When Google AI Ghosts the Experts Who Fed It

A Search Engine World take on the MediaPost report, what it says about AI Overviews, and why site owners should pay attention right now.

Google keeps asking publishers, creators, and SEO professionals to trust the process. The problem is that trust has become a one-way transaction.

A recent MediaPost article captured the frustration plainly. Independent publishers were invited to Google’s Web Creator Summit, asked to provide detailed feedback on search and AI Overviews, and then left with little to show for it. The people who helped build the open web were asked for input, handed over their data, explained the damage, and then got silence back.

That silence matters because this is not an abstract product story. It hits real businesses, real payrolls, and real sites that spent years building topical depth, trust, and original reporting.

What happened

By now the pattern is familiar. Google invites a small group of creators or publishers into the room. The company listens. Notes are taken. People leave thinking there is at least a chance that detailed examples, traffic data, and user experience concerns will shape the next phase of search. Then the machine keeps rolling.

According to reporting around the summit, some attendees submitted long, detailed write-ups explaining how AI Overviews and ranking changes were damaging their sites. That is not casual feedback. That is unpaid consulting from the very people whose work powers search results in the first place.

Publishers are not upset because change happened. They are upset because they were asked to help explain the fallout, and the fallout kept growing anyway.

Why this story matters to SEOs and site owners

This is bigger than one summit and bigger than one article. It gets to the core bargain that defined search for years. Publishers made content. Search engines indexed it. Users clicked through. Sites earned attention, leads, subscriptions, ad revenue, and brand recognition.

AI Overviews change that bargain. The engine still needs source material, but it no longer needs to send the visitor back to the source at the same rate. That is a hard shift for any business model built on organic discovery.

The ugly part: the sites most likely to feel the squeeze are often the ones with the deepest first-hand knowledge, the strongest niche expertise, and the least room to absorb traffic shocks.

The traffic problem is no longer theoretical

Pew Research found that when users encountered an AI summary in Google results, they clicked on a traditional search result in 8% of visits. When no AI summary appeared, that happened in 15% of visits. In plain English, the click path to publishers got a lot thinner once the AI answer box showed up.

That does not mean every query is doomed, and it does not mean publishers should stop publishing. It does mean the old assumption, publish good work and search will send the audience, no longer holds the same way it did even a couple years ago.

Old search compact	AI answer era
Publishers create the material	Publishers still create the material
Search engine points users to sources	Search engine often answers before the click
Visibility and traffic are linked	Visibility can rise while visits fall
Feedback loops improve ranking and content strategy	Feedback loops are weaker and harder to trust

What Google gets wrong here

The central mistake is not just product design. It is relationship design.

If you ask publishers for detailed feedback, then keep shipping features that reduce clicks, compress attribution, and weaken the business case for independent publishing, people stop hearing partnership. They hear extraction.

That is why this story lands so hard with SEOs. We have all seen versions of it before. Google says it wants better content. Independent experts publish better content. Large systems flatten the distinction, scrape the value, and send back less reward.

At some point the message to site owners becomes brutally simple: feed the system, but do not expect the system to feed you back.

What site owners should do now

Reduce dependence on a single search channel. Email, direct traffic, communities, forums, podcasts, video, and brand search now matter more than ever.
Double down on material that is hard to summarize cleanly. Original testing, first-hand experience, proprietary data, tools, calculators, and live inventories still create friction for copycat answers.
Track queries that trigger AI Overviews and compare them against click and conversion loss, not just rankings.
Build pages that earn visits after the summary. Strong visuals, deeper comparison layers, downloadable assets, niche tools, and current data can still pull users through.
Stop treating impressions as victory. If sessions, leads, and revenue drop, the pretty chart is costume jewelry.

The real risk for the web

If expert sites lose the incentive to publish, the search ecosystem gets thinner. Fewer specialists will invest in tests, guides, comparisons, and deep reporting if the answer engine absorbs the value and returns a fraction of the traffic.

That creates a strange loop. AI systems need credible source material. But if the economics of producing that material collapse, the source layer gets weaker. The machine starts eating its seed corn.

That is why the MediaPost piece matters. It is not just another complaint about rankings. It is a snapshot of a broken exchange between platforms and the people who made those platforms useful.

Final thought

Google can call publishers into the room. It can ask for memos, examples, and goodwill. But if the result is silence followed by another expansion of AI answers that reduce publisher traffic, the message is clear enough without a press release.

The experts were good enough to train the system, but not important enough to hear back from it.

For site owners, that means one thing. Build for search when it pays, but build beyond search so your business is not standing on a trapdoor.

Sources: MediaPost, Pew Research Center.