ResilientNiche
← Blog12 min read

The New Funnel: From Search to AI Recommendation

The old funnel was Google, click, read, contact. The new one starts in ChatGPT and ends with a buyer who already trusts you. Here's how it works.

Photo of Malik Browne

Malik Browne

Built BakingSubs to 162,500 Copilot citations and accelerating. Now teaching the system behind it.

  • seo-shift
  • strategy
  • chatgpt

The funnel buyers actually walk through in 2026 looks nothing like the one most coaching sites are built for. Old funnel: someone Googles "life coach near me", clicks a few results, reads a homepage, fills out a form. New funnel: someone asks ChatGPT, gets three names back, picks one, verifies on Google, then contacts. Different shape, different stages, different things to optimize.

Key takeaways

  • The old funnel had four stages (search, click, read, contact). The new funnel has five (AI query, AI shortlist, single click, Google verify, contact), and a buyer can drop off at any one.
  • The hardest stage to win is the AI shortlist. Only 3 names get returned per query, and getting on that list is what the Citation Cluster Method is built for.
  • A buyer who arrives from an AI recommendation has already pre-qualified you. They convert at much higher rates than cold Google traffic, even on a slow site.
  • Your homepage is no longer the first impression. Your ChatGPT or Perplexity answer is. The homepage now does verification work, not selling work.
  • BakingSubs earned 144,321 Microsoft Copilot citations in a quarter by building for the new funnel, not the old one.

The old funnel is what your website is still built for

The old funnel had four stages, and almost every coaching site you've ever seen was designed for it. Someone typed a query into Google. They saw ten blue links. They clicked two or three. They scanned, decided, and contacted one.

Every part of an old-funnel site reflects that reality. The hero section had to grab attention because it was competing with nine other tabs. The bio had to sell hard because the visitor was comparing in real time. The CTA had to convert on first read because the buyer was still cold.

This funnel still works for a few queries. Local searches with a map pack. Brand searches where someone already knows your name. But the queries that used to feed coaching pipelines (the "how do I find a coach for X" queries) have been hollowed out by AI answers. Google now answers the question itself, or hands it to ChatGPT, Claude, Perplexity, or Microsoft Copilot to answer.

When buyers stop clicking the ten blue links, the whole old-funnel website stops doing its job. If you've watched your traffic drop and not understood why, the answer is usually that the funnel changed underneath you.

The new funnel has five stages, and most coaches lose at stage 2

Here is the new funnel, stage by stage.

  1. AI query. A buyer asks ChatGPT, Claude, Perplexity, or Copilot something like "Who's a good life coach for new managers in their first leadership role?" They do not type this into Google anymore. They type it into a chat window.
  2. AI shortlist. The engine returns 2 to 4 names with a sentence about each. This is the new front door. If your name is not in this list, the rest of the funnel does not exist for you.
  3. Single click. The buyer picks one name from the shortlist (usually the top one, sometimes the one that matches their specific situation best) and clicks through.
  4. Google verify. Before contacting, they Google your name. They want to see a real human with a real footprint. LinkedIn, podcast appearances, a few reviews, a clean website.
  5. Contact. Form, email, or call booking.

Most coaches lose at stage 2. They never get on the shortlist. Their site is invisible to the engines, so the buyer never even reaches a stage where the homepage matters.

The reason most coaches lose at stage 2 is that AI engines do not pick names the way Google picked rankings. Google rewarded backlinks, age, and on-page SEO. AI engines reward clear authorship, topical depth on a narrow subject, and easy-to-quote answers. A 12-page coaching website that says "I help everyone reach their potential" cannot get on a shortlist for "first-time engineering manager", because the engines have nothing specific to pull.

A buyer walking both funnels, side by side

Let me trace one buyer through both funnels so the shape is clear.

Meet James. He's a senior software engineer in Manchester who just got promoted to engineering manager. He's six weeks in. He's drowning. He decides he needs a coach.

Old funnel (2019 version of James):

James types "executive coach for engineering managers UK" into Google. He gets ten results. He clicks the first three. The first site has a generic homepage with stock photos. He bounces. The second is a coaching directory with 200 names. He bounces. The third is a real coach but her bio is vague and she works mostly with C-suite execs. He bounces. He gives up and tells himself he'll figure it out alone.

New funnel (2026 version of James):

James opens ChatGPT and types "I just got promoted from senior engineer to engineering manager. Six weeks in and struggling with the people side. Who are some coaches who specialize in this exact transition?"

ChatGPT returns three names. One is Priya, a coach in London who has published 18 posts specifically about the senior-IC-to-EM transition. Her name comes up first because ChatGPT has pulled from her writing repeatedly in the past. James clicks her site.

Her homepage does not try to sell him. It confirms what ChatGPT already told him. Yes, this is the right person. Yes, she works specifically with new engineering managers. Yes, she has a clear method. He Googles her name to verify. He sees her on two engineering podcasts and a Substack she writes. He books a call.

Notice what's different. In the new funnel, James has already decided before he reaches Priya's website. The site is not selling. It is verifying. And the engines did the selection work upstream, which means the entire game is about getting onto that shortlist in stage 2.

This is why the channels that still work for coaches in 2026 look so different from the 2019 list. The channels that fed the old funnel (Google ads, SEO, cold outreach) feed a funnel that buyers no longer walk.

Each stage of the new funnel needs different optimization

Because the funnel changed shape, the things you optimize for at each stage changed too. Here's the mapping.

StageOld funnel goalNew funnel goal
1. QueryRank in Google for the queryBe present in the engines' training and retrieval for the query
2. Shortlist(didn't exist)Get named in the AI's top 3
3. ClickBeat 9 other tabs with a strong heroConfirm what the AI already said
4. Verify(was step 3 above)Have a clean, consistent footprint across Google
5. ContactPush hard with CTAs and offersMake booking easy, get out of the way

Stage 2 (getting onto the shortlist) is the new center of gravity. It's also the stage almost nobody is building for, because most coaching marketing advice is still teaching old-funnel tactics. Long lead magnets. Big email sequences. Daily LinkedIn posts. These can help, but none of them get you onto a ChatGPT shortlist on their own.

What gets you onto the shortlist is a cluster of well-written, specific posts that answer the exact questions your buyers are typing into AI engines. That's what the Citation Cluster Method is built to do, and it's how BakingSubs earned 144,321 Microsoft Copilot citations in a single quarter without ads, backlinks, or social media.

What stage 3 (single click) means for your homepage

The biggest practical shift is what your homepage now needs to do.

In the old funnel, your homepage had to sell. The visitor was cold. They were comparing you to three other tabs. The hero had to grab them, the bio had to convince them, the CTA had to push them.

In the new funnel, your homepage has to verify. The visitor arrived warm. ChatGPT or Perplexity has already told them you're the right person. They are not comparing. They are confirming.

A verification-mode homepage looks different. It opens with a clear statement of who you work with and what transition you help them through. It shows recent writing or talks on that exact topic so the visitor sees the AI's recommendation reflected back. It has a simple "book a call" link, not a 14-field intake form.

The mistake most coaches make is keeping the old-funnel homepage when they're getting new-funnel traffic. The buyer arrives ready to book and gets hit with a hard sell aimed at a cold visitor. They bounce, confused. The AI did its job. The homepage undid it.

What stage 4 (Google verify) means for your wider footprint

The verify step is the new funnel's quiet killer. Buyers who arrive from an AI recommendation almost always Google your name before they contact you. They want to see that you are real.

If they Google you and find one website, one Twitter account from 2019, and nothing else, they hesitate. The AI said you were an expert. Google said you barely exist. They go back to the AI and pick the second name on the list.

A clean footprint does not mean you need a podcast and a book. It means a few signals consistent with what the AI said. A LinkedIn that matches your site bio. Two or three appearances on niche podcasts in your field. A handful of reviews or testimonials with real names. Maybe a Substack or newsletter archive. The volume does not matter. The consistency does.

This is also why bare-bones "coach business cards" lose to coaches who write. The writer has a footprint Google can show. The card has nothing. When the buyer verifies, the writer wins by default.

Why this is good news, not bad news

Most coaches read this and panic because their site is built for the old funnel. The panic is misplaced. The new funnel is actually friendlier to small, expert-led operators than the old one was.

In the old funnel, you competed with everyone who could outspend you on SEO and Google Ads. A solo coach lost to a marketing agency or a coaching directory with a team of writers and a budget. In the new funnel, you compete on how specific and clear your writing is on one narrow topic. Big sites with broad coverage cannot win narrow shortlists. A solo coach with 20 posts about one specific transition can.

This is exactly the dynamic that lets a niche site beat much bigger competitors in AI search. The engines reward depth on a topic, not breadth across topics. Small expert-led operators are built for depth.

The same pattern applies outside coaching too. A workplace mediator with a tight focus on tech-company conflict beats a general HR consultancy. A nutritionist who only writes about perimenopause beats a general wellness site. An ecommerce founder who writes deep on one ingredient or one technique beats Amazon for the recommendation query, even though Amazon owns the transaction.

Frequently asked questions

How do I know if buyers are using the new funnel to find me?

Ask three of your last five clients how they first heard about you. If any of them say "I asked ChatGPT" or "I read about you somewhere and then Googled you", you have new-funnel buyers already. If your traffic from Google is down but your inbound calls are steady or growing, that's also a signal. The new funnel sends fewer clicks but warmer ones.

Do I need to abandon SEO entirely?

No. SEO still matters at stage 4 (Google verify), and brand searches still go through Google. What's changed is that ranking for broad how-to queries no longer pays off the way it used to, because the AI answers those queries directly. Spend less time chasing rankings on generic queries and more time on writing the specific, expert content that gets you onto AI shortlists. The full picture is laid out in what replaces SEO when buyers stop Googling.

Which AI engine should I focus on first?

ChatGPT has the largest active query volume from buyers right now, so it's the one most worth showing up in first. Microsoft Copilot is the easiest to earn citations in because it pulls heavily from helpful written content (this is where BakingSubs got most of its 144,321 citations). Perplexity matters because it shows sources prominently, which means a citation there is highly visible. Claude weights author signals heavily, so it rewards clear personal bios and consistent authorship.

How long does it take to start appearing in AI shortlists?

The fastest signals show up in 6 to 10 weeks if you're writing on a tight, specific niche and publishing consistently. Generic content can take 6 months or never start. The shorter timeline isn't about volume. It's about specificity. A coach who publishes 8 posts about one exact buyer transition usually starts appearing in shortlists for that transition before a coach who publishes 40 generic posts ever does.

Is the AI funnel different for coaches vs consultants vs other expert-led businesses?

The shape is the same. The vocabulary changes. A consultant might be searched for as "who's good at SaaS pricing strategy for early-stage B2B?" rather than "who's a good coach for X". A mediator might be searched as "how do I find a workplace mediator who works with engineering teams?". The five stages are identical. The optimization at each stage is identical. Only the query phrasing shifts based on what your buyers actually type.

The first concrete step is to find out what stage of the funnel you're losing buyers at. If you're invisible at stage 2, no amount of homepage polish will help. If you're on the shortlist but losing at stage 3, the problem is your site, not your visibility. Run the AI Visibility Check to see where you actually stand across ChatGPT, Claude, Perplexity, and Copilot, then build for the stage that's leaking.