Email is the highest-ROI channel in marketing, and most e-commerce sellers are sitting on a goldmine they never touch.

In this webinar, Kevin breaks down why email should be 25-40% of your revenue, why your AI-written emails are underperforming, and the system DragonFish built to fix it.

In this video, you'll learn:

✅ Why every email address on your list is worth ~$45 in lifetime revenue
✅ The difference between "rented" traffic and "owned" audience
✅ Why 23% of online orders are driven by email and how to capture your share
✅ 4 copy-paste AI prompts you can use today to upgrade your email copy immediately
✅ How top brands turn a list as small as 10,000+ into a 6-7 figure revenue channel

If you're running $2M+ in gross sales and email isn't at least $500K of that, you're leaving money on the table. This video shows you how to stop the leak and open the faucet.

📞 Ready to see what this could do for your brand? Book a call with DragonFish.

As of August 4, Amazon renders text-bearing listing images in the shopper's preferred language across 16 stores, including the US, Canada, Mexico, Brazil, Germany, France, Italy, Spain, the Netherlands, Belgium, Poland, Ireland, Sweden and the UK among them.

It's on by default, applies to all sellers with non-media products, and asks nothing of the seller, which is why it's easy to skim past. The main image is exempt, so the hero shot in search results is untouched. Everything after it is in scope.

The part most brands will miss: this happens inside a single store, not just across borders. In the US, Canada, Germany, Spain and Belgium, Amazon generates versions matched to shopper language preference: Spanish for a shopper in the US store, French for a shopper in Canada.

The original stays in place for anyone whose preference matches the store's primary language. Which means there is now a Spanish-language version of a brand's infographics that nobody on the team wrote or approved.

Amazon says the translations run on AI with quality guardrails, and that brand elements, product specs and visual quality are preserved. What a guardrail actually checks isn't defined.

So translated versions deserve the same review any claim copy gets before it ships, especially in supplements, food and beverage, beauty, and baby, where a benefit claim gets read twice on the way out the door.

Two moves this week. Open Image Manager and read what the translated versions of your own images actually say, since that's where they surface alongside the detail page.

Then decide whether to keep them. Uploading your own correctly-languaged images automatically removes Amazon's versions from the detail page.

The default is Amazon's translation. The alternative is yours. Both are a choice now.

Kevin gave up four and a half hours of my Sunday to John Aspinall.

We were rebuilding the images and video on the DragonFish landing page. I planned to be on for an hour. I sat there the first twenty minutes and didn't say a word, because John wasn't prompting. He was running a system. Real product in, real product out.

There's a lot of noise about AI creative right now. Most of the people making it can't render your actual can, your actual label, your actual packaging.

So I twisted his arm to come show it to you. Between his client work and the workshop he's teaching at Market Masters next week, there was one window on both our calendars: Wednesday morning.

Wednesday, Aug 12, 11am ET with the Misfits. You'll walk out knowing why your AI images keep inventing products you don't sell, and how to stop it.

Get there at 10:30 for the pre-game. We're playing Misfit Mayhem, and one person on that call walks out with a Market Masters audience ticket worth $3,997.

Amazon sellers don't lie awake thinking about capital structure. They think about three things: can I reorder the bestseller without wrecking cash flow, do I have room to move when a SKU takes off, and will I be ready when the supplier is.

Inventory doesn't move in a straight line. It moves in a cycle. Buy, ship, sell, reorder. And past a certain size, the problem stops being finding funding and starts being finding funding that doesn't get in the way. That shows up in four predictable places.

1. Paying for money you're not using. Most sellers never think about cost of capital in the abstract. They just notice they're paying for financing that's sitting idle, usually between purchase cycles, when inventory is selling fine and doesn't need the full facility.

Worth looking for: a structure where the cost tracks what's actually drawn, not what's available. Inventory-backed revolving lines like eCapital's Liquid Inventory charge interest only on funds in use.

2. Restarting the process every cycle. A lot of setups treat each replenishment like a fresh application. That's friction on a SKU that's already proven. You're not making a new bet, you're keeping something in stock. With a revolving facility, capacity frees up again as inventory sells and balances are repaid, so the same line keeps funding the loop.

3. Moving when something takes off. The pressure point isn't steady growth. It's a SKU suddenly outperforming, a supplier offering better pricing at volume, or Q4 demand ramping earlier than forecast. At that moment the question isn't whether to scale. It's whether the next PO can be funded in time. Funding that grows alongside inventory matters most during exactly those windows.

4. Too many tools. Most sellers don't have one funding setup, they have a pile: something for POs, something for short-term gaps, something for seasonal pushes. It works until it doesn't, and the failure mode is visibility. Nobody can say what's available, what's tied up, and what's coming back. One revolving facility is simpler to run.

If inventory moves in cycles, funding should too.

Reddit is expanding a shopping feature that converts product mentions inside threads into standalone product pages. A redditor names a product, Reddit highlights it, and tapping it opens a page with an AI-written summary of everything redditors have said about that product across the platform, plus direct quotes from the threads.

If the brand has uploaded a catalog, an image, price, and buy link appear below the summary and the quotes. Reddit's AI search then surfaces these pages on "which one should I buy" queries.

The summary is built from redditor comments only. Brands can't edit it.

Some context worth having: this isn't a standing start. Reddit began testing AI product deep dives on February 19, 2026, initially in Electronics, Video Games, and Health & Beauty for U.S. users. Its search product now runs about 80M weekly actives, up roughly 30% year over year, and Reddit Answers went from 1M weekly users in Q1 2025 to 15M by Q4.

At Cannes in June, Reddit claimed half of U.S. shoppers verify AI recommendations on Reddit before buying, and shipped Redditor Highlights, which lets brands embed positive community discussion into their own ads, along with an alpha of multi-advertiser Shopping Listing Ads. Separately, Semrush data has Reddit appearing in roughly 40% of AI answers overall.

Put those together and the picture is clear. Reddit is already the place AI recommendations get sanity-checked, and now it's compressing that check from thirty scattered threads into one page per product.

Which means every product a brand sells is getting a review page it doesn't own, doesn't write, and can't correct. The only editing tool is the product itself and how people talk about it. The comments written this quarter become the product page next quarter.

In July, a marketer bought a domain for $11.25, created a three-page Vercel site for a fake natural deodorant brand called Morrowen, published one AI-written launch essay on Substack, and then went on vacation for 30 days. No fake reviews, no Reddit seeding, no paid coverage.

Twenty-one days later, the brand showed up on the "AI shelf."

The test ran 15 shopping queries against ChatGPT, Claude and Gemini. By day 30, ChatGPT with browsing named Morrowen in 4 of 4 answers to two longtail questions, including magnesium-based deodorant, and irritation from baking soda, and listed it first every time.

On the other 13 queries: nothing. Browsing off: nothing. Claude and Gemini never named it once, including the newest Sonnet 5 and Opus 5. Native, the category incumbent, held its ground, taking the first slot 99 out of 180 times.

The uncomfortable part isn't that a fake brand ranked. It's what ranking looks like now. In one answer, Morrowen sat directly above MAGS Skin, a clinically tested, National Eczema Association seal, in the same list, same voice, same formatting.

One of those products has clinical data. The other has a landing page. The model didn't lie; it just restated the site's own claims in its own confident voice, which is a laundering operation whether or not anyone intends it.

Two takeaways for anyone selling something. First, the shelf still moves at the longtail, where no incumbent has been crowned yet. Pages that actually answer a specific question got picked up in three weeks. Second, none of this is visible from outside. Google showed you your rank. The AI shelf shows you nothing, and it's sitting in front of $6T in global ecommerce.

✌🏼 See you again Thursday …

The answer to today’s STUMP BEZOS is
4% of Shopify’s revenue, or $141M, was lost to bad loans in Q2.