points at the gated section. That keeps your subscriber content eligible in search without being treated as cloaking. Make sure the open lead paragraphs answer the core question on their own, so an AI summary has something accurate to quote.
4. Write the open layer for citation
Put the direct answer near the top, use clear H2 questions, and include named data with dates. Our guide on how to get cited by ChatGPT and Perplexity covers the formatting details that help engines pick you as a source.
5. Decide your crawler policy per company, not in one switch
Blocking works only against engines that honor it, and a license can matter more than a block. Review which AI user agents you allow, which you block, and whether any licensing program fits your size. Our breakdown of AI content licensing explains the models available to independent publishers.
6. Measure citations as well as visits
Forbes says AI platforms remain a black box even with paid tracking tools. Start simple: test your key topics in several assistants each month, log whether you are cited, and track referral traffic from AI sources in your analytics. Without that baseline you cannot tell whether a paywall change helped or hurt.
7. Price the gated layer like a product
Subscribers pay for outcomes, not for access to the same articles they could find summarized elsewhere. Bundle exclusive data, a members newsletter and tools into one clear offer. Our guide to subscription revenue walks through pricing and retention basics.
Quick Answers
Do paywalls stop AI from citing your content?
Often, yes. In the 5W test, hard and metered paywall publishers received 0% of AI citations across 40 queries. Larger data from Goodie shows some engines still cite paywalled publishers heavily, so results vary by engine and by licensing deals.
What is the best paywall strategy for small publishers?
For most independent sites, a freemium or registration model works best: keep explainers and guides open to earn search and AI visibility, and gate original data, tools and premium analysis.
Will a paywall hurt my Google rankings?
Not if you implement it correctly. Google supports paywalled content when you use its structured data to mark gated sections, which separates a paywall from cloaking.
Should I block AI crawlers if I run a paywall?
Decide engine by engine. Blocking only affects companies that respect robots.txt, and research cited by Goodie found a roughly 7% drop in weekly visits after blocking. Weigh that against licensing options and the value of your open content.
The Bottom Line
A paywall strategy built for 2020 assumed search engines were the only gatekeepers. In 2026, AI assistants decide which sources readers see first, and they mostly cite what they can read. Keep the content that earns discovery open, gate the work people will pay for, mark it up properly, and measure citations alongside subscriptions. That way you stay visible and still get paid.
Sources
5W and Everything-PR, The Paywall Penalty (July 20, 2026); Goodie, AI citations and news publishers study (2026); Liamvi, review of the paywall penalty test; Google Search Central, structured data for paywalled content; Digiday, Forbes and the Google AI pilot (October 20, 2026); AdExchanger daily news roundup (October 22, 2026).