Here’s the one thing I learned from looking at almost a year of search data across 11 of my sites: if Google can answer the question itself, people don’t click. If they need the actual thing, they still do.
Someone who searches for a famous person’s name can usually get what they want right on Google’s results page, and they have little reason to visit my page. Someone who searches for the full text of a poem has to click, because Google won’t show them the whole poem. Everything else in this post is the data behind that idea, and what it means if you build websites.
The short version
- Google showed my pages about 2.7 million times in October. People clicked about 1 time in 100.
- Searches for a quick fact, like a name or a definition, almost never got clicked.
- Longer, more specific searches, and searches for something you have to read in full, got clicked five to eight times as often.
- Visits from AI assistants are small, about 3% as many as from search, but on one of my affiliate sites they’re the visitors who buy.
Most people never leave Google
Two terms first, because the whole post rests on them. An impression is one time Google shows my page in the search results. A click is when someone actually visits. The click-through rate is the share of impressions that turn into a visit.
In October 2026, Google showed pages from these 11 sites roughly 2.7 million times. They got roughly 29,800 clicks. That’s a click-through rate of about 1.1%, so about 99 times out of 100, someone saw my page and didn’t visit. These are what people call zero click searches: the search ends on Google.
That’s not because the pages rank badly. My biggest site has averaged around position 5 or 6 since July, so it’s usually on the first page, and it still gets only about 1.4% to 1.5% of its impressions as clicks. People are seeing it. They’re just getting what they need without clicking.
Google answers more of the question itself
One sign of why is Google’s AI answers, the AI Overviews and AI Mode summaries that sit above the regular results. Since June 2026, Search Console shows how often your pages appear inside those AI features. On my four biggest sites, about one impression in five from June to October happened inside an AI answer. On the history site it was 23%, and on the health site 5%. (The exact shares may run a little high, because Google’s AI report also counts Discover.)
When the AI summary already answers the question, there’s little reason to scroll down and click. So the real question isn’t whether search still sends traffic. It’s which searches still do.
When people still click
I looked at the actual searches that brought people to these sites over 28 days. To keep the sites private, I’ve described each search by type instead of quoting it. Here are some typical examples:
| What someone searched for | Does the answer fit in a sentence? | Click-through rate |
|---|---|---|
| A famous person’s name | Yes | 0.32% |
| The “full name of” a famous person | Yes | 0.18% |
| The meaning of a famous line from a play | Yes | 0.27% |
| A figure from mythology (ranked about #1.5; 2 clicks from 3,173 impressions) | Yes | 0.06% |
| The title of a classic essay | No, you have to read it | 3.94% |
| A poem’s title plus its author | No, you have to read it | 4.72% |
The pages ranked in the top 10 for all of these. The difference is what the person wanted. A fact fits in a sentence, so Google gives it to them. A poem or an essay doesn’t, so they come to the site.
The same thing shows up in the bigger numbers. Searches of one or two words, which are mostly a name or a topic, were clicked 0.39% of the time. Searches of five words or more were clicked 2.85% of the time, and searches looking for a text or a quote 3.26%, at about the same rankings. Fact questions did better than one-word searches as a group, at about 2%, but as the table shows, plenty of them still barely got a click.
Which pages still get visited
Sorting the pages by type tells the same story:
| Type of page | What it gives people | Click-through rate |
|---|---|---|
| Directory listings | A specific provider to contact | 2.43% |
| Full texts and quotes | The thing itself, to read | 1.56% |
| Reference and history | Facts Google can summarize | 0.78% |
| Health and how-to | Answers Google can summarize | 0.76% |
Directory listings did best even though they rank lower, around position 13 on average. If you’re looking for a specific provider near you, you have to click. Full texts and quotes got about twice the click-through rate of reference pages.
It isn’t this tidy everywhere. On my biggest site, full texts got about three times the click-through rate of reference pages (2.08% against 0.71%). On my literature site, there was no real difference. The directory result also comes from a single site. But the overall direction is clear: the more a page offers something Google can’t show in a snippet, the more often people click.
AI assistants send a trickle, not a flood
Some people expect AI assistants to make up for lost search traffic. On these sites, they don’t, at least not yet. Over 28 days, AI assistants sent 787 visits, against 27,030 from search engines. That’s about 3%. Those visitors also stayed about half as long: around 33 seconds of engagement per visit, against around 68 seconds for search visitors. Most likely they came to check one thing the assistant pointed to, then left.
There’s one exception worth watching. One of my newer sites is an affiliate site, and it gets hardly any traffic from Google. It does, however, earn commissions regularly from visitors sent by AI assistants. It looks like people who arrive that way have already done their research in the chat and are ready to buy. It’s early, and I’m still collecting data on this, so I’ll share a proper analysis in the coming weeks.
What this means if you build websites
My own experience lines up with the numbers. It’s much harder now to start ranking a new site. You need solid content and you need links, and there’s no shortcut around either. But I’ve also found that if you build something people actually need or find useful, they’ll link to you on their own.
Put that together with the data, and the advice is simple:
- Don’t build a site around quick answers. Searches like “how many ounces in a cup”, “what time zone is Denver in”, or “how old is” a famous person get answered right on the results page. If the answer fits in a sentence, Google will show it and you’ll get the impression without the visit.
- Build around the thing itself. The full text of a poem. A directory listing with a provider’s hours and contact details. A calculator that does the conversion for you. A side-by-side spec table for the products someone is choosing between. People have to visit to read, contact, use, or compare these, so they’re the pages that get clicked, and they’re also the pages people link to.
- Go after specific searches. “Running shoes” is a topic, and Google already fills the page for it. “Best trail running shoes for flat feet” is a question someone still needs a page for. In my data, searches of five words or more were clicked about seven times as often as searches of one or two words.
- Watch your click-through rate, not just your clicks. From September to October, my literature site, which is about a year old, saw its clicks go up about 76%, from 2,160 to 3,800. Over the same month, its click-through rate fell from 1.2% to 0.9%, because a lot of the new impressions didn’t turn into visits. Clicks can grow and hide the problem. Click-through rate tells you whether Google still needs your page.
How I got these numbers
The data comes from Google Search Console and Google Analytics for 11 of my sites: mostly reference and history sites, plus health, how-to, product, directory, and tool sites. I left out sites with too little data, which ruled out most of my newest ones, and I’m not naming any of them.
A few limits worth knowing. This isn’t a before-and-after: AI Overviews rolled out across the US in May 2024, and Search Console only keeps 16 months of data, so I can’t compare against the time before them. The sites are young and still growing, which is why I focused on click-through rate rather than raw clicks. The search and page breakdowns cover the top searches and pages Search Console reports for October 10 to November 6, 2026; it leaves out rare searches. The AI share covers only my four biggest sites, and the biggest one carries a lot of weight in it. I also left product, review, and tool pages out of the page comparison, because they don’t rank high enough yet for a fair read.
Sources
- My own Google Search Console and Google Analytics data for 11 sites, pulled November 9, 2026. Monthly totals are rounded as Search Console displays them.
- Google Search Central Blog: Introducing Search Generative AI performance reports in Search Console (June 3, 2026)
- Google Analytics Help: Search Console keeps data for the last 16 months
- Android Authority: Google announces AI Overviews for Search, rolling out this week in the US (May 14, 2024)
Top photo: Alejandro Escamilla via Wikimedia Commons (CC0).