Paul from Rankster laid this out recently and it's worth passing along.

Most of what TikTok Shop does for you never shows up as a TikTok Shop sale.

A shopper can buy from you in three places. TikTok Shop is easiest because they're already scrolling. Amazon is next because they trust it and their card is saved. Your own site is the holy grail and the hardest, because nobody hands a random website their card details when Amazon already has them.

So the videos run. People watch. Some don't buy on TikTok Shop. They type your brand name into Amazon instead, because that's where they buy everything.

Branded search climbs. Organic rank climbs. Sales you never paid for. That's the halo, and Paul's point is that it's the only part of month two that pays you back later.

You can't track it with the pixel

The TikTok pixel can't see Amazon. Amazon won't let a third-party pixel near a listing and isn't about to start sharing. Amazon Attribution is supposed to fill the gap. It doesn't when TikTok is involved. Paul's words: don't trust it, long story.

The customer sees the video, does nothing, then searches your brand on Amazon three days later and buys. No link. No click. Nothing to attribute.

So you measure the shadow instead.

Back in 2023 Paul ran ads for a grooming brand. Attribution reported nothing. He opened Brand Analytics, went to Search Query Performance, and tracked the brand term week by week. It climbed while the views climbed. No tracking link. Just two lines moving the same direction.

One gate: you can only judge a halo once you have serious view volume. A brand with 7.2 million views in 30 days has a halo to find. A brand with 12,000 views doesn't. Don't go hunting for one you haven't earned.

The tools finally caught up

The average TikTok Shop outreach tool now tracks this. Paul's rundown:

Reacher correlates your daily branded search against your TikTok views, with a lag setting so a video today matches an Amazon order three days later. It's the only one of the three that touches branded search.

EUKA puts Amazon sales, TikTok views, and TikTok Shop sales on one timeline, plus your top Amazon search terms. It has MCP access, so you can point Claude at it.

Cruva links TikTok products to specific ASINs, so you can ask whether this product moved when videos about this product landed.

Parallel lines, not proof. But it's the first time anyone put those lines on one chart.

Are you tired of paying creators for generic, overly polished videos that look great but fail to convert?

Most brands believe that high-production video is the key to stopping the scroll, but in the world of AI search and UGC (User-Generated Content), authenticity is what actually drives sales. In this episode of Marketing Misfits, Kevin King flies solo to sit down with Janvi Shah, the Harvard-educated co-founder of Hue.

Janvi breaks down exactly why her platform manages over 4,000 vetted micro-influencers to help top Fortune 500 brands and e-commerce companies turn authentic video into a massive conversion tool.

She reveals why you need to stop scripting your creators, why slightly critical 4-star reviews actually perform better, and how "Answer Engine Optimization" (AEO) is completely changing the way AI platforms like ChatGPT and Perplexity find and recommend your brand.

If you are an e-commerce seller trying to figure out how to scale video content without wasting your budget on the wrong influencers, you need to watch or hear this episode.

🔥Subscribe to the Marketing Misfits Newsletter

On September 2, Anthropic released Claude for Commerce. The pitch fits in five words: agents for both sides of the sale. Two AI helpers for online stores, all the code on GitHub, free for anyone to take.

What it is

Helper number one is a shopping assistant that lives inside your store. A customer types "I need kids' gift ideas under $45" and it answers like a clerk who knows every shelf. It searches the catalog, weighs the trade-offs, shows a couple of products, and builds the cart right in the chat. It also handles basic customer service and checkout.

Helper number two is a store manager assistant. Anthropic's demo shows a merchant named Avery opening a morning briefing: sales up 1.7%, 5 orders and 6 listings need attention.

Two listings are at zero stock. One SKU sells out in under two days. Returns on a travel pillow are running 11%, all citing firmness, so maybe the listing copy is wrong. One ad campaign returns $6.02 per dollar, another $2.42.

Avery types "we're nearly out of the reef decals, stage a restock so this month's pace won't run us dry." The assistant does the math (1.7 a day, 60 units is 35 days of cover), drafts the order, and waits.

That last part matters. The assistant never changes anything on its own. It drafts, you approve.

How it works

This is not an app you buy. It's a starter kit. Any store or marketplace can clone it, plug in their catalog, and turn it on. You pay Anthropic for the AI's thinking. They stay the brain, you stay the store.

It ships with eight working demos across retail, travel, telecom, and entertainment. Wix had a working agent in fifteen minutes. Fetch Rewards had both running in under an hour.

Safety rails are built in. The shopper agent is blocked from inventing prices or products. It has hard limits on cart size and refunds and hands off to a human past those. The merchant demo even catches two buyer messages that tried to trick it into refunding orders and cutting every price to $1. It flagged them as fraud.

Who's behind it

Launch partners: Shopify, Klaviyo, Square, Wix, Intuit, Priceline, Visa, Mastercard, and Accenture. Shopify's Vanessa Lee said they want merchants "everywhere customers are shopping, and increasingly that means agent conversations."

Anthropic claims carts up to 35% larger, 60% higher purchase completion, and 40% higher conversion for small businesses. Vendor stats. Treat them that way.

Why it matters for sellers

First, this is not a one-off. Anthropic has been building the plumbing for months: a master product catalog, offer matching, product crawling. Same building blocks Amazon and Google run on. Their pattern is to dip a toe in, then go wider and deeper.

Second, a lot of paid tools just got a free competitor. On-site search and personalization vendors. Shopify apps that do inventory alerts. Marketplace platforms like Rithum, Sellercloud, Sellbrite, and Teikametrics. Anthropic just gave away a basic version of all of it. Small today. Small is how they start.

Third, the merchant side is built for smaller sellers, not enterprise. One operator, a few hundred orders a week, a dozen things to fix each morning. That's you.

Amazon has quietly moved a big chunk of product discovery into Alexa+ and Rufus, where shoppers ask questions instead of typing keywords. "What's a good travel pillow that won't go flat?" gets a short list, not a search page. If your product isn't on that list, you're invisible, and Seller Central won't tell you.

Lumian built a tool that pulls the curtain back.

You drop in an ASIN. The tool reads your listing and generates the questions a real shopper would ask Alexa about that kind of product. You can edit the list before it runs. Then it actually asks Alexa for Shopping those questions, records where you rank (and who ranks above you), and shows you which questions your listing can't answer.

The output is a report: this ASIN showed up in 7 of 10 discovery prompts, here are the 5 questions your listing content leaves unanswered, here's what to fix. It resolves parent/child variants so you're not chasing the wrong SKU, and the reports are shareable, so agencies can hand them to clients.

Most "AI visibility" tools guess. This one runs live queries against Alexa and shows the actual rank. That turns a black box into a checklist. The gap analysis is the useful part: it maps the difference between what shoppers ask and what your bullets, A+ content, and Q&A actually say. That's the same fix that helps you in Rufus, ChatGPT, and Google AI Mode, so the work compounds.

Try it: lumian.ai/alexa

Two new data sets landed this month on what happens when a shopper clicks through from ChatGPT, Perplexity, or Gemini to a store.

Ethercycle pulled 24 months of raw analytics from 10 established Shopify stores. That's $274 million in revenue, 94 million sessions, and 2.2 million orders. Year to date in 2026, AI assistants drove 0.10% of revenue.

About $75,000 out of $76.5 million. Small. But those visitors converted at 2.66% against a 1.91% portfolio average, and 85% of that revenue came from first-time buyers. Ten stores, one agency, last-non-direct-click attribution, so take it as a signal, not gospel.

Adobe went bigger. Across more than a trillion US retail visits, AI referrals in July 2026 were up 62% year over year and converted 60% better than non-AI traffic. Add-to-cart was 28% higher, engagement 14% higher, bounce rate 33% lower. It's vendor analysis, not an incrementality test, but the direction matches Ethercycle's.

Your warm email list can't reach strangers (cold can - checkout Dragon Fish). This channel does, and the strangers show up pre-sold. They already asked the assistant, got a shortlist, and clicked through to check one item. The page they land on decides the sale.

What to do this week:

Tag AI referrers in your analytics so the traffic shows up in reporting at all. If you can't split AI-referred orders from direct today, that's the whole job. It's an afternoon with an export and an agent that knows which referrers count.

Then run a monthly pull with three numbers per cohort: conversion rate (AI vs. everything else), contribution margin after shipping and returns, and repeat-purchase rate at 90 days.

Pull the ten pages your AI traffic hits most and read them like a buyer who already did the research. Accurate attributes, real stock status, current price, shipping and returns in plain language. Then compare that cohort's gross margin against paid search before you decide what it's worth.

✌🏼 Have a great College Football opening weekend (Gig ‘em).

See you again on Monday.

The answer to today’s STUMP BEZOS is
Facebook will earn $83.29 per user.