Google's Search Monopoly Was Built on a Deal. AI is Rewriting It.

Google's power was never built only on crawling the web.

It was built on a bargain that enough of the web accepted.

The bargain went like this: let Google crawl, copy, analyse and index your pages, and Google will send people back when those pages are useful. Publishers could sell advertising or subscriptions. Retailers could sell products. Service businesses could generate enquiries. Forums could grow communities. Search users got a map of the web. Website owners got a reason to keep publishing.

That was never perfectly fair. Search rankings were always an uneven distribution system. Google decided what appeared, in what order, and with which visual treatment. But the basic exchange was legible enough that most sites played.

The web has seen this pattern before. Universal search moved images, maps, video, news, and shopping into the results page. Featured snippets lifted answers out of articles. Zero‑click search made some informational traffic disappear before it reached the source. Social referral shifts taught publishers that a platform could create a traffic model and then change the terms. Mobile‑first design changed where the journey started and what users expected from a page.

Machine‑mediated discovery is another turn of that screw. The interface can now satisfy more of the user journey before the click.

The toll booth in the previous article is one response to that pressure. This article is about why the pressure exists in the first place.

This is the bit to inspect: can a page create value only when it gets the visit, or can it shape demand without analytics ever naming it as the landing page?

The Old Deal Depended on Traffic

Traditional search did not have to be altruistic to be useful. Google could monetise the results page because the results page was a gateway. Users searched, scanned links, clicked sources, and then did something elsewhere.

That elsewhere mattered.

The open web's economic model relied on destinations: articles, product pages, booking flows, service pages, documentation, comparison pages, forums, local listings, videos, recipes, support pages, and tools. Search engines extracted value from organising those destinations, but they also returned value by sending intent towards them.

The deal was therefore asymmetric, but functional. Google got the index and the advertising surface. Website owners got discoverability and referral traffic. Users got answers and routes.

AI Overviews, AI Mode, ChatGPT Search, Perplexity and other answer systems do not remove that pattern entirely. They still cite, link and retrieve. But they change the centre of gravity.

The user can ask a more complex question and receive a synthesis. They can compare options, follow up, ask for a recommendation, or narrow a choice without visiting several sources. In Google's own description of AI Mode, query fan‑out breaks a question into subtopics and issues multiple searches at once. Deep Search can go further, issuing many searches and creating a cited report.

That is useful for users. It is much less comfortable for websites whose value depended on being visited during the messy middle of research.

Search is Becoming an Answer and Action Layer

Google's AI Mode announcement is worth reading closely because it is not only about answers.

It describes AI Mode as a deeper search experience with follow‑up questions, links to the web, Deep Search, live capabilities, agentic actions, shopping, personal context, and custom charts. It gives examples of buying tickets, making restaurant reservations, booking appointments, and agentic checkout‑style shopping.

That is a long way from ten blue links.

The interface is no longer only helping the user find a destination. It is absorbing parts of the research, comparison and transaction workflow. The website may still be involved, but sometimes as a source, sometimes as an inventory endpoint, sometimes as a checkout destination, and sometimes merely as a cited footnote.

This matters because different website types used to monetise different parts of the journey.

Publishers monetised attention. E‑commerce sites monetised product evaluation and checkout. Local businesses monetised discovery and contact. Service businesses monetised trust‑building and enquiry. Review sites monetised comparison. Forums monetised long‑tail expertise and participation.

Machine‑mediated search does not remove all of those jobs, but it changes where they happen. The benefit moves towards interfaces that can answer, compare and transact. The risk moves towards sources whose economics depended on being visited during the middle of the journey.

The Same Source Can Be Used Without the Same Value Returning

A page can now be useful to an answer system without receiving the user's visit.

That is where the old bargain starts to wobble.

If a publisher explains a policy change clearly, the answer system can summarise the explanation. If a product page lists dimensions, compatibility, price, and reviews, an assistant can use those facts in a comparison. If a service page explains who a specialist helps, a retrieval system can recommend that specialist. If a local business page lists opening hours and facilities, an agent can use it to shortlist options.

Some of those outcomes still create value. A recommendation may be more valuable than a casual page view. A user who arrives after an assistant has already narrowed the options may convert better than someone browsing vaguely from search.

But the attribution changes.

Traditional SEO measurement could at least start with rankings, impressions, clicks, landing pages, and conversions. AI‑mediated journeys can split those signals apart. A source may influence a recommendation without receiving a click. A click may arrive later through direct traffic, branded search, a booking platform, an app, or a browser agent. A user may remember the recommendation but not the source. An agent may fetch the page, make a decision, and report only the outcome.

A page may shape demand without ever becoming the landing page that analytics credits for the conversion.

That makes the old traffic bargain harder to evaluate.

Zero‑Click Search Was the Warning Shot

The web was already moving in this direction before generative AI became the headline.

Featured snippets, answer boxes, knowledge panels, calculators, maps, shopping units and local packs all kept more activity inside the search results page. AI answers extend that pattern because they can compose a response across sources instead of presenting one extracted snippet.

Public data about the scale of the change is contested. Pew Research Center reported that users were less likely to click links when Google's AI summaries appeared in the results. Similarweb, Chartbeat, Authoritas, publishers and Google have all offered different measurements, caveats and objections. The exact number depends on query class, market, publisher size, search feature, device, brand strength, and methodology.

It would be lazy to claim one statistic settles the argument.

The safer conclusion is still serious: search interfaces are taking on more of the information work that previously happened after a click. For websites built around referral traffic, that is not a small UI change. It is a change in the distribution model.

Google's Problem is Also Its Advantage

Google is in a difficult position because its old product and its new product pull in different directions.

Classic web search depends on a strong open web. AI search depends on access to the same web, but may reduce the number of visits that make publishing worthwhile. Google can say AI search creates new opportunities for discovery, and sometimes that will be true. It can also be true that some publishers, comparison sites and long‑tail content owners lose traffic because the answer is now resolved earlier.

Both can be true at once.

That is what makes the economics difficult. AI search may improve user satisfaction whilst weakening some of the content supply it depends on. It may send higher‑intent traffic whilst reducing lower‑intent visits. It may expose lesser‑known sources in citations whilst concentrating the user's attention inside the answer interface.

The old deal was not perfect, but it had a visible return path. The new deal is still forming.

The Assumption That Breaks

It is easy to misread this as a copywriting problem.

The broken assumption is that the click remains the normal unit of value. It still matters, but it is no longer the only useful sign that a page shaped demand.

A page now needs to work as a destination, a source, a citation candidate, a comparison input, and sometimes an agent‑readable fact set. That does not mean flattening everything into FAQs. It means being clearer about the job each page does.

For an article, that might mean a stronger opening answer, visible authorship, current context, precise claims and links to supporting material. For a service page, it might mean naming the buyer problem, showing fit, proving experience and making the offer specific enough to be retrieved. For a product or location page, it might mean clean attributes, prices, opening hours, availability, facilities, schema, and stable URLs.

The article on GEO versus SEO makes the central point: GEO sits on top of a lot of competent SEO rather than replacing it. The difference is that the page now has to survive extraction, summary and comparison.

Measurement has to become less tidy because organic search traffic is no longer enough as a single proxy for visibility. The evidence is scattered across brand demand, assisted conversions, referral logs from AI platforms where available, server‑side bot traffic, citation monitoring, direct journeys, conversion quality, and changes in customer questions.

Some of that measurement will be messy. It is still better than pretending the old dashboard describes the whole journey.

The Deal is Not Dead, but It Needs New Terms

The web still needs search and discovery. Retrieval systems still need sources. Users still need somewhere to go when an answer is not enough. Commercial websites still need pages that explain, persuade, transact, support, and prove.

So the old deal is not dead.

It is being renegotiated through infrastructure, licensing, crawler controls, structured data, AI search interfaces, publisher programmes, legal pressure, and user behaviour.

Cloudflare's Pay Per Crawl and AWS WAF AI Traffic Monetization show that part of the renegotiation has moved into the infrastructure layer. The old crawl‑for‑traffic bargain used to be handled mostly through robots.txt, search policy and commercial tolerance. Now edge services can ask whether a crawler is allowed, blocked, charged, or asked to complete a machine payment flow before content is served.

RSL is another version of the same pressure. OpenAI's crawler separation is another. Google's Google-Extended is another. Content licensing deals are another. So are lawsuits. So are technical SEO and GEO workflows that make content easier to cite without surrendering ownership entirely.

The harder question is not whether Google remains important. It does. The question is whether the site is built for an older search journey where the click was the main event, or for a newer journey where the site may be visited by a user, summarised by an answer engine, evaluated by an agent, or used as evidence in a recommendation.

Wrapping Up

Google's search power was built on a bargain: let us organise the web, and we will send users to you when your pages are useful.

Machine‑mediated search rewrites that bargain because the interface can now do more of the work itself. It can research, compare, summarise, narrow, and sometimes act. That does not make websites irrelevant. It changes what websites are asked to be.

The next version of search visibility will not be measured only by rankings and clicks. It will be measured by whether a site can be discovered, trusted, summarised, attributed, recommended and converted from journeys that may not look like the old search funnel.

The website still matters. The route to it is no longer as direct.

Historically, websites were designed as destinations for people arriving from discovery systems. Increasingly, they also have to prove their meaning before the user sees a result.

Key Takeaways

• The old search bargain exchanged crawler access and indexable content for referral traffic.
• AI search weakens that bargain by satisfying more research, comparison and decision‑making before the click.
• Google's AI Mode shows search moving towards answers, research, shopping, bookings, and agentic actions.
• Traffic‑impact numbers vary by source and methodology, but the direction is clear enough to treat as a strategic change.
• Websites now have to work as destinations, citations, summaries, comparisons, and agent‑readable fact sources.
• Measurement needs to include assisted journeys, citation visibility, bot traffic, brand demand, and conversion quality, not only organic clicks.