AI Is Already Choosing Your Competitors. Here's Why.

For twenty-five years, the marketing question was simple: how do we rank on Google? Today, the question has changed — quietly, but completely. A growing share of your prospects no longer Google a problem. They ask ChatGPT, Perplexity, Claude, or Google's own AI for a recommendation. And the firms that get named are rarely the best ones. They're the most legible.

The new buying funnel

Here's what's actually happening. A buyer realises they need a tax attorney, a fractional CFO, an estate planner, an architect. Twenty years ago they'd ask a friend. Ten years ago they'd Google it and read three websites. Today, they open ChatGPT and type, "Who are the best estate planning attorneys in Austin for blended families?" In four seconds, they have three names, a one-paragraph rationale for each, and a pre-formed opinion. By the time they visit your website, the decision has already been narrowed to a shortlist of two — and you're either on it or you're not.

The harsh reality: AI tools don't pick the most qualified firm. They pick the most retrievable one. That's a different game entirely, and most established firms haven't realised they're playing it.

Why your competitors get named and you don't

Large language models are trained on, and increasingly grounded in, the open web. They don't browse your beautiful homepage. They don't watch your video. They don't appreciate the typography. What they consume is structured, citable, retrievable text — and a lot of it. The firms that show up in AI answers tend to share a few unglamorous traits:

• They have detailed, expert-written content on the specific questions buyers ask, in plain language a model can lift.
• They're cited consistently across third-party sites — directories, association pages, podcast notes, press, panel mentions, podcast guest credits.
• Their entity (the firm, the founder, the partners) is clearly disambiguated. The model knows who you are, what you do, and where.
• Their site uses clean schema and structured data so the model can confidently associate the right firm with the right specialty in the right city.

If your competitor is a slightly less impressive firm with a blog that answers buyer questions in detail and a website that's easy for machines to parse, they will get named in the AI answer. You won't. The model isn't being unfair — it's being mechanical.

The "almost-decision" has moved upstream

I talk a lot about the almost-decision: the moment a prospect has nearly chosen you and is just looking for permission to commit. For decades, that moment happened on your website. Now it's happening one step earlier, inside the AI tool. By the time someone clicks through to your site, they're often not evaluating you against the field anymore — they're confirming a choice the model has already pre-made.

That's good news, if you're the named firm. It's terrifying news if you're not, because you'll never see the deals you didn't get. There's no missed-call log. There's no analytics event for "ChatGPT recommended someone else." You just notice, eventually, that fewer of the right calls are coming in.

What to do this quarter

You don't need to overhaul your marketing. You need to do four specific things, in this order:

1. Audit your AI visibility. Ask the major models the questions your buyers ask. Note who gets named. Note whether you're mentioned, mis-described, or invisible. This takes an hour and changes the conversation.
2. Strengthen the content models actually consume. Long-form, plain-language answers to the real questions in your category. Not thought-leadership essays. Not SEO mush. Genuine expertise, written like you'd explain it to a smart client.
3. Build third-party signal. Get cited. Be a podcast guest. Get listed where your peers are listed. Get quoted. Models reward consensus, and consensus is built outside your own website.
4. Make your site machine-legible. Clean schema, clear entity, unambiguous service and location pages. Boring work. Compounds for years.

The firms that move on this in the next twelve months will spend the rest of the decade being recommended by tools they don't control to buyers they never spoke to. The ones that wait will keep wondering why the phone doesn't ring like it used to.

It's not a traffic problem. It's a recommendation problem. And it's fixable — but only if you start treating AI as the new front door, not a future trend.