AI is changing the rules of search: fewer clicks, new data to analyze, and technical issues that can make you lose visibility on Google (and beyond)
When you check Google Search Console, do you wonder where your organic traffic has gone?
The answer could be simpler than you think: people ask AI, get the answer there, and the click remains nothing more than a distant mirage.
Welcome to a new episode of SEO Confidential, the series where we interview some of the most authoritative minds in SEO and GEO from around the world.
Today, we’re joined by Oleg Korneitchouk, an SEO and digital marketing consultant with more than 10 years of experience, specializing in link building, website architecture, and content strategies designed to deliver tangible results for businesses.
In this interview, we tackle the hottest topics in search, and Oleg’s answers couldn’t be clearer. We discuss the issues that matter most to us (and to you), including just how low—embarrassingly low—the CTR of AI Overviews really is, how much publishers risk being taken over by AI big tech companies, and how much valuable data on generative search is already available in Bing Webmaster Tools compared with Google.
We’ll also give you a preview of a technical detail that could affect you directly: a recent change by Google in how it handles JSON-LD markup that, in the worst-case scenario, can lead to a complete loss of visibility.
Get ready: after reading this interview, you may look at your SEO (and GEO) strategy differently.
“I think Google doesn’t show click data because the CTR generated by AI is so low that it’s embarrassing,” Oleg told us
How important is it for LLMs to have access to reliable editorial sources, and why are they willing to pay publishers for this content?
Very important. LLMs are machines that work through repetition, so when the world changes, the only way they can “know” about it is for someone to publish content somewhere accessible to LLMs, allowing them to add it to their “knowledge.”
Given the enormous amount of false and misleading information circulating online, LLMs need reliable sources of facts to distinguish truth from fiction. Traditionally, journalism and publishing have represented precisely this source of verified and trustworthy facts.
Do you think this type of agreement could become a sustainable model for publishers, or does it risk creating a new form of dependence on AI platforms?
I think it could make sense for some publishers, but not for everyone. The more proprietary and exclusive your data or information is, the less you should allow LLMs to access it. It would be more appropriate to leverage that exclusivity to build and grow your audience.
In general, I think many publishers would benefit from making agreements with LLMs part of their revenue streams, because this is a growing marketing channel.
If publishers have to negotiate individually with large AI companies, how can they defend the value of their content when they have much less bargaining power?
The risk is that small publishers will be forced to accept very low compensation simply to avoid losing visibility and access to AI platforms. How can this imbalance be avoided?
That’s the billion-dollar question, and I don’t have a satisfactory answer. If we look at what is happening around the world, governments have stepped in to protect news and media publishers from this imbalance, because for any democratic country, having an informed and aware population is important. This is the path that faces the least resistance.
Ideally, the industry should act as a united front and put pressure on policymakers to introduce changes capable of protecting a business model based on producing and distributing information. Information wants to be free, but finding, verifying, and distributing stories that can benefit society comes at a high cost. If publishers don’t put the interests of the industry above their own profits, AI companies will continue to take them over one by one.
Google has changed the way it interprets JSON-LD with double-encoded characters. What problems can websites using this type of markup encounter?
Worst-case scenario: a complete loss of visibility because Google will not be able to understand which queries the content is relevant for and where it should rank.
How can an SEO quickly check whether their site has been affected by this change, and what fixes should they implement?
I quickly put together a small tool to check for double encoding in JSON, which allows you to check any page for potential issues.
At the very least, I recommend checking the homepage, where the Schema.org Organization markup is present, as well as an article page with one or more special characters in the title. These are the pages most likely to contain double-encoded entities.
Google says it has a well-funded spam team dedicated to fighting abuse. Why, despite this, does so much spam content continue to rank highly in search results?
It’s a very difficult problem to solve, made even more complex by the fact that LLMs have brought the cost of producing well-optimized content down to zero.
Google is forced to rely increasingly on other signals, such as trust, authority, authorship, and user experience, and it continuously changes its ranking criteria to find the right balance and achieve its goal: reducing spam without penalizing legitimate organizations.
In my view, there are too many economic incentives to produce spam for it ever to be eliminated completely: strategies for manipulating search results are bound to increase. Google should try to increase the risk-to-reward ratio of spam, making these practices less attractive.
Google now shows in Search Console how often a site appears in generative features, but without providing data on queries or clicks. How useful is this report really for understanding whether AI visibility is generating tangible value for a website? Why do you think Google still doesn’t show which queries generate impressions in its AI features or how many clicks websites receive?
I don’t know whether Google has issued an official statement on its decision not to provide click and query data, but it would probably say that the decision is related to privacy concerns, given that queries directed at LLMs can contain much more personal information.
Personally, though, I think they don’t show click data because the CTR generated by AI surfaces is so low that it’s embarrassing and could become a weapon in the hands of the information industry, which could use it to sue Google or renegotiate its agreements with the company.
They don’t show queries because spammers could use this data to improve the relevance of their content for those queries, while Google prefers website owners not to produce content solely for the purpose of ranking in search results.
At the moment, the best way to use the generative AI data available in Google Search Console is to analyze the pages that are cited most often and look for patterns that explain why some pages are more likely to be used as sources than others. For example, you can check whether certain types of content perform above the site average.
I think Bing Webmaster Tools is doing a much better job of sharing generative AI search data with webmasters. It provides click and impression data, as well as reference queries that do not contain personal information, search intent, the main topic, and citation share.
This data can provide a starting point for understanding which entities and queries to optimize content for in order to gain greater visibility. So my advice is simple: if you haven’t already created a Bing Webmaster Tools profile, do it now!
Your brand needs to become THE answer AI gives
So, what should you take away from this conversation with Oleg?
Three things, in my opinion, that you should write down right away.
First: the CTR of AI Overviews is so low that even Google would rather not show you the numbers.
Second: if your industry relies on editorial content, the bargaining power you have with AI platforms is being decided now, not a year from now.
Third, and this is the most practical piece of advice you can take away today: go to Bing Webmaster Tools and look at the generative search data that Google still doesn’t give you. Queries, clicks, impressions: everything you need to understand what to focus on. It’s free and available right away.
I’ve made up my mind: becoming the answer AI gives should be your number-one goal.
LLMs amplify your authority by drawing on what has already been written about you elsewhere on the web in a clear and verifiable way. Let me be blunt: if your brand isn’t mentioned, it simply doesn’t exist in the conversation that matters.
That’s why visibility in 2026 is built upstream, through sources, structured data, and content designed to be read, understood, and referenced by machines before it is even seen by a user.
A huge thank you to Oleg for this incredibly enjoyable conversation and for such clear, direct answers, without unnecessary technical jargon: exactly the kind of content you can understand at a glance, even if you don’t live and breathe SEO every day.
And you, will you be joining us for the next episode?
I’ll see you next Monday, right here on SEO Confidential, with a new guest ready to tell you where search is headed.
See you Monday!
