Google Isn't the Only Place People Search Anymore

TL;DR: On May 20, Google made AI Mode the default and called it the biggest change to Search in 25 years. The practical effect for you: Google now answers questions instead of sending people to your page, and a measurable slice of your audience has drifted off to Kagi, Brave, Perplexity, and ChatGPT. The job stops being "rank number one on Google" and becomes "be the source that gets cited, on every engine your readers use." Start by checking whether any AI engine quotes you today.

Google stopped being a search engine on May 20

I run an SEO company with my wife. So when Google spent its I/O keynote on May 20 telling the world it had reimagined the search box and called it the biggest upgrade in over 25 years, I took it personally. The thing we help people optimize for had just changed underneath us, live, on stage.

Here is what actually shipped. AI Mode is now the default for everyone, running on Gemini 3.5 Flash. Google says it crossed a billion monthly users within a year of launch, with queries more than doubling every quarter. TechCrunch ran the headline "Google Search as you know it is over," and for once the headline was not the exaggeration. Time framed it as a change in how people use the internet, which is closer to the real stakes.

The mechanic that matters to anyone with a website: the default Google result is now an answer, not a list of ten links. Google reads the pages, synthesizes, and hands the user a paragraph. Your page is somewhere in the citations, if you are lucky, in small text most people never tap. (I should be honest here. I was one of the SEO people who, two years ago, told clients AI Overviews were a passing experiment. I was wrong, and I have eaten that opinion in public a few times since.)

The number that should scare you a little

"Google sends less traffic now" is easy to say and easy to wave off. The data is harder to wave off. Three studies, run by people who do not sell SEO software, all point the same direction.

Read those together and the picture is plain. When Google answers the question itself, roughly half the clicks that used to reach the open web evaporate, and the ones that survive almost never go to the sources the AI quoted. The user got their answer and rated the trip a success. They just never visited you to get it.

One caution before you panic. These are population averages, and intent decides your real exposure. AI Overviews concentrate on informational searches: Ahrefs, analyzing 146 million SERPs, found that 99.9% of the keywords triggering an overview are informational, while transactional queries trigger one far less often (and although that gap is narrowing, it is still wide today). So a "what year did X happen" page bleeds almost all its clicks into the answer box, while a "best CRM for small agencies" page, where the searcher still wants to compare and decide for themselves, holds up better. Pull your own Search Console data and segment by query intent before you assume the 38% applies to you. For an informational-heavy blog it is worse than 38%. For a transactional storefront it can be barely visible today, though it is climbing.

Where your audience actually went

The other half of the story is the one the I/O keynote did not mention: people are leaving the single Google box. Slowly, then in clumps.

StatCounter put Google at about 90% of global search in April 2026. That sounds untouchable until you read the trend line. Google's all-device share dropped roughly 1.5 percentage points year over year, the largest single-year erosion since 2009, and on desktop it sits near 79%, the lowest in two decades. The lost share scattered: a bit to Bing on the back of Copilot, a bit to AI assistants, a fraction to Yandex.

Then there is the long tail of people who pay to leave. Kagi, the subscription engine that shows zero ads, went from about 2,600 paying members in 2022 to roughly 50,000 by mid-2025 to nearly 70,000 now, by its own public stats. Seventy thousand people is a rounding error against Google's billions. But they are not a random seventy thousand. They are developers, founders, researchers, and SEO nerds. The exact people a B2B SaaS or an agency wants to reach, voluntarily routing around the ad-funded web.

And a growing share of "search" is no longer happening on a search engine at all. It happens inside ChatGPT and Perplexity, where the user asks a question and the model answers with a synthesized response and a handful of links. I do this myself now for anything research-shaped, and I felt slightly traitorous typing that, given what I do for a living. (Side note: my honest read is that ChatGPT replaced about a third of my own daily Google use this past year, and I am not a representative sample.)

So calibrate. For most sites we look at inside SEOJuice, Google still drives something close to 90% of search visits. Do not torch a quarter chasing the 3% who moved to Kagi. The point is the direction, not today's split. The single search box you have optimized for since you started doing this is fragmenting into a dozen smaller boxes, and the fragments skew toward exactly the audiences that convert.

What still works, and what is quietly dead

Some of the old playbook is now dead weight. Some of it matters more than ever. Worth being specific about which is which.

Quietly dead: building your content calendar around informational keywords whose entire payoff was a position-one click-through. "What is a backlink," "how does canonicalization work," the definitional stuff. Google now answers those itself, in the box, and the Pew data says almost nobody clicks through to the source. You can write the best "what is" page on the internet and watch it earn impressions with no clicks. We have a few of those in our own blog. They are pretty. They convert nobody.

Still very much alive, arguably more valuable now:

• Brand search. When someone types your name, no AI Overview stands between them and you. Branded queries are the one channel AI cannot disintermediate, which is why brand building quietly became an SEO tactic.
• Being citable. If Google, ChatGPT, and Perplexity are going to answer with synthesized text, the win condition is being the source they synthesize from and name. That is a different craft from ranking, and it is the whole reason "generative engine optimization" stopped being a buzzword.
• Showing up where humans actually gather. Half the comments on the news that prompted this article were some version of "I only use Google to find Reddit threads." That is a real behavior, not a meme. Communities, forums, and the places people trust other people are now part of the discovery map.

The new job: be the source, not the link

Here is the reframe I have landed on, and I will admit upfront it is a work in progress. The job is no longer to win a ranking. It is to be the thing that gets quoted, across whichever engine your specific reader happens to open.

That changes what "good content" means in concrete ways. It needs to state claims cleanly enough that a model can lift them. It needs real data and named sources, because that is what gets cited over generic prose. It needs to exist in more than one place, because your reader might be on Google's AI Mode, or in Perplexity, or paying for Kagi. None of that is exotic. It is mostly the same fundamentals, pointed at a new target.

I am not going to pretend we have this solved. Remember that 38% organic-click drop from the field study? We are watching it show up unevenly across our own pages, and we are still learning which formats survive and which do not. If you want the tactical version of this, the per-engine mechanics for getting picked up by ChatGPT, Perplexity, and Google AI Mode, we wrote that up separately. This piece is the why. That one is the how.

How we are starting to measure it

You cannot manage what you do not measure, and most analytics setups were built for a world where traffic meant a human clicking a blue link. So we started tracking two things that used to read as noise: how often AI crawlers hit our pages, and how often a visit arrives referred by an AI tool rather than a search engine. Both are now first-class metrics in how we judge a page, sitting right next to its rankings.

Concretely, the crawler side means watching your server logs and analytics for the bots that feed the answer engines: GPTBot and OAI-SearchBot from OpenAI, PerplexityBot, ClaudeBot, and Google-Extended, which is the token that decides whether Google may use your pages for Gemini and AI Mode at all. The referral side means filtering your traffic sources for visits arriving from chatgpt.com, perplexity.ai, and friends, which most analytics setups still bucket as "direct" or miss entirely. Inside SEOJuice we fold both signals into per-page tracking, so a page's AI-crawler hits and AI referrals live alongside its keyword positions.

What we see so far, across the sites we monitor: AI-crawler activity has gone from a curiosity to something a steadily growing minority of pages get on a regular basis. I am deliberately not going to hand you a clean percentage, because the honest answer is the referral signal is still small and noisy on our end, and inventing a tidy stat would be exactly the move this blog exists to call out. But the direction is up, and it is consistent across the sites, not a one-off. If you have never looked at whether AI engines crawl or cite your site at all, that is the cheapest place to start.

Frequently Asked Questions

Is SEO dead now that Google uses AI?

No, but the definition narrowed. Ranking for informational keywords to earn click-throughs is dying for a lot of query types. Being the source that engines cite, owning your brand search, and earning trust in communities are all very much alive. The skill set shifts more than it disappears.

Does Google AI Mode kill my organic traffic?

It reduces it on the queries it triggers, by around 38% on average in the one randomized study we have, with the steepest losses on simple informational searches. Commercial and comparison queries where users still want to evaluate options are far less affected. Check your own Search Console by query type rather than assuming a single number.

Should I block AI crawlers from my site?

For most sites, no. Blocking the crawlers that feed ChatGPT, Perplexity, and Google AI Mode removes you from the answers your future customers will read. The exception is content you genuinely need to keep out of training and synthesis. It is a real trade-off, not a default.

How do I get my page picked up by Google AI Mode and ChatGPT?

State your key claims clearly and early, back them with named data and sources, keep your facts consistent across pages, and make sure the AI crawlers can actually reach you. The per-engine specifics differ, and we cover the mechanics for each engine in a separate guide.

Related reading:

• Optimizing for Perplexity, ChatGPT Search, and Google AI Mode
• From SEO to GEO: Search Optimization Didn't Die, It Got Smarter
• Google AI Overviews Cause a Massive Drop in Search Clicks
• Optimizing for Zero-Click Searches
• Ask Engine Optimization: The Next Big Thing

Run a free AI visibility check to see whether ChatGPT, Perplexity, and Google's AI Mode actually cite your site today, before you decide what to fix.