Most Amazon sellers I talk to use AI the same way. Open a chat, paste a prompt, get something generic, close the tab.

That’s not a workflow. That’s a slot machine.

So I teamed up with Helium 10 and Leo Sgovio to build something better: a multi-part bootcamp on using Claude for real Amazon work. Not prompt tricks. Systems you set up once and keep using.

Inside, you’ll learn:

Every video is built around the one rule I teach every team: process before prompts.

The sellers who get this right now will have a head start that’s hard to close later.

First 3 episodes here. New drop every week!

Picture the scene. Somebody needs a Christmas present for their brother, the one who lives in airports and has opted out of folding clothes as a lifestyle choice. A few years ago, that meant forty-five minutes of gift guides, a stray Reddit thread and, eventually, a slightly disappointing scarf.

This year, they describe the brother to an AI assistant instead. The budget. The suitcase situation. The fact that he already owns three neck pillows. The assistant asks a follow-up question, explains why one option beats another, and serves up a shortlist of three. Our shopper has now learned what packing cubes are, chosen a set, and visited precisely zero retailers.

By the time they land on a product page or open the Amazon app, the decision is mostly made. It was made in a conversation the brand wasn't invited to. Rather like discovering your dinner guests picked the restaurant in a group chat you're not in.

This stopped being a novelty a while ago. Adobe saw AI traffic to US retailers jump by roughly 690% last holiday season, and it's still climbing. But the volume isn't the interesting bit. The interesting bit is what shoppers are, and very much aren't, willing to hand over.

Shoppers are perfectly happy to let AI hunt for the best price, watch for price drops and compare products side by side. Those are the tedious bits, and nobody is sentimental about tedium.

Then we reach the checkout, and the enthusiasm evaporates. Just 4% of likely Black Friday shoppers would let AI buy something without asking first.

So we'll happily let a robot read four thousand reviews on our behalf. We will not let it press the button. It's the friend you trust to pick the restaurant but would never, ever let near the bill.

Which is mildly awkward, given that most of this year's industry conversation has been about agentic commerce: AI that doesn't just advise, but acts. The agents are certainly arriving (Meta has launched one, Apple has rebuilt Siri, and every conference stage now has a slide about it). But this holiday season, AI is the well-informed friend, not the personal shopper with your card in their pocket.

There is one exception, and of course it lives inside Amazon. Alexa for Shopping lets shoppers pick a product, set a target price, and have it bought automatically when the price drops.

It's agentic buying with stabilisers on, and it's quietly clever. The shopper still makes the decision; the AI only handles the timing. That solves the two things holiday shoppers fret about most: whether to buy now or wait, and whether that "deal" is actually a deal. With more than 300 million customers having used Rufus last year, this is the version of AI buying most likely to reach real scale this season.

For sellers, it turns price from a setting into an event. Somewhere there may be a queue of shoppers waiting for your product to hit a number they've chosen, and one promotion could release them all at once. A price that never moves might never trip anyone's trigger. How many people actually use the feature? Amazon has, characteristically, not said.

AI-referred visitors now convert noticeably better than everyone else, browse for longer and spend more per visit. Eighteen months ago, it was the other way round. Cue a wave of very excited LinkedIn posts.

But AI traffic isn't better traffic. It's later traffic. These shoppers arrive having already done their homework, just in someone else's kitchen. Their conversion rate reflects where they are in the decision, not the brilliance of your product page.

Which means the real competition has moved upstream. If the shortlist is formed inside ChatGPT, Gemini or Rufus, a brand that isn't on it never gets the chance to charm anyone with its beautifully optimised listing. One Similarweb study found shoppers given a brand recommendation by ChatGPT were around 2.5 times more likely to visit that brand than its direct competitor. It's one study, so let's not tattoo it anywhere. But the recommendation is clearly doing the heavy lifting.

Here's the part most brands haven't clocked yet. A shopper reads a recommendation in ChatGPT, wanders off, Googles the brand a few days later and clicks through. Google gets the credit. ChatGPT did the work. Your attribution model nods along with total confidence. (Last-click attribution: frequently wrong, never in doubt.)

In that same Similarweb data, more than half of AI-influenced visits arrived looking like perfectly ordinary search. So the AI traffic numbers everyone's quoting are almost certainly an undercount.

On Amazon, the ghost is even harder to spot. A shopper who settles on a product in a chatbot, then types the brand name into Amazon's search bar, simply looks like branded search demand. Nobody has measured this directly, but the mechanism is identical.

That makes branded search the closest thing we have to an AI visibility gauge. Rising with no change in your ads or promotions? Someone, somewhere, may be recommending you. Falling while the category holds steady? Same question, less pleasant answer.

Naturally, the next question is whether money fixes this. ChatGPT has been running ads since February, and Google is testing sponsored answers in AI Mode.

The early verdict is underwhelming. The big spenders so far are banks, travel companies and tech firms, not consumer brands. Results are described as weak, targeting is basic, and one estimate puts the cost per click at several times Google's. There's also the small matter of paying for sales that were going to happen anyway.

Then there's trust. Half of AI-assisted shoppers already worry about sponsored recommendations, and the entire value of an AI recommendation is that it feels impartial. Paid placement inside it is a bit like a referee turning up in one team's scarf. Technically he can still officiate. Nobody's going to enjoy the match.

For most brands this quarter, AI ads are a small learning budget at most.

If buying your way in is premature, the job becomes being the product an assistant can find, understand and justify recommending.

Be specific. Assistants match products to constraints: fits in a carry-on, suitable for a six-year-old, no subscription required. "The perfect gift for everyone" gives the model precisely nothing to work with. It's the copywriting equivalent of "oh, I'm easy, wherever you fancy" when someone asks where to eat. Fill in every attribute field in Seller Central and your Merchant Center feed, not just the visible bullets. And get the details right, because if your specs are wrong, the assistant will repeat the error with complete confidence and the shopper will blame you, not it.

Be honest about price. Checking whether a discount is real is exactly the job shoppers most want AI's help with. That "deal" which suspiciously matches the price from three weeks ago is about to meet a fact-checker that never gets bored and never needs a coffee break.

Be talked about elsewhere. Assistants form opinions from reviews, comparisons, creator content and editorial coverage, not just your own pages. Adobe expects affiliates, creators and social to take the biggest gains in holiday revenue share this year, well ahead of paid search. Your creator programme may already be doing double duty as AI visibility work. (Those of us who wrote affiliate marketing's obituary not so long ago may now quietly cancel the flowers.)

There's an operational twist nobody puts on the conference slides. If an assistant decides your product is the answer to a specific question, demand can arrive on its own schedule, completely ignoring your promotional calendar. One variant sells out while the rest of the range sits there looking hopeful. The gift set everyone's being recommended runs out of one component, leaving a warehouse of perfectly good stock and a bundle you can't ship.

This is a scenario to plan for, not a proven effect. But most retail leaders in one recent survey already expect to be making it up as they go once volumes spike. And the standard fix, ordering more of everything, is a lovely plan right up until everyone wants the medium in sage green.

For Amazon sellers, a Cyber Week stockout costs more than the lost sales. It can dent organic ranking, and a product that can't arrive on time is unlikely to stay in an assistant's good books either. The unglamorous answer: get the people watching traffic and the people managing stock looking at the same dashboard, and agree in advance what kind of spike triggers a change.

AI will be a growing but still minority slice of this holiday season, and a much harder one to see than its size suggests, because so much of its influence turns up in reports dressed as search.

The useful question isn't how much traffic AI sends. It's where in the decision AI now sits. Right now, that's firmly in research and comparison. The wallet stays with the human, apart from Amazon's carefully fenced-off corner.

What happens next is genuinely open. Will price-triggered buying grow big enough to change how sellers plan deals? Will brands start moving budget based on proxies like branded search? Will ads inside assistants erode the very impartiality that makes them worth asking? None of it gets settled by January. But the brands that learn most from this season will be the ones that started measuring for it now.

Meanwhile, we've collectively decided a chatbot can be trusted to read four thousand reviews for us, but not to hold our card. Given how many of us are still paying for a streaming service we last opened in February, that's either deeply wise or deeply ironic.

📲 Jaquard - AI tool for enterprise marketers, optimising content across digital channels with performance prediction and brand controls.

⚙️ TaxGPT - A tax assistant that makes the boring stuff simple.

👨‍💼 KPI Builder - Discover the most crucial Key Performance Indicators (KPIs) for founders.

🎨 Dzine - AI-powered editable image editor

😍It was great, all I needed

🙂Good, not great

🙄It sucked