May 15, 2026  ·  8 min read  ·  Google · AI Search · SEO

Google's AI Search Needs the Open Web More Than It Wants to Admit

Google suddenly wants more links and subscription hooks inside AI Overviews. That is not a gift to publishers. It is the clearest sign yet that AI search cannot just keep strip-mining the web forever.

A dark editorial illustration of an AI search interface with glowing citation links reconnecting to independent websites across the open web

Google is suddenly acting like links matter again.

That is the real story underneath this week's AI search update. According to Ars Technica, Google plans to make websites more prominent inside AI Overviews and AI Mode, and it is looking for publishers to test a subscription integration that ties a reader's site subscription to their Google account. Google says early testing showed people were much more likely to click when a site they already subscribe to appeared inside the answer.

Read that slowly. The company that spent the last year training people not to click is now experimenting with ways to make clicking happen again.

This is not generosity. It is dependency management.

Google knows AI search has a contradiction sitting at the center of it. The product gets better when it can absorb more of the web, summarize more of the web, and keep people inside Google's interface longer. But the input layer for that product is still the open web itself. If the websites creating the useful stuff keep losing traffic, subscriptions, and ad revenue, the answer engine eventually starts eating its own food supply.

AI search wants the web to behave like free raw material. The problem is the web is an ecosystem, not a quarry.

That is why this move matters. It is the first real tell that even Google understands zero-click AI search has limits.

Up to 90% click drops Penske Media has alleged that searches with AI Overviews can slash clicks by as much as 90 percent. Whether every query is that brutal or not, the direction is obvious: the answer box keeps more of the value for itself.

This is not a course correction out of kindness

Google would love for people to read this as balance returning to the system. Look, more citations. Look, better visibility for publishers. Look, maybe even subscription-aware links. Problem solved.

No. What this looks like to me is a company realizing that the old extraction ratio may have been too aggressive.

AI Overviews and AI Mode were always built around the same basic incentive: compress a messy web into a neat Google-shaped answer layer. From a user-experience standpoint, I get the appeal. People want fewer tabs, faster answers, less friction. From Google's standpoint, it is even better. More time on Google, more habit formation around Google, more leverage over the distribution layer, more ways to turn the web into a product feature instead of a destination.

But the moment Google starts openly talking about improving publisher prominence inside those AI surfaces, it is admitting the surface itself was not enough. If users were perfectly happy never leaving the answer box, Google would not be hunting for new click-through mechanics. If publishers were not getting squeezed hard enough to complain, threaten lawsuits, and push regulators, Google would not be testing subscription-linked visibility.

The update is a signal. The pressure got real.

The subscription API is the biggest tell

The most interesting part of the Ars report is not just "more links." It is the subscription integration.

That feature says something important about where Google thinks the valuable web is heading. Not toward anonymous commodity pages. Not toward infinite SEO mush. Toward content that has an actual relationship with a reader.

That matters because subscriptions are one of the few clean signals left that a website is producing something people value enough to pay for, log into, or maintain an ongoing connection with. If Google wants to surface those sites more aggressively inside AI answers, it is basically conceding two things at once:

That is a big deal. It means the answer engine era is already running into a quality problem. The more Google collapses the web into summaries, the more it risks degrading the incentive to make the original stuff worth summarizing in the first place.

And once that happens, everybody loses. Publishers lose revenue. Readers lose depth. Google loses fresh, trustworthy material. The whole thing starts feeding on thinner and thinner derivative sludge.

I wrote in my post on topical AEO and SEO that the game now is not just ranking pages. It is becoming source material. This update reinforces that. Google still wants sources. It just wants them on terms that preserve Google's convenience.

A moody illustration of websites and publishers feeding an AI search engine through glowing data lifelines while links return value back to the web

Google cannot keep pretending the traffic problem is imaginary

One of the more irritating parts of this whole era is that Google keeps acting like the traffic collapse conversation is overblown. Maybe in some aggregate metric deck somewhere, impressions are up and user satisfaction looks fine. Cool. That does not mean the web underneath feels healthy.

Publishers are not imagining the shift. Independent site owners are not hallucinating the drop in referral traffic. Searchers are not wrong when they say they are seeing more answers and fewer reasons to click. The system changed. Everybody can feel it.

And honestly, this is the part I think people in tech still understate: the web is not just a content warehouse. It is a production system. If the economics of publishing stop working, quality drops later, not instantly. There is a lag. That lag tricks platforms into thinking they can keep squeezing forever.

They cannot.

You can get away with starving the upstream for a while. Then one day the upstream gets worse, narrower, more locked down, or more synthetic. At that point the answer engine has a quality crisis that looks mysterious from the outside and completely predictable from the inside.

That is why this week's "more links" shift feels so important. It is not just a UX tweak. It is Google quietly acknowledging that an AI answer layer still needs living pages, living brands, and living businesses underneath it.

What publishers and marketers should take from this

If you run a site, this is not the moment to trust Google more. It is the moment to understand what Google is telling you accidentally.

The takeaway is not "great, links are back." The takeaway is that direct audience and original value matter even more now because the platforms have finally started to notice they cannot fake the whole stack by themselves.

If I were building for the next year, I would focus on a few things:

The ugly truth is that AI search is not going to reverse course out of principle. It will only rebalance when starvation starts threatening the product itself. That means publishers need to stop waiting for fairness and start optimizing for durability.

The open web is still the real product

The funniest thing about AI search is that it keeps trying to present itself as the final interface, while depending completely on a system it did not build and cannot replace.

Google can wrap the web in answers, reorder it, summarize it, personalize it, and measure it to death. Fine. But it still needs people out there making things worth finding. It still needs forums, blogs, docs, local sites, niche publishers, reviewers, researchers, and weird obsessives who care enough to publish specific knowledge on the open internet.

That is the part the AI layer never gets to outgrow.

So when Google starts putting more links back into AI search, I do not see a benevolent platform finally remembering publishers exist. I see a company discovering that the web is not infinitely extractable.

That is a much more interesting story.

And if Google really has learned that lesson, good. It needed to. But the real test is not whether the answer box gets a few prettier citations. The real test is whether the economics of creating useful pages start to recover in any meaningful way.

Until then, I am not calling this a fix. I am calling it what it is: a giant platform noticing the machine still needs something alive underneath it.

← All posts
🌲

Forest SD

Tech, AI, digital culture. San Diego. Writing about what is actually happening, not what the press releases say.