BakeBetter Apps·Explainer Hub
← All Explainers
Affiliate × AI × Paid Media

The $200K/Day Affiliate Playbook

How Jordan ("the homeless affiliate") and Kirsta rebuilt their entire ad stack around AI agents, bid cap scaling, and hyper-personalized landing pages — plus every funnel, skill, and mechanism they exposed on this call.

$200K
Peak day ad spend
$40K+
Best profit day
225M
Views — Tai Chi offer
13
Tactics deconstructed
Claude Code SkillsMeta Bid CapsUTM Dynamic Landers Quiz FunnelsUpsell Refund HackCross-Market SwipesFaceless Scaling
CHAPTER 01

The AI Stack Every Affiliate Should Be On Now

"If you're just using ChatGPT, you're behind." Jordan and Kirsta's first point: the game moved from chat LLMs to agentic coding tools that can one-shot internal systems in a single prompt. What took them months of back-and-forth with Cursor last year — their Facebook ad launcher — Kirsta now rebuilds in a single session.

Quote

"We're in the phase where this is the second evolution… AI agents will build your Facebook ad launcher or your AI creative tool in one prompt. If you're not using it, you're behind."

— Kirsta

The 3-layer AI stack

1

Pick an agent-driven platform

Cursor, Anthropic's Claude Code, Antigravity, Lovable, or OpenClaw. Don't live inside ChatGPT — pick something that can execute, not just chat.

2

Build or install Skills

Create reusable Skills that load automatically when invoked — no more re-prompting the same system every session. Skills for copywriting, Shopify store builds, advertorials, ad launching, email flows.

3

Chain skills into production pipelines

One skill calls the next. Kirsta's agents now build full Shopify stores + landing pages via API — he's even letting agents build his email flows autonomously.

⚠ Watch-out on OpenClaw / local models

Josh admits OpenClaw is roughly "where Cursor was last year." You need a real local rig (gaming-grade RTX 5090 or AMD Ryzen AI Max / "Halo" workstation with heavy RAM) to run 48B+ parameter models locally. On a plain Mac Mini, you're only running the agent — not the model — and you hit API rate limits hard. For now: use Opus for planning, local for execution.

What AI still can't do well for landers

Copy, outlines, angles, advertorials → 90% automatable. The gap is images — not the generation, but which image matches which paragraph. The creative IP still lives in Kirsta's head. Jordan's hack: accept generic AI images, brute-force test 1000s of variations, and let Facebook's auction pick the winners before investing in custom creative.

CHAPTER 02

The Skill Chain: From Zero to Live Campaign

Jordan's exact production pipeline — each skill hands off to the next. When a campaign needs to ship, he invokes them in order. This is the assembly line that turns an offer into live ads in minutes, not days.

1

Market Research Skill

Built around Mark Builds Brands' YouTube prompts + Stefan Georgi-style deep research. Takes up to an hour to fully map the market, avatars, and desires.

2

Advertorial Rewrite Skill

Feed it parallel-market swipes (skincare → dental, both aesthetic). Never swipe same-market competitors — that's just rewriting someone's control. Rewrite a parallel market that hasn't hit your niche yet.

3

Image Ad Skill (native + ClickBank style)

Two separate skills — one per ad format. Always start with image ads before burning budget on video.

4

Congruence Loop

Run ads → identify winning advertorial → build new ads congruent to it → identify winning ads → build new advertorials congruent to them. The winners converge.

5

Video Ad Skill (angle-driven)

Only produce videos for angles the images have already proven. Don't waste video budget on untested concepts.

6

Launcher Skill

Manually logs into their custom launcher and fires the campaign — Kirsta hasn't wired the Meta API directly yet, and doesn't need to. Headless browser automation is fine.

💡 Pro tip — skill invocation

Skills solve the "context gets lost" problem. Instead of re-prompting a system every session, the skill auto-loads on invocation. Build skills for any repeat workflow: store builds, lander copy, email sequences, ad compliance checks, offer research.

CHAPTER 03

The Custom Facebook Ad Launcher

Kirsta's flagship internal tool — originally took months of Cursor work, now rebuildable in a single prompt. It's a Cursor-built web app hosted on App Platform, talking to a custom Facebook App that gives it permissions to all their Business Managers.

What the launcher does

  • Launches ads at scale (hundreds per day)
  • Performance analysis baked in
  • Custom bidding strategy (Kirsta's scaling model, in a few clicks)
  • Ad spying — inspect competitor ads inline
  • Offer management — auto-detects the offer from the ad account, pre-fills copy, headlines, URLs
  • Clean ads manager replacement
  • One-click post ID scaling — find winning ads by criteria → copy IDs → publish
Time saved per scale operation
~2 hrs → <30 sec

The normal flow (enter ads manager, find ad, edit, open Facebook comments, grab ID, create campaign, paste) collapses to a click-click-launch.

How to build one yourself

1

Register a custom Facebook App

Needed for API access. Install on your primary Business Manager → share access to linked BMs. Only needs Meta review if you productize and sell it publicly.

2

Build the web app in Cursor / Claude Code

One-shot prompt with a detailed spec. Kirsta emphasises: "give it crap input, you get crap output." Spend time on the requirements doc.

3

Host on DigitalOcean App Platform

Connect your custom Facebook App credentials. Use internally — no Meta approval required.

⚠ AI API bill reality check

Kirsta: "I don't even want to look at the monthly AI API bill. Are we actually saving money vs hiring a Filipino VA?" — Weigh quality-control time + API costs against human cost before over-automating.

CHAPTER 04

Bid Cap Scaling — The Real Unlock at $200K/day

The single biggest lever Kirsta used to push to $200K daily spend. Most advertisers are stuck on auto bid because that's what they were taught. Auto bid introduces volatility — Meta is an auction, inventory fluctuates daily, and auto lets the algorithm burn your budget chasing unavailable inventory.

The portfolio mix at scale

Bidding modelShare of spendJob
Bid capMajorityControl CPA at scale — force the learning down to your limit
Auto bidSupportingRefills the funnel with low-hanging fruit below the bid cap
Cost capSmall sliceSpecific use-cases; blend cautiously
Why blend auto + bid cap

"At exponential levels, you deplete all the low-hanging fruit below your bid cap threshold. If the bid cap stops spending and there's no funnel, Facebook thinks the bid cap is performing crap. That's an unravelling snowball going backwards."

— Kirsta

The bid-setting formula

Starting bid ≈ 0.72× to 0.80× of target CPA

Example: AOV = $100, target CPA = $80 → first bid around $72–78. If it's not spending, nudge 1.2–1.3× up. Don't jack bids aggressively high — that just turns the campaign into auto bid again.

⚠ If you're tempted to crank the bid

Don't. Fix the funnel. Kirsta: "In cases where I need to jack bids super high, it's better to look at the funnel and whatever else might be broken."

Testing sequence (new accounts)

1

Start on auto bid

Fresh accounts don't have enough signal for bid caps to spend consistently. Auto gets you data fast.

2

Get a feel for the right bid

Kirsta watches auto results, infers what the market bid should be (sub-CPA).

3

Layer bid caps on top for scale

Once Meta has data, bid caps will spend. Only now introduce them — never start there.

CHAPTER 05

Campaign Structure by Payout Size

Jordan structures ad sets differently based on the payout per conversion. This is the test framework the AI skill chain uses too.

Payout per conversionAds per adsetBudgetStructureWhy
$3 (e.g. loans)1 ad$10/day adsetABO — every ad gets its own adsetIndividual ad-level performance evaluation at low cost
Mid payout5 adsPer adsetEach adset = 1 angle/avatar/conceptTighter concept testing, still enough signal
$50–60+ (high payout)25 ads$50/day CBOLet Meta allocate spend across 25 adsTesting 25 ads individually at $50/day each = $5k/day burn — inefficient

The standard test batch

Testing budget per launch
25–50 ads

Jordan's default opener — 25 for smaller batches, 50 at most. Structure chosen per the payout matrix above.

CHAPTER 06

The Many-Lander Hyper-Personalisation Strategy

Kirsta's bet on where the market is going: "Whoever can pay the highest CPMs wins." Not the highest customer acquisition cost — the highest impression cost. Get there by raising conversion rate. Get conversion rate by matching each ad to a dedicated, angle-specific landing page.

Quote

"Dan Kennedy said whoever can bid the most to acquire a customer wins. I think what's actually true is whoever can pay the highest CPMs to reach a customer — that's who wins the auction."

— Kirsta

The approach — nerve pain example

Main market = nerve pain. Sub-avatars inside it have very different desired outcomes:

  • Wants to walk easier
  • Wants to drive again
  • Wants to play golf again
  • Wants to sleep through the night

Build one funnel path per desire. Headline: "How people are getting rid of nerve pain and getting back to driving in 6 weeks or less — without pills." Story, testimonials, before/afters all skewed to that avatar.

The skincare case study

💡 Winning angle Kirsta found

Core desire: "Look younger than my neighbour."
Landing page headline: "My neighbour and I used to look the same age. Now she looks 20 years younger than me. What happened?"
Story: Two women same age, AI-generated before/afters, one discovers the cream. Print.

Parallel-market angle pattern

Tinnitus

Veterans angle

"How I got rid of my tinnitus" — veteran story. Works because of gunshots/bombs association. High emotional resonance.

Nerve pain

Driving again angle

"6 weeks to driving again — without pills." Specific life activity → high intent clicks.

Skincare

Neighbour rivalry angle

Side-by-side ageing comparison → emotional driver is social, not vanity.

CHAPTER 07

Dynamic Headline Swap — The Low-Ticket Hack

Instead of building a full custom funnel per angle, run one VSL and hot-swap just the headline based on which ad sent the traffic. Same VSL, same body, same offer — just a dynamically-rendered headline matched to the UTM parameter.

1

Tag every ad with a unique UTM

The UTM carries the angle name — e.g. ?angle=veteran, ?angle=driving.

2

Custom JS reads the UTM at page load

Switches the H1 (and optionally the sub-headline) to match the ad's promise. Everything below can stay identical.

3

Use Vyopt (or similar) if you don't want to code

Kirsta's mentioned tool. Described as "super scrappy, built by some guys in a basement" — cheap, specific, works.

💡 Low-ticket application

Perfect fit for TSL/info-product funnels where ripping the full sales letter per angle is prohibitive. Hyper-specific pre-lander or advertorial → same core VSL → same order form. Massive CVR lift for near-zero extra build cost.

⚠ Testing landing pages has real cost

Facebook can read early signal on ads (CTR, CPC) without needing conversions. It cannot read if a lander converts — you have to spend on testing. Which is why lander testing should be quality over quantity, and headline swap is a cheat code.

CHAPTER 08

The Upsell Refund Killer (+ the AOV Move)

Optimise for upsells, not front-end purchases

Kirsta's publishing unlock: change the Meta conversion event from initial purchase to upsell purchase. Downstream effects:

  • Lead quality goes up exponentially
  • AOV goes up
  • Front-end "vanity" volume goes down, but real-revenue leads go up

The upsell copy → refund rate link

Refund rate change on one of their offers
20% → 12%

Zero product change. Just rewrote the upsell messaging.

Their old upsell said: "The product won't work unless you also buy this." That single sentence was telling buyers: "You just got sold something broken." Buyer's remorse, refund.

New framing, built off Alan's upsell matrix: Faster, easier, bigger. Upsell is a bonus accelerator — not a patch. Refund rate halved.

Principle

"The upsell impacts refund rates more than any other step of the funnel — more than the product itself."

— Kirsta
CHAPTER 09

Funnel Case Studies Exposed On The Call

Real offers Jordan & Kirsta pulled up live as references. Study these — they're every archetype.

$15M+ spend

Tai Chi offer

225M views on YouTube. Fully AI-generated. Rolled into truckers, seniors, multiple avatars. "AI slop is printing" — same mechanism, different demographic VSL per vertical.

→ Template replication at scale
Publishing co

Awesome REI (Awesome Real Estate Investing)

Did ~$100M / year. Houses multiple experts under one brand: lending, tax liens, wholesaling. Front-end $47 info + $5k/year high-ticket membership on the back-end. Quiz funnels — rare in info.

→ Portfolio-of-experts publishing model
Low-mid ticket

O Formula (oil brokering)

Unique bizop — "become an oil broker." Runs VSL direct, mid-ticket. Works because it's prestigious, passive-feeling, and sounds investment-grade. Ideal for 60+ audience ("Facebook's finest clickers").

→ Anti-saturation bizop angle
Acquired

Foreclosure Academy

Bash's old offer. Bizop for older, moneyed audience. Super unique mechanism. Sold for big money — acquirer still runs it at scale.

→ Unique mechanism = exit asset
Health publisher

Dr Josh Axe / Dr Pompa-level plays

AI-slop style ads → lead-gen kit → sales call → multiple millions/month. Doctors hate being marketers — the opportunity is partnering with them on front-end media buying.

→ Find the doctor, own the ads
App — $400K/day

Fasting / health apps (Lithuanian-owned)

Health Insider-style affiliates. Go negative for 3 months on acquisition; average retention 7–8 months. LTV covers everything.

→ Many-lander strategy originator
NHB graduate

YouTube automation (Kur-name) offer

Traffic → quiz → gated VSL → 5-day $27 evergreen challenge (drip videos). Built from a real challenge that did millions → evergreened cleverly.

→ Challenge → evergreen conversion
AG1-style exits

Affiliate-founded consumer brands

Most big scaling supplement/wellness brands are owned by former affiliates. The mindset shift (long-term brand vs short-term campaigns) is the gatekeeper — not skill.

→ Affiliate speed + brand longevity
CHAPTER 10

The Market Right Now — Compliance & Survival

Hit rates are down vs last year. Scaling is harder. Jordan blames the Meta auction tightening; Kirsta adds a likely regulatory pressure (FTC or DOJ letter) pushing Meta into aggressive compliance.

Symptoms they're seeing

  • Mass ban wave on accounts — their own BM gets friction adding payment methods now
  • Specialist ban-appeal friends can't keep up with inbound DMs
  • Meta filing lawsuits against black-hat advertisers
  • Bizop guys are the only cohort still ripping at scale
  • War + macro conditions compounding
💡 Survival stance

Jordan's read: the compliance squeeze is good for long-term. Black-hat exits the pool → white-hat publishers on ClickBank / brand-friendly platforms get their oxygen back. Position for post-squeeze — don't race the bans.

⚠ The scaling → FTC pattern

"Every affiliate that scales super big almost gets sued by the FTC. That's a marker of scaling super big." Big high-ticket scalers position as investments (AI, Amazon FBA, real estate) rather than "start a business" — which is exactly where regulators target. Certification and passive-income framings face the same heat.

CHAPTER 11

Team, Downtime & Offer Portfolio Building

Their team grew from 2 people to "frankly small but real" and instantly exposed an ops problem — weeks of under-allocated work because they hadn't learned project management yet.

What they do in downtime

1

Refine the tool set

Skills, agents, launcher improvements. Build the war chest.

2

Horizontal prep

They plan to horizontally scale 10–30 verticals via Shopify stores + advertorials when the next ramp hits.

3

Mindset resilience

Dips are normal. Last year's dip at this time is what funded this year's scaling infrastructure.

Pattern

"While we might not be killing it 100% of the time, we're innovating and building on our tools and team — so when the next boundary opens, we push through it with more firepower."

— Kirsta

Jordan's delayed-gratification problem (and fix)

Affiliate brain is wired for same-day launch → same-day kill. E-com sourcing + fulfilment breaks that loop. Jordan's personal commitment: take on projects with longer payoff horizons now that cash flow covers expenses. Stop picking only low-cycle-time offers.

CHAPTER 12

Cross-Market Swipes (The Jordan Rule)

The rule

"I don't do the whole ripping thing — that's kind of dumb. If you're running dental, give the AI a skincare advertorial to rewrite. Don't swipe your own market or you're just copying someone's control."

— Jordan

Parallel-market framework

Your marketParallel-market swipe sourceWhy it works
Dental (aesthetic)SkincareBoth aesthetic-driven buyer intent
Nerve painJoint supplements / back painShared "getting activity back" desire
TinnitusHearing / sleepShared neurological frustration pattern
Skincare — neighbour angleWeight loss — "friend" angleSocial comparison emotional driver
Real estate bizopOil / commodities bizopShared investment-feeling framing
💡 Creative steal method

Pull the winning advertorial from a parallel market your niche hasn't seen, feed to your rewrite skill, instruct: "rewrite for [my market], preserve structure and emotional beats, change only the product-specific details." You get a "new-to-market" creative that's already been auction-tested in a sibling vertical.

CHAPTER 13

Personal Brand & The Homeless Affiliate Effect

Jordan's public identity = "the homeless affiliate." The bit drove outsized attention — his earlier interviews out-viewed industry veterans in half the time. Key insight: you don't need to flex cars and watches to get affiliate attention — a strong narrative hook works better.

Distribution channels that matter for affiliate
X > Telegram > Facebook

X is "a cesspool of all the faceless people scaling crazy." Telegram groups host $70M/year operators. Facebook = brand-friendly, boomers, but still home to the single best marketing group (Alan / NHB lineage).

What makes affiliate personal brand work

  • Show face — but don't fake luxury. Narrative > flex.
  • Get interviewed on shows where affiliate managers and offer owners watch.
  • Conferences = recognition loop. Affiliate World etc. — the value is being the person people know.
  • Partner split: one "yapper" for networking, one operator for execution. Both needed.

Josh's model — publishing deals without selling

"On publishing calls I don't sell the creator. I'm so sold on the idea that I'm doing it with or without them. I'm just evaluating whether they want to be the face. If you believe in the creator, they feel it — and you attract rather than pitch."

— Josh
RECAP

Action Checklist — Ship This Week

This week

This month

This quarter

Closing principle

"AI is only limited by your imagination. If you've done the manual work yourself, you know what skills to build. If you haven't — your imagination is too skewed and you need operators to do the creative thinking for you."

— Josh