SerpApi.Org
First, understand how it works.
If you run marketing, take a breath. The shift to AI search is here, but the work you must do has shifted. Before you tear up your playbook, there are four things to understand about how AI search works: SEO isn’t dead, AI search runs on the past, everyone sees a different answer, and you can’t measure where you stand by asking AI yourself. Get that straight and the rest gets a lot less scary.
For 20 years, search meant Google. You typed in a question or a brand name and got a page of results. A few paid ads sat up top, where companies bid on keywords to grab more share of voice, and then came the blue links. We spent two decades learning how to rank first on that page.
Over the last few years, that behavior has moved. People now ask ChatGPT, Claude, Perplexity, Gemini, Grok, and others, the same things they used to type into Google, except now they get about three to four answers instead of 10 links. This isn’t a fringe habit. Bain & Company found that about 80 percent of consumers lean on AI-written summaries for at least 40 percent of their searches, and roughly 60 percent of searches now end without anyone clicking through to a website. The answer is the destination now.
That’s a real change. Here’s what to understand before you react.
SEO isn’t dead
You’ve heard that SEO is dead. I don’t agree. The large language models (LLMs) behind AI search are trained on the open web, and they pull from it, and that web is the one SEO built. If your content is structured, credible, and easy to find, you’re feeding the same sources these models draw from. If it isn’t, you’re invisible to them too.
What changes is the emphasis, not the work. Ahrefs studied 75,000 brands and found that brand mentions across the web track with AI visibility far more closely than backlinks. Getting talked about in articles, guides, and other people’s sites now counts at least as much as your own pages. So, keep doing SEO. Just stop treating your own website as the only thing that matters.
AI search runs on the past
LLMs learn from existing data, so their knowledge is historical by design. That has a consequence most marketers miss. If you change your positioning today, launch a new message, or fix your site, don’t expect AI search to reflect it tomorrow. The models were trained before you made the change, and it takes time for new content to be crawled, absorbed, and surfaced.
Plan for the lag. This is a months-long build, not a switch you flip. The brands that appear in AI answers next year are the ones putting in the work right now.
Everyone sees a different answer
This is the part that trips people up. There’s no single fixed ranking in AI search like there was on Google for two reasons.
The first is personalization. There’s a name for the mechanism behind it: fingerprinting. It’s a tracking technique that stitches together small signals about you. These signals include your browser, device, location, time zone, and language, and it is compiled into a profile unique enough to pick you out of millions of people. Your fingerprint follows you even when you’re not logged in, and even in incognito or private mode.
AI tools tailor their answers off the same idea: what they can tell about who’s asking, plus whatever they remember from past conversations. So, if two people ask, “What’s the best project management tool for my team,” ChatGPT can hand them different brands based on what it already knows about each of them. The recommendation changes depending on who’s doing the asking.
The second is that these systems aren’t deterministic. Open an incognito window, run the same query twice, and you’ll often get two different answers.
Combine these two factors and LLM answers start to feel like snowflakes. The answer you see isn’t the one everyone sees and isn’t even the one you’ll see again.
You can’t measure where you stand by asking it yourself
The worst way to learn how your brand is performing in AI search is to ask ChatGPT yourself. You’ll get one answer, shaped for you, that probably won’t repeat if you run it again.
It gets worse if you ask about your own brand by name. These models are built to be agreeable, so the moment you suggest your company belongs in the results, it goes along with you. Ask if you’re a top three player and it will often just say yes. That’s not the truth; it’s a flattering, tainted answer.
The first step to appearing in AI search is measuring. Before you spend a dollar improving your AI visibility, find out where you stand, and do it at scale, across many runs, so the personalization and randomness average out. Then get specific. Figure out your ideal customer profile, identify the exact prompts that customer would type, and measure how you show up on those. A generic ranking barely tells you anything. The prompts your buyer uses are the only ones that move revenue.
For now, sit with the basics: SEO still matters, AI learns from the past, everyone sees something different, and you can’t manage what you haven’t measured. Get these four principles straight, and you’re already ahead of most of the market.
This post originally appeared at inc.com.
“Click here to subscribe to the Inc. newsletter: inc.com/newsletters"
