Steve Merrill

Helping small business owners find more success and fulfillment by escaping the entrepreneurial job and make more money.

[Steve Merrill @merrills]

There are two very different things an AI can do with your store. Read it, or buy from it.
Right now, almost all the action is on the first one. Get cited, get mentioned, show up in the answer. That's where the work is today, and it matters.
But it's only half the shift.
The next change is agents that check out on their own. A shopper tells an assistant to buy the thing, and the assistant completes the purchase without the person ever visiting your site.
The platforms are building the plumbing for it now, and it's coming.
Being readable gets your name into the recommendation. Being buyable will get you the order. When agents start completing purchases, a store they can read but can't check out on will watch the sale go to whoever got ready first.
So get readable now. Then start getting your product feed and checkout ready for the day an agent can finish the purchase.

[Steve Merrill @merrills]

Stop trusting the conversion rate your platform hands you.
Shopify and your analytics each count sessions differently, and the number they show you can be off enough to make you call it wrong. I have seen the dashboards disagree by a wide margin on the same store in the same week.
So I calculate it myself. Total orders divided by real visitors. One number, defined the same way every time, that I actually trust.
That is the number I use to judge whether a page change helped or a traffic source is junk. When a metric moves around based on which tool you asked, you cannot make a decision with it.
Pick your own definition of conversion rate. Compute it the same way every week. A number you control beats a prettier one you cannot explain.

[Steve Merrill @merrills]

Getting your store mentioned by AI is not a plugin you install once.
A lot of people treat it like a checkbox. Add some schema, drop in a file, done. Then they wonder why a competitor keeps showing up in the answers and they do not.
AI surfacing is daily research. The models change how they pull and rank. New questions get asked. Competitors publish. Your product line shifts. What got you cited in July can quietly stop working by October.
We treat it the way a good ad manager treats an account. Check what the AI is actually saying about the category. See who is getting named. Find the gaps. Feed the model better, clearer, more current information. Watch what moves.
The stores winning in AI search treat it as an ongoing practice. Set it and forget it is how you disappear. Keep researching, keep feeding it, and keep showing up in the answer.

[Steve Merrill @merrills]

One of the stores I work with keeps getting the same question from customers: how do I know you're real and not some drop-shipper in China?
Shoppers used to ask that on your website. Now a lot of them ask ChatGPT instead.
And the AI can only answer with what it can find about you. If that's thin, it describes you the same way it would describe any faceless store.
So for this client, we gave the AI real proof to work with. Product details pulled straight from the actual catalog. Real reviews. A real founder, a real story and a real location.
Nothing made up. If the AI has to guess, it either skips you or gets you wrong.
Trust has always been the whole game in ecommerce. The difference now is that a machine often answers the trust question before a shopper ever reaches your site.
Give it real, specific facts it can check. That's how you come across as a store people can actually buy from.

[Steve Merrill @merrills]

Adding product options can quietly lower your sales.
Every color, size, and variant you add is one more decision you hand the shopper. Past a point, choice stops feeling like freedom and starts feeling like work. People who feel that friction leave to think about it, and thinking about it usually means not buying.
I watched a store trim a bloated variant menu down to the few that actually sold. Conversion went up. Support tickets went down, because there was less to be confused about. The catalog got easier to run at the same time.
This runs against instinct. More options feel like more ways to say yes. On the page they are more ways to stall.
Look at your best product. Count the choices between landing on it and buying it. Cut the ones nobody picks. A simpler decision is a faster sale.

[Steve Merrill @merrills]

Read It or Buy From It: Getting Your Shopify Store Ready for Agentic Checkout
Most of the advice about AI and ecommerce right now is about getting mentioned. Show up in the answer. Get cited when someone asks ChatGPT for a recommendation. That matters, and we spend real time on it for the stores we work with.
It is also only half of what is happening.
The bigger shift is quieter. AI assistants are starting to complete purchases on their own. A shopper tells an assistant to buy the thing, and the assistant places the order without the person ever landing on your site. Those agent-driven purchases have multiplied several times over this year, and the major platforms are building the plumbing to make it routine.
That splits the job into two parts. An AI can read your store, or it can buy from it. Being readable gets your name into the recommendation. Being buyable gets you the sale.
Why readable stops short
Getting cited is a content game. Clean data, real answers to real questions, a site the model can crawl. Do that well and the AI starts naming you when someone asks for options in your category.
Then the shopper says order it for me, and the assistant needs a path to actually do that. If your store cannot hand an agent a clean price, live stock, and a way to complete the transaction, the recommendation turns into a dead end. The model mentions you, then buys from whoever it can actually check out with.
You can win the recommendation and still lose the order. That is the gap most stores do not see yet.
What buyable actually means
Buyable is a technical state, not a vibe. An agent completing a purchase needs a few things to be true at the same time.
A product feed that is accurate and current. Price, availability, variants, and identifiers that match what is really on the site. When the feed and the storefront disagree, the agent has no safe way to transact, so it moves on.
Live validation. The agent needs to confirm price and stock at the moment of purchase, not trust a cached version from last week. Platforms like Shopify are exposing ways for assistants to check that in real time.
A checkout an assistant can actually complete. The fewer custom steps and required fields between add to cart and order placed, the more likely an agent gets through it without a human.
None of that is glamorous. It is the same discipline that makes a store convert for people, pointed at a new kind of buyer.
The feed is doing more work than you think
For years the product feed was a Google Shopping chore. Now it is one of the most important assets you own for AI commerce.
When an assistant recommends and then buys, it leans on structured, first-party data about your products. Your own feed is the source of truth the agent checks, more than anything a third-party site says about you. If that feed is thin, stale, or contradicts your product pages, you are handing the agent reasons to skip you.
This is why we treat feed quality as a revenue project, not a maintenance task. Consistent titles. Real specs. Identifiers that line up. The plain facts a buyer, or a machine, would ask for before committing.
What to fix this quarter
You do not need to rebuild your store. You need to close the gap between readable and buyable. A short list gets most stores there.
Audit your product feed against your live storefront and fix every place they disagree.
Make sure price and inventory are validated live, so an agent trusts what it sees.
Strip the friction out of checkout. Count the steps and fields, and delete the ones that exist for you and not for the customer.
Keep your product data specific and honest, so the agent has facts it can act on instead of marketing copy it has to discount.
Each of these helps your human conversion rate too. That is the nice part. Getting ready for agents is mostly just running a tight store.
The bar moved
A year ago, being readable by AI was ahead of the curve. Most stores were not even thinking about it.
This year the shoppers who lean on assistants are letting those assistants buy. The sale goes to the store that is ready to be bought from, not only quoted.
Get your feed clean, your data honest, and your checkout out of the way. Readable was last year's bar. Buyable is the one that pays.

[Steve Merrill @merrills]

If you sell both direct and through retailers, the direct side is your marketing engine.
A lot of brands treat wholesale as the grown-up channel and direct-to-consumer as a side project. That has it backwards. When customers see your product everywhere and ask for it by name, retailers reorder to keep up. The demand you build direct is what pulls the wholesale orders through.
I have watched this with product brands over and over. Turn up the direct marketing and the reseller orders climb a month later, without a single new wholesale conversation.
Your ads and your content do double duty. They sell a unit today and they make your product something stores have to carry tomorrow.
So do not starve the channel that creates the demand. Feed the direct side and let it pull everything else along.

[Steve Merrill @merrills]

OpenAI hired one of the people who ran search ranking at Google.
That is a small headline with a big signal. The era of AI answers being a free-for-all is ending. Ranking inside AI is about to get structured, measured, and competitive, the same way Google search did.
Right now most stores show up in AI results by luck. Good luck runs out. Once there is a real ranking system, the brands that prepared early sit at the top and the latecomers pay to catch up.
We watched this exact movie with Google. The stores that took SEO seriously in the early days built an advantage that took competitors years and real money to close.
AI search is at that same early point today. The rules are still soft. Showing up still costs almost nothing.
The window where being early is cheap does not stay open long. Get your store readable and cited now, while it still costs mostly effort.

[Steve Merrill @merrills]

For about 600 years, the game was to spread information as far as possible. The printing press, radio, the web, all of it pushed content outward to more people.
AI just reversed the direction.
Instead of sending people out to a dozen sources, the model pulls everything in and hands back one answer. The reach that used to be the goal now happens inside a system you do not control.
For a store, that changes the job. Publishing and waiting for traffic is no longer enough. You want to be the source the model reaches for when it builds that single answer.
That means clean product data, real content that answers real questions, and a site the AI can actually read. The brands that get pulled into the answer win the sale. Everything else slowly fades from view.
That direction flip is the whole story. Build for the pull. That is where the sale happens now.

[Steve Merrill @merrills]

The funnel that books me 12 to 15 calls a week runs on about $100 a day in Meta spend.
One ad points to a free audit. The audit is the offer, not the call. People want to see their own store's score, so they raise their hand.
The audit gates the call. By the time someone books, they have already seen what is broken and they want it fixed. Roughly 40% of the audits turn into a booked call, because the audit did the selling.
Most people run ads straight to a calendar and wonder why nobody books. A calendar is a chore. A free look at your own store is a curiosity.
Give people a reason to raise their hand before you ask for their time. The order matters more than the budget.

[Steve Merrill @merrills]

Your cheapest ad is the one that defends your own name.
When someone types your brand into Google, that click should be yours. It almost always costs less than any other keyword you run, because you are the most relevant result for your own name.
Skip it and you leave the door open. Competitors can bid on your brand and sit above you, so a shopper who came looking for you sees them first. You pay in lost sales instead of ad spend.
A branded search campaign is a few dollars a day for most stores. It protects the traffic you already earned with every other channel. The person searching your name is the warmest lead you will get all week.
Most stores pour money into cold prospecting and leave their own name unguarded. Cover the branded term first. Then go find new people.

[Steve Merrill @merrills]

I ran a clothing brand to about $10 million a year before I sold it.
The number I lived by was simple. Revenue minus ad spend.
Gross sales feel good. They also hide the leak. You can post a record month and keep less than the month before, because you paid more to get there.
The number that runs a healthy store is what is left after you buy the traffic. Revenue minus ad spend. That is the pool everything else comes out of. Product, payroll, rent, you.
Once I managed to that number, the decisions got simple. An ad that grows gross sales but shrinks that pool gets turned off. A quiet email flow that costs nothing and adds to it gets more attention.
Most stores I look at can tell me last month's revenue in a second, then go quiet when I ask what a single order clears after product, shipping, and fees.
That gap is where budgets go to die. You cannot scale spend on a number you are guessing at.
Know what one order actually leaves you. Then grow.

[Steve Merrill @merrills]

Ironically, product copy that sounds too much like AI is the copy AI won't cite.
Think about why. When a description is all polish and adjectives, "premium," "elevate your everyday," "curated for the modern lifestyle," it carries no facts. A model reading it for a shopper finds nothing to verify and nothing to quote, so it passes.
The pages that get pulled into AI answers read like an operator wrote them. Materials. Measurements. Who it's for and what it actually does. The plain stuff a buyer would ask on the phone.
There's a lesson under this for all your content, not just product pages. The more your writing sounds like generic marketing, the less any reader trusts it, human or machine. Specifics build trust. Vague enthusiasm burns it.
Write your product pages like you're answering a real question from a real customer. That's the voice both shoppers and models reward.

[Steve Merrill @merrills]

Optimizing for AI search runs on a completely different protocol than SEO. Treating them as the same is why stores stall.
Old SEO was a popularity contest. Keywords, backlinks, climbing a ranked list of ten blue links. You were trying to be higher than the next result.
AI works more like a librarian than a scoreboard. Someone asks a specific question. The model assembles one answer from the sources it trusts and understands. There's no page two. You're either in the answer or invisible for that question.
So the work changes. Instead of chasing rankings, you're making your product legible and trustworthy to a machine. Clear structured data. Plain descriptions. Mentions on the sources the model already relies on. Consistency it can verify.
Stores that bolt AI onto their old SEO checklist keep waiting for a ranking that never comes. The ones that treat it as its own discipline start showing up in answers. Different game, different rules. Play the right one.

[Steve Merrill @merrills]

I watched an artist double her prices on everything over three years without lowering a thing.
No discount strategy. No race to the bottom. She spent those years getting her work in front of the right people. Gallery relationships. A real art installation in Venice she didn't even profit from directly, because a partner footed the build.
That installation worked as proof. It put her work somewhere that made people take it seriously, and the direct orders followed. When the perceived value went up, the price went up with it, and buyers came anyway.
This is the part ecommerce brands forget while they're fighting over ad costs. Demand gets built by where you show up and who's seen standing next to you.
Most stores try to grow by spending more to reach more people. The quieter play is raising what your brand is worth, so the same audience pays more and feels good doing it. Placement over volume. Every time.

[Steve Merrill @merrills]

You can reverse-engineer exactly why AI recommends your competitor and not you.
Here's the move. Open ChatGPT or Perplexity and ask the buying question your customer would ask. "Best waterproof hiking sandals for wide feet." Watch which brands it names and which pages it cites underneath the answer.
Those cited pages are the map. They're the sources the model trusts for that question. Some are review sites. A few are forums. One might be the competitor's own product page, written in a way the machine could actually read.
Now you have a to-do list instead of a mystery. Get your product mentioned on those trusted pages. Rewrite your own page so it says plainly what the model is looking for. Go earn the same signals the winner already has.
Most people treat AI visibility like a black box. It leaves a paper trail. Ask the question, read the citations, and go collect the ones you're missing.
What does an AI say when you ask for your own category? Worth finding out before a competitor does.

[Steve Merrill @merrills]

If your average order is $18, paid ads will quietly bleed you dry.
The math is brutal and most stores never run it. It costs roughly the same to acquire a customer whether they spend $18 or $80. So on a low-order-value product, the ad eats the whole margin and then some. You're paying to lose money on the first purchase and praying they come back.
There are two honest ways out. Raise the order value before you scale spend, with bundles, minimums for free shipping, and add-ons that make sense together. Or accept that this product grows through organic and email, not paid.
What doesn't work is pouring ad budget onto a cheap product and hoping volume saves you. Volume just makes you lose money faster.
Know your true profit per order first. Then decide if paid can even carry the weight. The ad platform will happily spend your money either way.

[Steve Merrill @merrills]

Without AI I'd have three or four more people on payroll right now.
I don't say that as a brag. I say it because I think this is the part of the AI conversation we're missing. Everyone's debating whether AI will take jobs. Meanwhile, companies are hiring slower, replacing fewer roles, and figuring out how to do more with the teams they already have.
I've got systems doing the work that used to need a full seat. Pulling data. Drafting the first version of everything. Watching accounts and flagging what changed. I hire when a human clearly beats the machine, and only then.
The adoption is slow and uneven, so the shift is hard to see month to month. Under the surface it's real. One unfilled role at a time.
The move for anyone running a small business is simple. Be the person operating the systems. Learn the tools while they still feel like an edge instead of a requirement. The leverage is sitting there for whoever picks it up first.

[Steve Merrill @merrills]

I run almost every client month to month. No term contracts. No lock-in.
People think that's risky. It's the opposite. It means I have to earn next month during this month, every month. The pressure keeps the work honest.
What actually retains a client is the report. Every week they see the same three things: where we were, where we are, and where we're going.
When someone can watch the line move, the renewal stops being a conversation. The card charges and nobody flinches, because the value showed up before the invoice did.
A contract makes someone stay when they want to leave. Good reporting makes them not want to leave in the first place. One of those is a much better business to run.
It also changes how I sell. I don't have to promise a twelve-month transformation to justify a signature. I promise what my team will actually do this month, then show the numbers that prove we did it. The proof carries the relationship, so I never have to oversell to keep it.
Show the progress often enough and retention takes care of itself.

[Steve Merrill @merrills]

https://t.co/cBbu6LuE7T