In this episode of Search Off the Record, Martin speaks with Nikola Todorovic (director of Software Engineering at Google Search) about how AI is changing Google Search. They discuss the evolution from traditional search to AI Overviews and AI Mode, how Google tests and launches search changes, and why query behaviour is becoming more conversational and complex.
Nikola also explains the role of machine learning in Search, how features are evaluated before launch, and what site owners and SEOs should focus on as AI becomes a bigger part of the search experience. If you work in SEO or web development, this episode offers a clear look at how Google approaches AI in Search and what it means for the future of search visibility.
Episode transcript → https://goo.gle/sotr109-transcript
Listen to more Search Off the Record → https://goo.gle/sotr-yt
Subscribe to Google Search Channel → https://goo.gle/SearchCentral
Search Off the Record is a podcast series that takes you behind the scenes of Google Search with the Search Relations team.
#SOTRpodcast #SEO #GoogleSearch
Speakers: Martin Splitt, Nikola Todorovic
source
The advice given by Nikola Todorovic sounds reasonable and timeless at first glance: website operators should simply continue creating genuine value for users — with personal experiences, tests, context, and human judgments. In reality, however, this recommendation is cynical and detached from reality for many small and medium-sized webmasters.
Google (and other AI systems) trains its models and generates AI Overviews and AI Mode responses largely on the basis of other people’s content. The search engine benefits enormously from the years of free work done by millions of webmasters, bloggers, and niche experts. At the same time, the traffic that previously flowed to their websites is increasingly replaced by summaries displayed directly on Google’s results page.
For large publishers with strong brands, this may be manageable. For small and medium-sized site operators, it is often existential. Why should anyone spend weeks testing products, writing personal experience reports, taking photos, and creating detailed guides if Google extracts the essence and presents it as its own summary within seconds — sometimes with a link, but often with little or no meaningful, clickable traffic? The compensation for this effort becomes unpredictable and tends to decrease steadily.
The podcast suggests that people “just” need to keep creating value. This ignores economic reality: No one produces high-quality original content for Google for free. Personal experiences, real tests, and nuanced human assessments are not created by magic — they require time, money, and effort.
When the return on investment systematically declines, rational webmasters will do one of two things: Dramatically reduce or stop creating new content. No longer make their content freely available (paywalls, login requirements, closed communities, newsletters, etc.).
We are already seeing this in many niches: high-quality sites stop publishing new articles, turn into pure archives, or disappear entirely. The much-praised “human touch” that Google claims to value so highly is increasingly disappearing behind walls — because it is no longer available “for free.”
Google still seems trapped in the old logic that the open web will remain an almost inexhaustible, free raw material. This was largely true in the 2000s and 2010s. In the age of generative AI, this assumption is naive.
If fewer and fewer webmasters are willing to provide high-quality content for free, the very source Google depends on will dry up. The consequences: AI Overviews and AI Mode will increasingly rely on older, less relevant, or lower-quality sources; search results will lose freshness, depth, and authenticity; the user experience will suffer in the long term — despite short-term impressive summaries.
Personal experience reports or well-founded human assessments can only be synthetically generated to a very limited extent. When the original sources dry up, exactly what Google currently markets as its differentiating factor compared to pure chatbots will be missing.
The appeal to “just keep creating value” is a classic Google response: it shifts all the responsibility and all the risk onto content creators, while Google itself maximizes the economic benefit. As long as the company is not willing to give webmasters a fair, predictable share of the value it generates (whether through better traffic, better monetization opportunities, or other models), this advice remains an empty, feel-good platitude.
The future of the quality web does not depend on webmasters continuing to work “anyway.” It depends on whether it is still worth it for them. Currently, there are many signs that for a large part of the decentralized web, it no longer is. Google will sooner or later feel the consequences of this misjudgment itself.
Berita bagus 🙂 . Terimakasih . Sukses Selalu 🙂
yeah yeah
everything but to add a normal funcional search on youtube
the current search is terrible
cant filter anything, no precise date range, no precise duration range, no sorting by newest
That's all good that you implement. How about fixing DMCA abuse first?
Google dorking is still the best way to find things. For example finding out there is a website that archives all of presidents truthsocial posts. Wouldn't be able to find it without doing Google dorking searches.
My results are becoming less and less relevant to the point I have to force google to find a certain phrase or word.
Seems they compare my search phrase to the highest bidder for companies that pay for result position and then I get totally irrelevant results. Sucks now.🫤
Pozdrav Nikola Todorovic.
Da li bi pomogao kolegi iz Srbije da dobije bilo kakav posao u Google-u?
U svakom slucaju, bravo za tebe. Svaka ti cast sto radis za njih i cestitam ti na uspehu u tvojoj karijeri!
AI Overviews are reshaping what SEO actually means. Watched this—Nikola's point about query behavior shifting conversational hit hard. We're seeing it live: clients asking "best way to" instead of "best [product] for". The rank-and-click model is dead. Now it's about being the source Google AI pulls from. Schema markup, content depth, E-E-A-T—these aren't optional anymore. They're how you stay visible when AI summarizes answers instead of showing 10 blue links. Real question: how many agencies are actually preparing their clients for this shift versus still chasing position 1?
I would be more enthusiastic about the AI in the search results if the output wasn't authoritative looking summaries that often include information that is out of context, from an unreliable source, or is just wrong. I've seen that from queries where I do already have some expertise, and by actually clicking the links to the source web pages (which often aren't original sources, but other sites regurgitating information from the original sources). And I've seen some funny responses where the AI Overview just makes stuff up (mostly when I search for something that isn't found clearly in the search results). Most people aren't going to click the links if it looks like their question was answered with authority.
Thanks good stuff.
Martin Splitt, How does the engineering team ensure that the AI prioritizes content from experienced and trusted creators over automated or AI-generated spam?