STUMP BEZOS
ChatGPT is losing marketshare fast to Gemini and Claude. A year ago ChatGPT held 76.4% of the AI market. What is it now?
[ Answer at bottom of email ]
👀 NEW WAY of AMAZON SEO: NOUN PHRASE OPTIMIZATION
ReFiBuy's Andrew Bell just dropped a deep guide arguing that the keyword era is ending. The sellers who win in the Rufus/Alexa+ world will be the ones who optimize around complete product phrases, not single keywords. Here's the skim.
The big shift
Amazon search now runs on two layers working together. A9 (the old engine) still decides whether your product can be found. It reads your listing words and builds the candidate pool.
Sitting on top, Rufus/Alexa for Shopping interprets what the shopper is actually trying to accomplish, applies their context and budget, weighs the evidence on your page, and picks the winner.
The kicker: AI agents now take one prompt, like "find something for my kids and our home" and fan it out into dozens of searches by product type, room, style, audience, use case, and constraint.
You're no longer trying to rank for a keyword. You're trying to be eligible across the whole family of queries one shopping mission generates. Bell calls this query-plan coverage.
The three acronyms to know
ACO (Agentic Commerce Optimization): engineering your ASIN so it can be retrieved, interpreted, trusted, and selected across both layers. Good ACO is great SEO, but great SEO alone isn't enough anymore.
QPO (Query Planning Optimization): predicting all the search paths a mission might generate, then asking: how many can my product be found by, and how many can it honestly win?
NPO (Noun Phrase Optimization): the language layer that makes it all retrievable. It’s organizing your product content around the phrases that define what it is, who it's for, where it's used, what constraints it meets, and why it can be trusted.
What a "noun phrase" actually is
A phrase built around a head noun that gets more specific as you stack modifiers:
lamp → floor lamp → metal floor lamp → dimmable metal floor lamp for reading nook
Each modifier (material, style, use case, audience, constraint, proof) reduces ambiguity and unlocks more query branches you can answer. Shoppers and AI agents both search this way — in full descriptive phrases, not lone keywords.
The anatomy to build around: Head noun · product type · material · style/theme · use case · constraint (outdoor-safe, dimmable, lightweight) · audience · proof layer (UL-listed, official seller, ETL-certified).
Why this matters now: COSMO
Amazon's own COSMO research shows the system reads beyond attributes and titles to the commonsense relations linking products to human intentions. A phrase like "large coastal metal wall art for living room" is a compact relation statement.
It presupposes material, style, size, and room all at once. You can't directly optimize COSMO (sellers never see that internal graph), so treat NPO as relation-aware language optimization: structure content around phrases that imply the relationships that matter, then watch whether your marketplace signals improve.
Actionable rules for operators
Use accurate noun phrases that reflect real product fit. Never force irrelevant terms (don't add "memorial statue" unless it truly is one).
Separate parent phrases from child phrases (metal wall art vs. large abstract metal wall art for living room). Much of the demand lives in the child refinements.
Use mega phrases and noun stacks only when they read naturally. Organize meaning, don't spam.
Build phrase evidence from the market, not your imagination: autocomplete, ad signals, reviews, Q&A, customer language, and catalog data.
Read organic and paid signals together as they reveal different forms of demand and content-market fit.
Find where you're already visible vs. where relevant demand is still underrepresented.
Measure it as a system, not one metric. Track across: visibility (are you in the consideration set?), attention (does the language earn interest?), confidence (does the PDP kill uncertainty?), fit (does it satisfy the constraint?), conversion, demand capture across clusters, and learning from controlled changes.
⚡ Do This Now: 7 Specific NPO Moves
Map your query plan in 5 minutes. Type your main product term into the Amazon search bar and screenshot every autocomplete suggestion. Then type it into Alexa for Shopping ("show me [product] for [use case]") and note the follow-up prompts it offers. That list is your query plan — those are the child phrases you need to be eligible for.
Pull Search Query Performance (Brand Analytics → SQP). Sort by your top ASIN. List every phrase where you get impressions but a low click or purchase share. Those are phrases where you're visible but not winning. Fix the language and evidence there first, don't waste time on terms you already dominate.
Rebuild your title as a noun-phrase stack, not keyword salad. Structure it: head noun → product type → material → key attribute → use case/room. Example: "Metal Wall Art — Large Abstract Coastal Panels for Living Room." One coherent phrase a human and AI can both parse beats 200 characters of comma-jammed keywords.
Put one constraint or proof phrase in your first two bullets. "Dimmable," "outdoor-safe," "UL-listed," "dishwasher-safe," "fits standard 12-inch shelf." These are the exact filters agents apply when narrowing a mission. If the phrase isn't on your page, you get filtered out before comparison.
Add an audience + occasion line to your bullets or A+. "Gift for horse lovers," "for a first apartment," "for a reading nook." Agents fan missions out by recipient and situation. Most listings only describe the product, never who it's for or when it's used.
Mine your reviews and Q&A for real phrases. Ctrl-F your reviews for how customers actually name the product and its use ("used it for my camper," "held up in the rain"). Copy their exact noun phrases into your content. That's market evidence, not guesswork.
Set a single controlled test. Change one thing — add a use-case phrase to a bullet — and watch that phrase's SQP click/purchase share for 2–3 weeks before touching anything else. Measure whether the language moved visibility, not just gut feel.
The winning ASIN isn't the one that ranks for one keyword — it's the one that's eligible, relevant, persuasive, trusted, and machine-legible across the entire mission.
Dan Ashburn's team launched their own AI platform on Friday. Not another ChatGPT wrapper, not skills bolted onto Claude. Their own engine, built for one job, running Amazon businesses.
Every other seller AI lives inside a general model that gets confidently wrong with your money on the line. Titan controls the whole environment instead. What the AI sees, how it works, how it thinks, and what catches it when it's wrong.
You tell it the outcome you want and it deploys up to 8 specialist agents on your account. PPC, keywords, listings, data. It runs on over a billion US data points and 13M tracked keywords, joins your Slack like a teammate, and runs Missions in the cloud while you sleep. Nothing changes without your approval.
One seller connected their account and Titan found a 45% traffic leak in two clicks. They fixed it that day and sales jumped 25% the next.
Dan opened a free WhatsApp group, AI for Sellers, plus the master prompt guide his team runs on their own brands. First time he's let non members in.
Join free -> AI for Sellers WhatsApp group
🔭 YOU GOTTA SEE THIS
Are you artificially lowering your store's profitability just to increase your conversion rate?
Most e-commerce founders obsess over top-of-funnel conversion metrics or acquiring cheap email leads, but they are ignoring the single most important number: Customer Lifetime Profit (CLP)
In this episode of Marketing Misfits, Kevin and Norm chat with Matthew Barnes, an earthquake engineer turned e-commerce conversion scientist.
Matthew explains why traditional A/B testing and basic conversion rate optimization (CRO) are completely backwards. He breaks down how he built The Tool (Proteus Digital Lab) a highly advanced testing and intelligence platform that connects Shopify data, Meta ad touchpoints, and Klaviyo data to prove which customers actually generate profit and which ones you are acquiring at a loss.
If you run an e-commerce brand and want to understand how to optimize your store like an engineer, do not miss this technical, deep-dive episode!
🔥 Subscribe to the Marketing Misfits Newsletter
🌎 INTERESTING STATS
Click image for more info
🎤 YOUR BIGGEST COMPETITIVE EDGE ISN’T AI - IT’S YOU
If you were at Ecom Mastery AI featuring BDSS in Nashville in April, you probably remember Makenna Riley's keynote on standing out when everyone's automating everything, and why real connection is the one thing AI can't replicate.
Now Makenna and her mom/business partner, Forbes Riley, just launched a brand-new book: Pitch Secrets A–Z. It’s a fast, practical guide to communicating with confidence, growing your influence, and connecting in a way that actually moves people (and moves product).
If you don't know Forbes: 30+ years on TV, over $2.5 BILLION in product sales, and a career spent making entrepreneurs and brand owners unforgettable on camera, on stage, and in the boardroom. When she talks about pitching, sellers listen.
Because in a noisy AI world, your story, your energy, and your ability to connect are still the hardest things to copy.
🛠️ BDSN SOFTWARE TOOL of the DAY 🛠️
Put Google's AI Search Engine Inside Your Shopify Store
If you run a Shopify store, here's a stat that should stop your scroll: shoppers who use search convert at dramatically higher rates than browsers, yet most Shopify search bars are dumb keyword-matchers that return junk the second someone misspells a word or searches by intent instead of exact title.
Nimstrata fixes that by dropping Google Cloud's AI Commerce Search, the same discovery technology behind some of the biggest retailers on the planet, directly into your Shopify store. No replatforming, no dev team, no code changes. It installs in under 30 minutes and is fully managed.
1. Semantic Search That Actually Understands Intent Instead of matching strings, Nimstrata understands what the shopper means including synonyms, misspellings, and natural-language queries all return relevant results. Cotopaxi reported a +45% higher conversion rate from search after switching.
2. Revenue-Optimized Collection Pages Your collection pages get re-sorted automatically to surface the products most likely to convert, so browsing turns into buying instead of bouncing.
3. Personalized Recommendations Google's recommendation algorithms train on your store's data such as live user events, real-time catalog updates, and data enrichment to serve each visitor products they're actually likely to buy.
4. Enterprise Speed & Reliability Sub-200ms response times, a 99.95% uptime SLA, and infrastructure already proven across 1B+ searches. It's fast enough that shoppers never feel it working.
5. No-Code, Fully Managed Real-time catalog sync with Shopify, predictive autocomplete, dynamic filtering, and an analytics dashboard that’s all handled for you.
If search on your store still feels like it's from 2015, this is the closest thing to hiring Google's search team for a monthly subscription. Worth a look for anyone serious about squeezing more revenue out of the traffic they already have.
👉 nimstrata.com
🧠 AMAZON LETS YOU SEE EVERY AD THAT EARNS A SALE
For years, Amazon's ad reporting has been playing favorites by handing 100% of the sales credit to whichever ad got the last click before checkout. Every campaign that warmed the shopper up along the way? Zero credit.
Which means a lot of you have been cutting "underperforming" campaigns that were actually doing the heavy lifting.
That's about to change. Amazon has rolled out a new Attribution toggle in the Sponsored Products reporting console that lets you view multi-touch attribution right next to the old last-click model.
Why this matters for your account:
Sales credit now gets shared across every ad that influenced a purchase, not just the final tap.
You can finally see the full funnel including which campaigns drive awareness, which drive consideration, and which close the deal.
Your broad and upper-funnel keywords stop looking like dead weight. They'll get credit for the assists they've been quietly racking up.
You'll stop killing your supporting cast. Comparing the two reports side by side shows you which "losers" are actually setting up your winners.
This is the closest Amazon has come to showing you how shoppers actually convert instead of just who got the last click. Toggle it on, compare it against your last-click data, and think twice before you pause that broad campaign. It might be the reason your exact-match terms are printing.
🔥 MORE HOT PICKS 🔥
🥃 PARTING SHOT
“You may encounter many defeats, but you must not be defeated. In fact, it may be necessary to encounter the defeats, so you can know who you are, what you can rise from, how you can still come out of it.”
✌🏼 Have a great weekend.
See you again on Monday.
The answer to today’s STUMP BEZOS is
ChatGPT is down to 52.7% market share.
