AI Search

SEO in the AI Era: What's Changing, and What Stays the Same

Google's AI Overviews and AI Mode can cut clicks to the top result by as much as 58%. You might think that's the end of SEO as we know it. Google itself says something different, and that's worth a closer look.

Wiktoria KolinkoWiktoria Kolinko
August 19, 20267 min read

A year ago, the question “does SEO still make sense in the AI era” sounded like a conference panel topic. Today it's a question clients ask outright, because they see traffic dropping in Google Search Console and don't know if it's seasonality or something bigger. If you're looking for a full action plan, we already have a complete guide to GEO on the blog. This piece does something different: it stops at the question GEO takes as its starting point, whether SEO still makes sense at all, given what the click data looks like.

The end of the click as the only measure of success

Let's start with the numbers, because a lot of unconfirmed estimates circulate on this topic. Ahrefs, analyzing 300,000 keywords from Google Search Console data, measured a 58% drop in CTR for the number-one result in December 2025 when an AI Overview appears on the results page, up from 34.5% measured eight months earlier. The drop isn't limited to the top position: the page in second place loses about half its clicks, and even the tenth position sees a decline close to 20%.

In AI Mode, Google's full-screen search mode with a generative answer, the situation is even more pronounced: around 93% of sessions end without a visit to any external site.

This isn't a temporary novelty effect. The share of queries that trigger an AI Overview at all keeps rising month over month.

What Google itself says about this

Google has weighed in on this directly, not through expert speculation. In its official developer documentation, Google states that good SEO practices remain relevant for AI features in Search (AI Overviews and AI Mode), and there are no additional requirements or special optimizations needed to appear in them.

That's an important sentence, because it cuts off speculation that you need to completely rebuild your site for some new, mysterious AI algorithm. At the same time, it doesn't mean nothing changes: what changes is what the same, long-established set of criteria evaluates, once generating an answer is on the table, not just a list of links.

What doesn't change: fundamentals and E-E-A-T

The technical basics of SEO, meaning crawlability, indexing, load speed and correct HTML structure, work exactly the same way regardless of whether a human reads the answer or a model generating an AI Overview does. Google clearly states this as the first condition for visibility in AI features.

What's gaining importance is the concept of E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness), a framework Google has developed for years in its guidelines for quality raters. Models generating AI Overview and AI Mode draw on the same ranking systems as classic search, so content that already performed well thanks to credibility and expertise also has an edge when it comes to being cited in AI answers. I laid out the full list of signals AI weighs when choosing brands for its answers in the GEO guide. Here it's enough to remember that this isn't a separate, new rulebook, just an extension of what you already know.

What actually changes: from ranking to citation

The biggest change is in what you're actually working toward. Classic SEO aimed at a ranking position: high meant visible, visible meant clicks. With a synthesized answer generated by AI, ranking still matters, because Google still has to pick the pages it draws from first, but the final step is no longer position, it's citation, whether the model pulls a sentence from your content that's worth quoting at all.

This difference has concrete consequences. A page can rank highly and still never get cited if its content is too generic for a precise statement to be pulled from it. And the reverse: a page outside the top ten sometimes gets cited in an AI Overview or a ChatGPT answer if it happens to contain a specific, easily quotable sentence on exactly what someone asked. Ranking and citation are two separate, overlapping evaluation systems, not one continuation of the other.

The exact same mechanism applies to link building, just moved down to the level of a single publication. I covered this in more depth in Link Building in the AI Era, where I show, among other things, how one specifically worded sentence about a brand started getting repeated by ChatGPT almost word for word. A link that used to be enough on its own now works best alongside a specific, citable sentence around it.

What this looks like in practice

Let's pull it together: the data shows clicks dropping, Google confirms the fundamentals haven't changed, and the difference comes down to whether your content can be cited, not just ranked highly. A few concrete things follow from that, worth changing starting now:

  • Write so a fragment can be pulled out and cited on its own. A specific answer in one or two sentences, near the start of a section, not diluted across a whole paragraph.
  • Don't drop the technical fundamentals. Speed, indexing and correct heading structure still determine whether Google even considers the page in the first place.
  • Build credibility that's visible on the page. An author with real experience, sources, update dates: exactly what E-E-A-T evaluates.
  • Treat visibility more broadly than just your Google position. Check whether your brand shows up in ChatGPT or Perplexity answers too, not only in search rankings.

If you want a step-by-step implementation plan, complete with metrics and tools for measuring AI visibility, the GEO guide lays it out in detail.

SEO is not dead. What we call success is changing

The data on falling CTR is real, and pretending otherwise won't help anyone. An agency that today sells a client nothing but a rise in Google rankings is selling an increasingly smaller slice of the whole picture.

That doesn't mean throwing out everything you know and starting from zero. Google itself confirms the fundamentals haven't changed: a fast, well-indexed site with credible content is still a necessary condition. What's changed is only the final step, one that didn't exist before at all, namely a model's decision on whether your content can even be cited as an answer.

It's the exact same logic I described when it comes to link building: what matters there isn't just a link's presence, it's whether the sentence around it can be cited. SEO in the AI era isn't a separate discipline next to classic SEO. It's the same foundation, with a new reader added, one that evaluates content by its own rules.

If you're wondering where to start building that visibility beyond your own site, get in touch. We'll match the format, a link, a mention or a sponsored article, to whatever your brand is actually missing.

FAQ

Najczęściej zadawane pytania

Krótkie odpowiedzi na pytania, które pojawiają się najczęściej w kontekście tego artykułu.

Yes. In its official developer documentation, Google states that good SEO practices remain relevant for AI features in Search and that no additional requirements or special optimizations are needed to appear in them. What changes is not the foundation but the final step: alongside ranking there is now a model's decision on whether your content can be cited.

Ahrefs, analyzing 300,000 keywords from Google Search Console data, measured a 58% drop in CTR for the number-one result in December 2025 when an AI Overview appears. Eight months earlier the figure was 34.5%. The second position loses about half its clicks and the tenth sees a decline close to 20%.

Ranking decides which pages Google considers at all. Citation decides whether the model pulls a quotable sentence from your content. A page can rank highly and never get cited if it is too generic, while a page outside the top ten can be cited when it holds a specific, easily quotable sentence.

It does, and it matters more than before. Models generating AI Overview and AI Mode draw on the same ranking systems as classic search, so content that already performed well on credibility and expertise also has an edge when it comes to being cited. It is not a separate rulebook, just an extension of what you already know.

Write so a fragment can be pulled out and cited on its own: a specific answer in one or two sentences, near the start of a section. Keep the technical fundamentals, build credibility that is visible on the page, and check whether your brand shows up in ChatGPT or Perplexity answers, not only in search rankings.

No contracts, no minimums

Ready to build your own success story?

Create a free account, browse 2,000+ domains with full SEO metrics, and place your first order today. Join agencies that are already scaling their organic traffic with TakeLink.