Tai Chi for AEO: a practical guide to AI search with a different perspective

AI search tends to reward consensus, and originality is how consensus begins. If you want ChatGPT, Gemini, Perplexity and Claude to describe and recommend your brand, you need four things: a clear identity, information you keep current, credibility that other people give you, and an original idea worth repeating. Those are four principles I've learned in more than twenty years of practicing tai chi.

What is answer engine optimization, and why does it matter now?

Answer engine optimization, or AEO, is the practice of influencing how AI systems discover, describe, and recommend your brand. It matters because the way customers find you has shifted under your feet. People used to search Google, scan a page of results, and click through to a website. Now they are asking a chatbot to tell them what they need to know, and the bot answers.

The scale is hard to ignore. Pew Research Center found that 42% of American adults use an AI chatbot to search for information. ChatGPT alone reportedly passed 1.2 billion weekly users in October 2026, and Google's AI Mode crossed a billion monthly users in May.

Meanwhile, the click is shrinking. According to Pew, when Google displays an AI summary, clicks on traditional search results drop from 15% to 8% of visits. The search marketing platform Ahrefs reported that the click-through rate for the top organic result was 58% lower on keywords that triggered an AI Overview, comparing December 2023 with December 2025.

For marketers, that means your customer is forming an opinion of you in a place you cannot see. In a tai chi fight, when your opponent loses balance, you guide them where you want them to go as they fall. Claude and ChatGPT have customers off balance in exactly that way, guiding them by skipping websites altogether and serving up a sentence. The good news is that you can use tai chi right back.

Why does AI search reward consensus?

A human buyer can be dazzled. They can be worn down by frequency, charmed by a Super Bowl spot, or talked into something by a good salesperson. A language model doesn't respond to those things in the same way. It has no ego to flatter and no attention you can demand. When it searches for information, it finds what exists about you, draws on patterns it's learned, and assembles an answer from what it deems relevant and credible.

LLMs predict plausible answers based on patterns in huge amounts of data. They are programmed to be helpful and accurate, which often means they favor established information over an unfamiliar claim, and they rely on signals of relevance and credibility. None of this means the most repeated claim is true. It means repetition and agreement can have an enormous influence on what becomes accepted as truth. The more sources say the same thing, the more likely that thing becomes part of the machine's understanding of the world.

That creates a problem for marketers. AI doesn't necessarily reward originality, but consensus is what everybody is already saying. The way through is that consensus has to come from somewhere. Somebody had to be the first person to say something, discover the number, notice the pattern, or ask the question nobody else thought to ask. Originality is where consensus begins, and four principles from tai chi describe how to get there.

The four principles of AI search at a glance.

Be straight

Change is constant

Relate

Be empty

Be straight: the machine has to know who you are.

Master Zhong taught Oliver (and Oliver taught me) three principles that shape how I think about tai chi and business. One of them is be straight. Know your center. Stand in it. Own it. Everything else in tai chi is impossible if you are leaning.

I have written many times that in the mind of a customer, you get to be one thing. A company that tries to be four things is usually remembered for none of them. What has changed is that the customer's mind is no longer the only place you get to be one thing. Your brand is now something a machine paraphrases, and what it says depends in part on how consistently your identity is represented across the information it encounters.

AirOps analyzed 21,000 brands and found that 68% of brand mentions were unique to a single AI model. ChatGPT, Gemini, Perplexity, and Claude can each build a different picture of your brand. You are trying to be described consistently by several different readers who never compare notes. That doesn't mean repeating the same sentence a thousand times. It means you need a clear idea worth repeating: one idea, a small vocabulary you use on purpose, propositions with proof instead of slogans, and the same essential answer to "what do you do" on your website, in your founder's LinkedIn posts, in a podcast interview, and in a review someone left on G2.

Example: how Traction built a vocabulary AI repeats.

Traction has invested in building a vocabulary word: liquid workforce. Since our pivot in 2019, we've used this phrase consistently to describe our business model. When I ask Gemini for an advertising or marketing agency that operates with a liquid workforce, Traction is at the top of the list.

On ChatGPT, the same query pointed straight at us.

Straight does not mean rigid. In tai chi, a straight spine is what lets you move freely. A clear narrative is what lets you publish on anything, in any format, on any platform, and still be recognizable when the machine stitches it back together.

Change is constant: nothing stays where you left it.

Another of Master Zhong's principles is that change is constant. Nothing holds still. Yin becomes yang becomes yin. The practitioner who treats a position as permanent is the one who gets knocked on their ass.

This is especially relevant to AI search because the information environment is changing all the time. New articles are published. Competitors introduce new products. Customers write reviews. Models are updated. Search systems change which sources they retrieve and cite.

AirOps found that 70% of the pages ChatGPT cited in its study had been updated within the past year, and that content updated within the previous three months was three times more likely to be cited. Pages that went more than three months without an update were also more likely to lose visibility. That doesn't establish that changing a date makes content more credible. It does suggest that keeping important information current matters.

Most companies still treat content as a campaign. Write it, launch it, report on it, move on. The pillar page from 2023 sits there doing its job, except the world it describes may have moved on. The tai chi way to think about this is not "publish more." It is that your content is a form, not a monument. You run the form again. You adjust it. Old posts get revisited with new evidence. Numbers get updated. A paragraph gets rewritten because you learned something.

Chime, one of the companies in an AirOps case study, had more than 700 posts and could refresh about 50 a quarter by hand. When the company automated its refresh cycle, AI citations on its priority questions reportedly tripled in four weeks. That's a vendor case study, so take the precise figure with a grain of salt, but the direction is clear.

Relate: the art of using no force.

Of all the principles I have learned in tai chi, relate may be the most difficult to explain. From the outside, push hands can look like two people pushing each other around. But that isn't what it feels like when you're sparring with someone who knows what they're doing. You don't feel a powerful shove. You feel yourself losing your balance, falling into a void, and you can't quite figure out how you got there. The person who sent you flying may have used almost no force at all. You don't create the force. You use the force that's already there.

The same AirOps study found that 85% of brand mentions in AI answers came from third-party content rather than the brands' own websites. Brands were 6.5 times more likely to be cited through someone else's domain than through their own. You can spend millions of dollars telling the world what your brand stands for, but when a customer asks an AI system about your category, much of what it says may come from comparison sites, reviews, analysts, Reddit threads, YouTube videos, and people who have used your product.

Many brands respond by pushing harder on their own domain. More blog posts, more landing pages, more "content." That kind of force-feeding ignores where much of the energy is coming from. The relate move is to understand the conversation already happening around you and become part of it. If people are asking questions about your category on Reddit, contribute something useful. If analysts are comparing solutions, give them evidence that makes your difference clear. If customers are telling stories about their experiences, make those stories worth sharing.

Relating goes further than earning third-party mentions. You can introduce an idea that other people carry for you: a fact, a number, a story, a way of describing a problem that suddenly makes sense. When people start repeating it, your original idea begins to become the consensus, and the machines begin to encounter it everywhere. That is much closer to what happens in push hands than buying a thousand backlinks ever will be.

Be empty: consensus begins with originality.

Language models learn from what has already been said. Ask one a generic question and you get something close to the consensus. The consensus is free, it is everywhere, and it doesn't need your website to exist. If your content says what everyone else's content says, you aren't only competing with the other brands in your category. You're competing with what the model can already produce on its own.

In tai chi, being empty is not about having nothing. It is about releasing unnecessary tension, attachment, and preconceived ideas. It's how you get straight. You don't decide in advance exactly how a movement must unfold. You make yourself available to what is actually happening so you can relate to it. In business, that can mean letting go of the assumptions everybody in your industry has accepted as truth. Why is everybody answering the same question? What are we missing because we're so busy trying to be better at what everybody else is doing?

The answer might be a proprietary statistic nobody else has, a customer story that reveals something nobody has noticed, or a new vocabulary for describing an old problem. Or it might be an entirely new question. Much of marketing is devoted to answering questions customers already have. We compete to be the best answer, the first answer, the most visible answer. But the brands that change categories often do something more powerful. They change the question.

Example: how Patiofyre changed the question.

Our client Patiofyre sells a wood pellet patio heater that has a four-foot infrared glass heating surface. They have competitors that use wood pellets. They have ones that use glass. And they have ones that use infrared. But they're the only ones that have all three. So they created content that clearly explains why an infrared glass surface is better for wood pellet heaters. ChatGPT now lists infrared glass as a feature to look for in a wood pellet heater and links to Patiofyre's website when it does.

That doesn't mean originality automatically earns citations. AI systems can ignore brilliant ideas, repeat bad information, and favor familiar sources over better ones. But an original idea that other people find useful has something generic content doesn't: a reason to spread. That's why your One Thing matters more now, not less.

How to build an AEO practice.

Nobody gets good at tai chi by doing it once. You do the form again and again. You learn something, forget it, rediscover it, and eventually begin to understand something your teacher told you ten years earlier. AEO works the same way. It often gets sold as a project: audit, fix, report, done. But the work doesn't stop when the audit is finished. Your competitors are still publishing, your customers are still talking, the engines are still changing, and you need to keep moving with them.

These habits support the work:

The opportunity is growing. The share of US ChatGPT answers that include a citation rose from 1.6% to 6.8% between June 2025 and May 2026, according to Similarweb. The hard part is that this has to happen every week, inside the normal work, by people who are also doing everything else. That is what makes it a practice.

Whoever tries less, wins.

Tai chi is deceptive. Watch somebody who is really good at push hands and it can seem almost effortless. Their opponent lunges, loses balance, and goes flying. The master barely moves. What you don't see are the years spent learning to be grounded, the sensitivity to another person's movement, the ability to change without resistance, and the willingness to let go of the instinct to push back. The absence of force is the result of the presence of a shit ton of practice.

Marketers have been conditioned to believe that the way to win is to push harder. More media. More impressions. More content. More optimization. AI search changes the equation. You can still buy attention, and you can still influence what people find. But you cannot simply purchase consensus. You have to give people something they believe is worth repeating.

The machine may reward consensus. But somebody has to give the world something to agree on.