AI Search Is Eating the Click, and Agents Now Build the Shortlist

AI search is cutting the clicks that used to reach your website, and AI agents are starting to assemble buyers’ shortlists before anyone visits a vendor site. Traffic is now the wrong scoreboard. The new one is simple to say and hard to win: when a buyer asks a machine about your category, does it name you?

This piece is the why. For the how, see chapter 6 of our guide: from traffic to citations. To measure it, see chapter 7: measure what grows revenue when clicks disappear.

Do AI Overviews reduce website traffic?

Yes. The best independent data shows a large drop in clicks when an AI summary appears.

Pew Research Center tracked the real browsing of 900 U.S. adults in March 2025, across 68,879 Google searches. When an AI summary appeared, people clicked a regular result on 8% of visits, against 15% when it didn’t. They clicked a link inside the summary on just 1% of visits. And 26% ended their browsing session right there, against 16% without a summary (Pew Research Center).

Ahrefs compared 300,000 keywords using December 2025 data. When an AI Overview appeared, the top-ranking page’s average clickthrough rate was 58% lower (Ahrefs).

And the AI surface is growing. Google says AI Mode passed one billion monthly users within a year of launch, with queries more than doubling every quarter (Google).

Is SEO dead?

No. But ranking is no longer the prize. Being cited is.

Seer Interactive’s 2026 analysis found that, throughout 2025, brands cited in an AI Overview earned 2 to 5 times the organic clickthrough rate of brands that weren’t cited (Seer Interactive). The answer box didn’t kill the value of search. It moved the value to whoever the answer names.

Google’s own advice points the same way. It tells site owners to focus on “unique, non-commodity content,” because people in AI search ask “longer and more specific questions” (Google Search Central). A first-hand view beats a summary of what everyone else already said (Google). That’s bad news for “7 tips” content. AI can write that for anyone.

Who does AI name, and why isn’t it your blog?

Mostly other people’s coverage of you.

Muck Rack analyzed more than 25 million links cited by ChatGPT, Claude and Gemini across 17 industries. Earned media made up 84% of citations. Paid and advertorial content made up 0.3%, and journalism alone 27%. Recency matters too: more than half of the journalism citations came from articles published in the past 12 months (Muck Rack).

So the budget question changes. If AI mostly repeats what third parties say about you, the work that moves AI answers looks like PR, analyst relations, speaking, partner content and customer stories others will publish. Your own site still matters as the place those sources check facts. It just isn’t enough by itself.

How are B2B buyers using AI to buy?

To gather, and then they check with a person.

In a Gartner survey of 645 B2B buyers, conducted August to September 2025, 45% said they used generative AI in a recent purchase, mostly to gather information on vendors and products. Buyers used an average of seven information sources. And 69% said they prefer to validate AI-generated insights with a sales rep (Gartner).

The same buyers don’t fully trust either side. Fifty-one percent said they’re more likely to get misleading information from generative AI, and 49% said the same of a sales rep. And 67% prefer a buying experience with no rep at all (Gartner).

Put together, that’s the new journey: a machine builds the first list, and a person confirms it at the moment of risk. Your content has two jobs. Give the machine facts it can quote. Give the rep proof the buyer will believe.

What happens when agents build the shortlist?

The agent becomes the first reader of everything you publish.

Google announced “information agents” in Search that run “in the background, 24/7,” watching the web for changes that match a person’s request and sending an update. It also announced agentic booking for local services, and for some categories Search can call businesses for you (Google). Those launches are consumer-first. But B2B buyers already work the way these agents assume, so plan for them now.

An agent can’t be charmed. It reads what’s specific, current and consistent. That means what you do, for whom, with what proof, at what scope. Vague category language and “contact us to learn more” give it nothing to work with. Contradictory facts about you across the web give it a reason to pick someone else.

What should a CMO do about it?

Change what you count, then change what you make.

1. Count being named, not visits. Pick five questions your buyers ask AI and check monthly whether ChatGPT, Gemini, Perplexity and Google AI Mode name you, and which sources they cite. Chapter 7 covers how to measure AI visibility.
2. Fix your facts everywhere. Your name, what you do and your proof should read the same on your site, LinkedIn, partner sites and directories. Inconsistency is how AI confuses you with someone else.
3. Fund the sources AI reads. Earned media, speaking and partner content belong in the same budget line as content now.
4. Publish what only you can. First-hand results, named examples and industry-specific answers. That’s why industry-specific content matters more now, not less.
5. Arm the rep for the validation moment. If 69% of buyers want a person to check the machine’s answer, the rep needs the same proof the machine quotes.

What does this look like on a real site?

I’ll use ours, because I can show you the numbers.

That’s the honest baseline for a 35-year marketer’s site. It’s why we rebuilt around buyer questions and earned media instead of more posts.

I’ve watched the buyer’s path move before. At RAPP (1991–2005), I led the consumer programs that generated over $1.5 billion in annual revenue for Dell. In 2012, mix modeling at a single Harrah’s property, done in partnership with Polaris Research, cut more than $30 million a year from marketing spend without losing performance (the Harrah’s story). Both times the lesson was the same: when the buyer’s path moves, the budget has to follow the path, not the habit.

Where do you start?

Find out where you stand before you spend. Ask about the measurement and AI-visibility audit: a fixed-fee look at which spend grows revenue and where AI answers name or miss your brand, with written decision rules. If you run a service business, Berry Buckley, which I co-founded, builds AI workflow agents for owner-led service businesses in Dallas–Fort Worth (berrybuckley.com).

The click was never the goal. It was just the easiest thing to count.