The old web deal was ugly but easy to read. You let search engines crawl, they indexed the world, and some fraction of people clicked through. Against that attention you sold ads, subscriptions, services, software or reputation. Traffic was a crude proxy for value, but at least you could see it.

Generative search breaks the deal in two places. Machines can use your pages without sending a human back, and readers can get a good enough answer without ever meeting the source. Traffic falls, and your instruments stop seeing much of what happens, so a business can stay useful to its market while becoming hard to observe from the inside.

The fights over "Google Zero," AI Overviews, crawler blocks and pay-per-crawl are therefore an early warning for any founder who uses public writing to create demand, and newspapers are only the first to feel it. Once your public shelf becomes training data, citation fodder or answer material with no reliable path back into a relationship, publishing can look successful while the business quietly loses its audience.

Traffic is still worth counting, but the better question is which intermediary turned the work into demand, proof or revenue, and what you kept from the exchange.

Readers now reach sources through gateways

The Reuters Institute's 2026 Digital News Report describes a long move away from going straight to a publisher and toward third-party gateways such as social, video, aggregators and now AI chatbots. Weekly chatbot use for news is still smaller than search or social, but the people who use it are not casual. They use assistants to compile, compare, verify and translate, which is high-intent behavior with a machine sitting between the question and the source.

The small numbers understate the shift because AI intermediaries sit at the moment a question is formed. People no longer start by browsing a publication; they ask something that answers, cites, summarizes and suggests the next step. Your site becomes one input among many to an answer assembled somewhere else.

Pew's 2025 browsing-data work found that Google users who saw an AI summary clicked traditional results less often than users who saw none, and clicks on the links cited inside the summary were rare in that sample. Google says AI search creates new discovery surfaces and that Search Console and site-owner controls still give publishers visibility. The two claims are compatible, since new surfaces can open while the old traffic accounting gets weaker.

Why the theft argument is incomplete

The easy reaction is moral: platforms are stealing content. For some rights holders that is the whole case, and fair enough. For a small strategic publisher, consultant, founder or analyst, I think it leaves out the harder, operational question of which assets should stay open, which should be blocked, licensed or paywalled, and which pieces of work create a relationship you actually own.

Cloudflare's 2025 pay-per-crawl idea is interesting because it drops the old choice between being indexed and being invisible. Some crawlers get free access, some pay, some hit a wall. That is a richer policy, yet a crawl fee or a citation replaces none of the customers, subscribers, sales conversations or retained audience that the business depends on.

The common founder failure is to optimize one lane and call it strategy. SEO chases visibility, legal fights scrapers, product ships a generic chat feature, growth pushes newsletter signups, and finance asks where the money is. In a market mediated by AI, those teams are all working on the same audience problem and should be judged against one scoreboard.

What to track

Keep a simple record of how public work moves through machines, platforms, readers and conversion paths. Keep it plain enough that someone actually maintains it, with enough detail to change what you publish, open or block.

Area What to record Question Rule
Crawl Access Which bots can fetch pages, which are blocked, which are rate-limited, and which have commercial terms. Is the company giving away durable source material without a return path? Open commodity marketing pages; govern source-rich analysis, datasets, tools, and templates.
Answer Presence Whether AI search and chat systems mention the work, summarize it accurately, or omit it from relevant queries. Are assistants turning the firm's point of view into demand or into anonymous background knowledge? Write pieces with distinct claims, named frameworks, and source trails that are worth citing.
Referral Quality Clicks from search, AI surfaces, social, newsletters, direct links, partners, and private forwards. Which channel sends readers who behave like prospects rather than passers-by? Rank channels by reply, subscription, meeting, share, and repeat-reading behaviour, not sessions alone.
Owned Relationship Email, Substack, RSS (really simple syndication) feeds, community, customer-relationship management (CRM) records, manual replies, customer list, and buyer conversations. What survives if search traffic falls by half? Every high-performing public piece should offer one explicit relationship path.
Conversion Right Who controls payment, delivery, access, subscriber identity, refunds, and renewal communication. Can the company monetize the reader without asking a platform for permission at the decisive moment? Use platforms for convenience, but keep the paid offer portable enough to move.
Proof Loop Partner feedback, reader replies, consultation requests, paid-pack downloads, citations, and sales objections. Did the article teach the company something commercially useful? Keep publishing only where the feedback improves doctrine, product, positioning, or revenue.

Deciding what stays open

Different content has different jobs. Pieces that set out your point of view want to be found, cited, forwarded and argued with, so let them travel. Paid worksheets exist to turn attention into a buyer relationship, so keep them out of answer boxes. Source-heavy research may deserve public excerpts, while working notes, scoring sheets and client-ready templates may belong behind a subscription, a direct sale or a partner review.

What you are managing is a set of retained options, and a page view is only one of them. Publishing on GitHub Pages, Substack, LinkedIn, X, YouTube or a private memo list buys you some mix of reputation, search presence, citation presence, partner trust, inbound deal flow, paid subscribers and product feedback. If AI search weakens one of those, another has to be strengthened to make up for it.

Public writing should therefore become more distinctive. Generic explainers written for search engines are easy for answer engines to absorb and easy for readers to forget. Distinct claims, decision rules, named frameworks, worked examples and proprietary checklists give someone a reason to remember where an idea came from. An AI system can summarize the text, but it has a harder time replacing the trust built when the same framework keeps turning up in real decisions.

What the dashboards miss

Google's publisher guidance is right as far as it goes: AI features show up in Search Console, and the old preview controls still matter. The mistake is to treat the old analytics stack as complete. If an AI answer absorbs a weak click, you may not care. If it keeps a serious reader from ever finding the author, the loss is quieter and more expensive.

So put softer evidence next to the dashboards. Did a founder mention the piece on a call, or a partner forward it inside the firm? Did someone subscribe after the third memo rather than the first? A company asking for the worksheet instead of another free essay counts, and so does a chatbot answer that attributes the work accurately enough to support trust.

Keep the traffic metrics, but rank them lower, because falling traffic is a symptom and the risk underneath it is losing control of your audience. Managing that risk means working on discoverability, relationship capture, paid conversion and reusable proof at the same time.

Start with your ten defining pages

Take the ten public pages that most define the company's point of view. For each one, fill in the table: crawler policy, likely role in AI answers, direct, search and AI or referral traffic where you can see it, newsletter or feed capture, replies, sales use, paid-asset tie-ins and partner feedback.

Then give each page a job. A few stay open and deliberately quotable. Another set should point harder toward a newsletter, feed or other direct relationship. A third becomes public previews for paid worksheets. Generic pages may be consolidated, and expensive source work with no conversion path may be blocked or licensed.

Counting published essays tells you little. A stack of them matters only if it builds into a recognizable position, distribution, trust and a commercial right. AI search makes this urgent because it separates attention from arrival: someone can learn from your work without ever touching your site.

That is survivable if you know what you want back from each asset. Work meant to build authority may need a citation, a launch note may need a click, a partnership may need a reply, and a paid instrument may need a download.

Source Notes