Google’s own advice on surviving AI search is not a technical roadmap. It is closer to a mindset shift, and it came from an unusually candid conversation between Google Search Advocate Martin Splitt and Nikola Todorovic, Director of Software Engineering at Google Search, on the company’s Search Off The Record podcast.

Splitt put the question directly: with AI Overviews and AI Mode now embedded in Search, what is Google doing to make sure the wider web ecosystem still thrives? Todorovic’s answer redirected the responsibility. There is no “magic wand,” he said, and no clean playbook for what an SEO should do differently in the new system. His guiding principle instead: site owners need to keep proving they deliver real value, because that is what keeps users coming back through Google in the first place.

“Provide value” is not a platitude, it is a description of the ranking signal

Taken at face value, “provide value” sounds like the kind of advice that fits on a motivational poster and tells you nothing actionable. But it is worth reading more carefully, given who said it and where he sits inside Google.

A Google engineer cannot spell out ranking mechanics in public, but he can point toward them indirectly. If Google’s systems reward sites that users are actively seeking out, then “value” is not vague; it is a description of the external, user-driven signals that already influence rankings. Todorovic extended the logic to any business model: subscriptions, products, publications, podcasts. None of them survive without value, AI-driven search or not. “If you don’t provide value, nobody’s going to buy your newspaper or book or listen to the radio or to the podcast,” he said.

That framing matters more now than it did five years ago, because it lines up with what Google has been doing structurally to the search results page. If you want the fuller picture of how AI Mode and AI Overviews are reshaping which sites get surfaced at all, see our coverage of how Google’s task-based agentic search is disrupting SEO right now.

Google’s real advice: master AI, don’t outsource your judgment to it

Todorovic also addressed the anxiety running through the industry, including inside Google itself, about whether AI will make jobs and businesses obsolete. His answer was not to avoid AI but to use it deliberately, as a way of adding value on top of existing work rather than replacing the work entirely.

He was specific about where the line sits. Using AI to bulk-generate content because it is cheap and easy “is not going to provide a ton of value,” in his words. Using it to sharpen grammar, tighten style, better understand your own data, or analyze competitors is a legitimate application. The distinction is between AI as a shortcut around quality and AI as a tool that sharpens work you were already doing well.

A practical AI prompt: reverse-engineering what your content actually answers

One useful application of that principle is a “reverse knowledge search”: feeding a finished article to an AI model and asking it to extract every question the text directly and completely answers, using only sentences that are actually present in the document rather than implied.

Run against a well-optimized article, this kind of prompt can surface the exact question a page answers best. In one documented case, the extracted question then ranked #1 in Google’s organic results, and separately ranked in Bing’s featured snippet, Bing News, and the top organic position. The content itself was not written or optimized with AI; only the grammar pass used it. The value of the exercise is diagnostic: it tells you whether a page is actually about what you intended it to be about, and whether it is drifting off-topic, without trying to reverse-engineer the search engine itself.

This lines up with a broader trend worth understanding if you are budgeting time for AI visibility work: AI search is not replacing traditional search so much as sitting on top of it, and the two are increasingly influencing the same ranking behavior. We break that dynamic down in how AI search is layering on top of Google, not replacing it, and in what Sundar Pichai’s own interview reveals about Google’s search direction.

The branding angle nobody expects from an SEO conversation

The “provide value” advice connects to a point that gets underweighted in most SEO discussions: brand recognition. At a past Search Central Live event, a conversation about ranking advice for a law firm’s website landed on the same idea Danny Sullivan later raised on stage that same day, that sites need to function as recognizable brands, not just optimized documents. Getting a business’s name into a potential client’s head before they search for it, through word of mouth and consistent branding, is itself a search advantage, because it feeds the kind of demand signal that AI-era ranking systems increasingly weigh.

None of this replaces the fundamentals. It reframes them. Businesses that can show, through user behavior and brand recognition, that they consistently deliver on what they promise are the ones AI systems and traditional rankings alike keep surfacing. If you want the operational version of that principle, our guide on what to do when AI Overviews are taking clicks from your site covers where to focus first.

What “using AI in the best possible way” looks like in practice

Translating Todorovic’s advice into a working checklist is more useful than repeating the phrase itself. A few applications stand out as genuinely additive rather than a shortcut around quality:

• Data and competitive analysis. AI is well suited to processing large volumes of ranking data, competitor content, or customer feedback faster than a human team can manually, freeing up time for the judgment calls that still require a person.
• Editing, not authoring. Using AI to catch grammar issues, tighten sentence structure, or flag unclear passages in content you already wrote preserves the expertise while improving the delivery.
• Content auditing. The reverse knowledge search prompt described above is a good example: it does not generate anything, it verifies whether existing content is actually about what you intended it to be about.
• Research acceleration. AI can summarize a competitor’s positioning or a market’s history quickly, but the strategic decisions built on that summary still need a person with domain expertise checking the output.

What does not hold up, by Google’s own account, is using AI to multiply content volume simply because generation is cheap. That approach produces exactly the kind of low-value output the “provide value” principle is meant to filter out, and it is unlikely to hold up as AI systems get better at recognizing thin, derivative material.

Frequently asked questions

What did Google actually recommend for AI search?

Google’s Nikola Todorovic said there is no single roadmap for adapting to AI Overviews and AI Mode. His core advice was to keep providing genuine value to users, since that is what continues to drive traffic through Google, and to use AI deliberately to strengthen existing work rather than to mass-produce content.

Should I use AI to write my content?

Google’s own guidance draws a line between using AI to bulk-generate content, which it says adds little value, and using AI to support work you are already doing, such as editing, data analysis, and competitive research. The latter is a legitimate use; the former is not likely to hold up.

What is a reverse knowledge search prompt?

It is a prompt that asks an AI model to extract the exact questions a piece of content directly and completely answers, based only on sentences present in the text. It is useful for verifying that a page is actually focused on the topic you intended, rather than drifting off-target.

Does branding actually affect SEO in the AI search era?

Brand recognition increasingly matters because AI-driven ranking systems weigh external, user-driven signals, including whether people are actively seeking out a specific brand by name. Strong brand recall built through word of mouth and consistent positioning feeds that demand signal.