Five years ago, when a service-business buyer started looking for a vendor, the first action was almost always a Google search. They'd type a query, scan the top three or four results, click into a couple, and start building a shortlist. The marketing playbook for that buyer was clear: rank for the right queries, write a strong service page, capture the click, and move them down the funnel.
That sequence is breaking. Not slowly. Quickly. And most service businesses haven't realized it yet.
The buyer who used to start on Google is now starting in ChatGPT, Perplexity, Gemini, or Claude. They ask a question (something like "What's the best video production agency for SaaS in the Midwest?" or "How should a $5M home services company think about marketing budget?") and they get an answer with three or four named companies inside it. By the time they hit Google at all, they're not researching the category. They're researching the shortlist they already have.
What's actually changing
The behavior shift isn't subtle. We see it in client analytics. Direct branded traffic is up. Organic traffic for top-of-funnel keywords is flat or down. Referral patterns now include "chat.openai.com" and "perplexity.ai" as recognizable sources, even when they don't show up cleanly in standard analytics. Sales conversations open with prospects who already know the company's positioning, sometimes more accurately than the company's own homepage describes it.
What's happening underneath: AI assistants are doing the synthesis layer that used to happen in the buyer's head. The buyer used to read four service pages and form an opinion. Now an LLM reads forty and forms the opinion for them. The shortlist is being built before any human ever sees a SERP.
If you aren't on that list, you don't get to compete. The buyer didn't reject you. They never considered you, because the AI never named you.
Why this is different from "just SEO"
The instinct is to treat this as a new SEO problem and ask which keywords to optimize for. That misses what's different. Traditional search ranks documents. AI search ranks claims about you. The system isn't picking the best page about home services marketing. It's picking the best summary of who does home services marketing well, synthesized from everything it has read about everyone in the category.
That changes what makes a piece of content valuable. The page that ranks #1 in Google might still rank #1, but it might never get cited as a source in an AI answer if it doesn't carry information the model considers worth quoting. And the page that gets cited might be a third-party blog, a podcast transcript, a forum post, a press mention. The locus of authority diffused.
The shortlist is being built before any human ever sees a SERP. If you aren't on that list, you don't get to compete.
The new question to optimize for
The question isn't "How do I rank for [keyword]?" It's "What does an AI assistant currently say about us when someone asks the question our buyer is actually asking?"
That's a testable question. Open ChatGPT or Perplexity, ask the questions a real prospect would ask in your category, and look at the answers. Three things almost always show up in the audit:
- You aren't named at all. The model has no clear signal that you exist in the category. This is the most common state and the one with the cleanest fix.
- You're named, but the description is generic or wrong. The model has fragments of information about you but no coherent story, so it generates a plausible-sounding but vague summary.
- You're named accurately, alongside competitors. The model has a clear picture of who you are and how you differ. This is what "winning" looks like at this layer.
Most service businesses are in the first state and don't know it.
What actually works
The work to fix this is less mysterious than it sounds, but it's different from traditional SEO. A few things that consistently move the needle:
Be quotable in your own copy
LLMs cite content that contains specific, declarative claims. "We help service businesses build marketing systems" is not citable. "We work with service businesses doing $2M to $25M in revenue, and we focus on operational systems before we touch ad spend" is. Specific = citable.
Get cited elsewhere
The model's view of you isn't built only from your site. Podcasts, guest posts, industry interviews, comparison articles, third-party reviews: those signals shape how the model summarizes you. If your founder is on three industry podcasts a year and quoted in two trade publications, the model has a stronger story to synthesize than if all the information about you lives behind your own homepage.
Make the entity layer clean
Schema markup, consistent NAP data, clear org structure on your site, an unambiguous "we are X, we do Y for Z" sentence on the homepage. These look like SEO basics. They are also what AI assistants use to disambiguate companies with similar names and verify they're describing a real entity.
What this doesn't change
It's worth being clear about what isn't different. Buyers still make decisions based on trust, fit, and proof. They still want to talk to a human at some point. They still close on the call, not on the page. Most of the deep work of marketing (positioning, offer, proof, sales conversation) is the same as it's been for twenty years.
What's different is the discovery layer. The funnel didn't shorten. The top of it just moved off your site, into a system you don't directly control, and you have to optimize for being represented well inside it.
Service businesses that get this early will benefit asymmetrically. The ones that don't will spend the next two years confused about why their pipeline is shrinking even though their Google rankings haven't moved. The metrics will tell a coherent story to anyone watching the right numbers, and a confusing story to anyone still watching only the old ones.
