Twenty five years ago, SEO meant tricking the Google bot. We stuffed pages with keywords and hid lists of search terms in text the same colour as the background. A travel page might read:

“Looking for cheap flights Sydney? Our cheap flights Sydney deals are the best cheap flights Sydney has to offer. Book cheap flights Sydney today.”

The bot loved it. Every human who landed there left. Then Google got smarter, the tricks stopped working, and good SEO became something every business simply had to do. A minimum standard, not a point of difference.

GEO is on the same curve, just moving faster.

GEO works. That’s the problem.

Let me be clear, GEO is worth doing. When we audited the Timezone and Kingpin brands for TEEG, a structured programme of schema and content improvements lifted AI brand citation by 37% in the first two months.

But anything that works that quickly gets copied just as quickly. Some of today’s tactics already have a familiar smell. Katja Forbes, who writes thoughtfully about the machine customer, recently put it this way: “The keyword-stuffing era is over. The adjective-stuffing era is here.” Same game, new name, and the same short half-life.

If you’re doing GEO, good on you. If you need help, no problem. But the conversation is moving on, because being found isn’t the same as being chosen.

The machine doesn’t just find you. It chooses for your customer.

In 2000, the bot found your page and a human did the deciding. Today, AI agents read, compare and recommend on your customer’s behalf, often before that customer ever reaches your site.

So great customer experience now has two audiences. Humans want story, feeling and a sense of who you are. Machines want structure, consistency and proof. Designing for both is harder than designing for either, and brands lose the recommendation in three ways.

1. They get averaged

Machines read the way a busy customer does: fast, once, looking for the point. If your messaging is vague, inconsistent or contradicts itself from page to page, AI falls back on the generic language of your category.

In the same piece, Katja showed Gemini flattening distinctive running shoe brands into “soft” and “reliable” until she pushed it to reflect each brand’s real voice. Most customers won’t push.

The fix isn’t writing for the bot. It’s clarity, which serves both audiences. When we rebuilt Queensland Public Trustee’s website, we not only improved user journeys for both humans (emotional) and machines (rational) we also consolidated more than 500 pages to just over 70 using content intelligence techniques that play to both audiences. Engagement rose 12%, page views 28% and search performance up 11%. Clearer for people, more legible to machines – that’s intelligent content.

2. They’re hard work

Machines favour technology that’s easy to talk to and content that’s easy to reach. Deep structures don’t get read. Duplicate pages get rejected. Agents run on their own instructions and skills designed to make them efficient, and if your platform makes their job difficult, they’ll move on to a brand whose platform doesn’t.

When a major HR compliance provider consolidated several product domains into a single front door, we were able to increase LLM referrals by more than 60%.

3. Nobody vouches for them

This is the one most businesses miss, and the one with the longest half-life.

It matters what you say and how you say it. It matters more what other people say about you. AI assistants weigh reviews, media coverage, expert commentary and third-party validation when deciding who to recommend. You can be perfectly visible and beautifully clear, and still lose to a competitor who simply looks more credible.

Your website tells AI who you say you are. Your reputation tells it whether to believe you.

The commercial stakes are real. When we benchmarked a major supplement manufacturer, its machine visibility was close to zero. We sized the growth opportunity at almost $130 million. Becoming visible would be a start. Giving people and machines reasons to trust it would be the next step.

That’s where earned reputation comes in. For Randstad, our sister agency History Will Be Kind delivered an earned media programme that generated 166 media hits and more than 60 direct client leads. It helped establish the brand as a leading industry voice, with coverage beyond its own channels to back that position.

Those results don’t measure AI recommendations. They demonstrate the value of having others tell your story, creating independent evidence for people and machines to draw on.

If you're not visible to machines or don’t have an earned media strategy, or aren’t sure what either is, now is the time to ask. Getting a machine to read your website is a web problem. Getting it to believe you is a business problem.

Preference is earned, not optimised

Visibility, clarity, technology and trust. Fix one in isolation and the gap simply moves somewhere else. Before you engage the next GEO 'expert' or spend money on building yourself a dashboard, ask yourself these four questionsbuy the next GEO dashboard, ask four questions:

1.      Can the machines see us?

2.      Do they understand what makes us different?

3.      Is our technology easy for them to work with?

4.      Does anyone credible vouch for us?

At Deepend | History Will Be Kind , these four questions sit behind Be Chosen, a suite of complex tooling run by 'experts in the loop' who'll take you through a one week, fixed price diagnostic program: Be Visible, Be Clear, Be Ready and Be Trusted. Run one or all four, and you’ll come away with a plain English roadmap your own team or agency can act on.

If you’d like to know whether you’re being found or being chosen, let’s start with a 20 minute conversation.