Needle Space Labs
← Field notes

Field note

Brand is the algorithm now.

By Andy Jordan

For twenty years, search ran on a quiet deal. If Google put you on page one, it was vouching for you. Ranking was the endorsement. So a lot of companies treated brand as the soft line item and search as the hard one: rank, get the click, report the ROI.

AI broke the deal. When a buyer asks an assistant what to buy or who to trust, there are no ten blue links. There's one answer, and the machine decides who's in it.

Brand didn't get less important. It became the gate. Because in an AI answer, being seen is not the same as being recommended.

The rules are changing fast

The speed is the story. Google says AI Mode passed one billion monthly users (opens in new tab) just one year after it launched, with queries more than doubling every quarter. In the same announcement, Google said you can now hand tasks to agents just by asking a question in Search.

That last part matters more than the user count. We're no longer only marketing to people who search. We're marketing to agents that search, compare, and shortlist on their behalf. Forrester's newest buyer study (opens in new tab), a survey of nearly 18,000 business buyers, found 94% used AI during their purchase. The first reader of your story may not be a person at all.

And the answers don't sit still. SparkToro and Gumshoe (opens in new tab) had 600 volunteers run the same prompts nearly 3,000 times across ChatGPT, Claude, and Google's AI. Fewer than 1 in 100 runs returned the same list of brands. Getting the same list in the same order took closer to 1 in 1,000.

So checking a prompt once tells you almost nothing. Ten runs isn't enough either. The teams doing this well run their most important prompts 50 or more times a month and track how often they appear, not where they rank. The scoreboard has moved from position to presence.

Being seen isn't being chosen

Wil Reynolds, the CEO of Seer Interactive, just published one of the sharpest pieces I've read on this shift: Re-imagining AI Visibility KPIs in a Query Fan Out world (opens in new tab). His most important finding is one every brand leader should sit with. In an AI answer, you can show up and still lose.

His example is a travel prompt where ChatGPT names Airbnb and VRBO, then tells the reader to look beyond them. Both brands were visible. Neither was recommended. Reynolds calls it the anti-recommendation.

That's the trap in most AI visibility reports. "Did we show up?" is a yes or no. The question that matters is what the machine said about you once you did.

Brand is the algorithm

Here's the part of Reynolds' piece that stopped me. Name a premium watch. Now a luxury car. You can, even if you've never owned either. Years of reputation put those names in your head before you ever went shopping.

A chatbot works the same way. The models absorbed the same reputations we did, and they reach for them first. Reynolds calls it hardwiring. When you ask an AI a question, it runs a chain of searches before it answers, which the industry calls query fan out. Watch those searches and you'll see the model already naming the brands, people, and publications it trusts. It decided who mattered before it looked anything up.

His proof is Banana Republic. Google "ethical jeans" and it shows up on page one. Ask an AI and it doesn't, in all of Reynolds' testing. Banana Republic had the page, the traffic, and the conversions. What it didn't have was a reputation for the thing it was claiming.

The newest data backs this up. AthenaHQ, a platform that tracks AI search, released its State of AI Search report (opens in new tab) last week, drawn from millions of answers across eight AI models. The average brand shows up in 16.3% of answers. Category leaders show up in 56.5%, about three and a half times as often. And AI answers cite a brand's own website only 16% of the time. The machine is describing you in other people's words.

That's why I keep saying brand is the algorithm now. You used to be able to rank your way into a category. Now you have to be known for it, by people and by machines.

What isn't changing

The tools change every quarter. What the machines trust hasn't changed at all. It's the same thing people have always trusted: someone credible, other than you, vouching for you.

Muck Rack analyzed more than 25 million links (opens in new tab) cited by ChatGPT, Claude, and Gemini this year. Earned media made up 84% of them: journalism, research, and independent coverage. Paid content barely registers. The models are reading what others say about you, not what you say about yourself.

There has never been a better time to work in content, PR, comms, or brand. The people who earn a company's reputation now shape what the machines say about it.

The same goes for people. AI engines lean on the experts and creators who already carry weight in a field, and so do buyers. Forrester found (opens in new tab) the typical B2B purchase now involves nine external influencers. Buyers check what AI tells them against peers, product experts, and analysts, and they're more likely to engage a vendor because of what an industry expert said than because of an AI answer. And some of the most credible voices you have are already on your payroll. Your engineers, scientists, and founders know things no campaign can fake. When they publish, speak, and show up in the conversations that matter, they become signals the models learn to trust.

One more thing that hasn't changed: nobody trusts marketing speak, and AI can spot it a mile away. Vague superlatives and "leading provider of innovative solutions" give a model nothing to cite. Specific claims with real proof behind them, in the words of people who know, do.

Brand is a system, not a campaign

If reputation decides who gets recommended, brand can't be a logo refresh or a launch campaign that ends. It has to be run, every week, like any other system the company depends on.

Reynolds frames his new metrics as search KPIs. Read them as a brand leader and every one is a brand metric. How much of what you publish do people actually pass to each other? Are the trade publications the AI trusts writing about you, and is that growing? Are the experts the model already knows saying your name? When you run your most important prompts 50 times this month, do you show up more than last month? And when you launch into a new market, does the world's story about you back up the claim yet?

Notice what's missing: rankings, clicks, impressions. Every signal is about whether people believe you. That's not a search problem. It's a brand problem, and it needs an owner.

It also ends the playbook a lot of companies are still running. Ten times the content at half the cost doesn't impress an editor or an expert, and they're the ones feeding the model. Cheap volume might move a dashboard. It won't move the machine.

How I run it

This is why Needle Space Labs maps every piece of a brand to the demand engine it feeds. Be known, be chosen, be bought. Reynolds' anti-recommendation is exactly the gap between the first two: a company the market knows but doesn't choose.

Closing that gap starts the way a reporter would. The story comes from interviews with founders, engineers, and customers, not from a prompt, and every claim ships with real proof behind it. That's what earns the editors, experts, and customers whose words the machines learn from.

Then we measure what the machines actually say. One of the agentic workflows I build tracks how ChatGPT, Claude, Perplexity, and Google's AI answers describe a company against its competitors, and flags when the description drifts. People decide what to do about it. That's brand run as a system: a story worth repeating, proof that holds up, and a way to see whether it's landing.

The companies that win the AI answer won't be the ones that show up most. They'll be the ones the machine, and everyone else, already believes.

Sources

Get a free 48-hour story audit← Previous note