Automation & Workflows · Updated July 2026

How to Use AI for Search: A No-BS Guide for Marketers

68% of Google searches now end without a click. Here's how I use AI for search every day, the four moves that matter, and my free no-BS guide to steal.

Part of the AI Search & GEO guide

A friend reached out looking for a marketing internship for his son. I had nothing but bad news: every company and university I worked with had just cut their intern programs, and the reason was always AI.

I told that story for most of a year. Then I checked whether it was still true. It isn’t, at least not the way I was telling it.

Employers told NACE they planned to hire about 4% more interns in 2026 after the sharp drop the year before. And when the Strada Institute asked executives what AI did to entry-level roles, 41% said it stripped out the routine work while 42% said it expanded the analytical and judgment work. The rung didn’t disappear. It moved up, to “can you ask a good question and judge what comes back.”

That’s the skill this guide teaches, and search is where it pays off first.

Search changed while everyone argued about it

SparkToro ran clickstream data on US Google searches from January through April 2026. About 68% of them ended without a single click to any website, up from 60% two years earlier.

When it first crossed 60% I wrote a survival guide for zero-click search and figured that was the alarming number. It was the warm-up.

AI Overviews now appear on more than a fifth of searches. When one shows up, click-through drops by nearly 60%.

Google’s AI Mode was still a sliver of real queries in that window, about a third of a percent. Then Google said at I/O this year that AI Mode passed a billion monthly users, with query volume more than doubling every quarter.

I’ve been writing about chatbots eating Google search alive for a while now. From the searcher’s seat it’s simple. The answer comes to you, and the ten blue links are the fallback.

The smallest bet that pays: 30 minutes a week

I’m not going to tell you to overhaul your week. Block thirty minutes, once a week, and spend it on one thing: searching and asking questions with AI instead of around it.

You search dozens of times a day already. Email threads, docs, the open web, your own files. Most people still do it the way they did in 2015, then paste the answer into a chat tool as an afterthought.

Flip the order. Start with the question, let the model do the first pass of finding and framing, and an hour of digging compresses into a few minutes.

Thirty minutes isn’t a course and it isn’t a certification. It’s one block on your calendar that you don’t cancel.

How to use AI for search: the four moves

These are the moves I run most days. You don’t need a new tool for any of them, just a different first step.

  • Start with the question, not the keyword. Type the full sentence you’d say out loud. Your prompts are almost certainly too short, and that’s the cheapest fix on this list.
  • Let the model do the first pass. Finding and framing is its job. Judging what comes back is yours, and that judgment is the rung that moved up.
  • Talk instead of type. Dictate the messy version.
  • Point it at your own material too. Email threads, docs, and your notes count, not just the open web.

That third one is the sleeper. I dictate to Gmail, Slack, docs, and my notes at well over 100 words a minute now, and most of that runs straight into a model.

Talking out a messy first draft of a question gets me a far better answer than the tidy one-liner I’d have typed. I wrote the whole thing up as the T.A.L.K. method, which is four letters and a five-minute conversation.

Pick the mode before you pick the words

This section didn’t exist when I first published this post, and it’s now the part I’d read twice. Searching with AI is three different jobs, and picking the wrong one burns your thirty minutes.

  • Quick answer. One question, one reply, thirty seconds. That’s most of your searches.
  • Deep research. You hand over a real question and the model spends five to thirty minutes planning, reading dozens of sources, cross-referencing them, and writing back a cited report. Since February you can point it at your connected apps, restrict it to sources you trust, watch it work, and interrupt to redirect it mid-run.
  • Agent mode. The model drives a real browser and does the clicking for you.

Deep research is the one marketers underuse. It’s also the one that most needs a human on the other end, because a cited report is a starting point and not a source.

Click the citations. I’ve had reports hold up beautifully, and I’ve had one cite a page that said something slightly different than the summary claimed.

If you want the agent version, I mapped the three surfaces where Claude can use a computer and the one rule for picking between them.

Start with the guide, not a course

I wrote a short, no-fluff walkthrough of how I actually use AI for search. No theory, no hype, just the moves I run most days.

Get it free here: The No-BS Guide to Using AI for Search

Work through it in one of your thirty-minute blocks. It’s built to be applied the same day, not filed away.

Searching with AI is only half the skill

Getting fluent at finding things is one side of the coin. The other side, the one that matters more every quarter, is making sure your own work shows up when other people ask AI those same questions.

Here’s the number that reframes it. SE Ranking looked at traffic across 101,574 sites and found AI platforms send about 0.32% of all website visits. That’s sixteen times what it was two years ago, and it’s still a rounding error.

So the clicks aren’t the prize. The mention is.

When a prospect asks for a recommendation in your category, you’re either inside that answer or you aren’t. ChatGPT drives roughly three quarters of that referral traffic, with Gemini near 12% and Perplexity around 7%.

I’ve already been wrong about how much traffic this costs, in the optimistic direction. What I’d tell you now is to skip the shortcuts.

The llms.txt file half the internet added did nothing measurable for AI citations across nearly 300,000 domains. Start with my guide to generative engine optimization, and the broader AI marketing hub for how the pieces fit together.

You don’t need any of that to get value this week. Searching better comes first. The rest builds on it.

The future isn’t waiting

Block the thirty minutes. Open the guide. Run one of the four moves tomorrow.

The routine half of this work is what got automated. The judgment half is the rung that moved up, and thirty minutes a week is a cheap way to reach it.

Get the next move before everyone else does

I send two emails a week breaking down a single AI-for-search or AI-visibility habit you can use the same day, drawn from what’s actually working in my own marketing. No tool dumps, no hype cycle.

Subscribe free and start with the search guide above.

-Alec