518: How AI Is Transforming Ecommerce With John Lawson

Podcast: Download (Duration: 45:45 — 52.6MB)

AI for ecommerce is not a future thing. It is a right-now thing, and the highest-leverage use cases are product titles and descriptions, customer avatars, audience segment reports, long-tail keyword mining, and pulling actionable insight out of competitor product reviews. My guest John Lawson has built custom GPTs that do all of the above in a single interface, and he built them without writing a line of code.

This is a My Wife Quit Her Job interview with John Lawson, CEO of 3rd Power Outlet (a shoelace and apparel accessories brand), platinum eBay power seller, top-rated Amazon merchant, small business influencer of the year, and author of “Kick Ass Social Commerce for E-preneurs”. John has been actively using AI in ecommerce for years and is one of the sharpest practitioners I know.

Here is his complete playbook for applying AI to a physical-products business in 2024.

Get My Free Mini Course On How To Start A Successful Ecommerce Store

If you are interested in starting an ecommerce business, I put together a comprehensive package of resources that will help you launch your own online store from complete scratch. Be sure to grab it before you leave!

Click Here To Download

Key takeaways

• ChatGPT is one year old. It went from text-only to voice, vision and image generation in that single year. Sellers who wait for the tooling to settle down are going to lose to competitors who are learning it now.
• The top ecommerce use cases: SEO-rich product titles and descriptions, customer avatars, audience segment reports, long-tail keyword reports, and mining competitor reviews for product-improvement insight.
• Custom GPTs let you package your best prompts, methodology, and reference docs into a reusable app. John built his (EcomAIBoss.com) using natural language, no coding.
• The best review-mining trick: pull all the reviews for a competitor product, throw out the top and bottom scores, and have ChatGPT extract the middle-range complaints and suggestions. That is the shortlist of features that would make a better product.
• GPT-3.5 still handles almost every text task well if you know how to prompt. GPT-4 unlocks images, plugins, browsing and custom GPTs.
• On-device LLMs are coming. John predicts a ChatGPT 3.5-class model running locally on your phone within about a year, which changes the privacy calculus for sellers worried about handing their data to a large platform.
• Content is king, context is queen. Facebook and Instagram ads worked because they layered context on top of content. The next platform where that plays out is TikTok, and TikTok Shop is where a lot of the emerging opportunity lives.

What ecommerce use cases is AI actually good at right now?

The strongest ecommerce use cases for AI in 2024 are written-content creation and audience research: product titles, product descriptions, SEO-rich category copy, detailed customer avatars, audience segment reports, and long-tail keyword reports. These are jobs sellers used to pay experts for or spend hours doing by hand. Now they land in the time it takes to write a good prompt.

The measurable impact is on sales, not just search rankings. John’s shoelace business is a good example.

Shoelaces sound generic until you realize the audience splits into skaters, sneaker collectors, everyday fitness buyers and distance runners, and the same product has to be pitched differently to each group. AI is what makes generating four separate on-brand descriptions for the same SKU practical.

The materials-and-weaves conversation that AI can carry for you

John’s specific win was getting AI to talk about the technical differences in his laces, like the materials used and the weave patterns, in a way that resonates with buyers. Explaining why “not all shoelaces are the same” is hard for a business owner who has been living inside the product. It is easy for a well-prompted GPT that has the specs in the prompt.

How to build a custom GPT for your ecommerce business (no coding required)

Custom GPTs are small apps that run on top of ChatGPT and are specialized for one repeatable task. John packaged his own prompts, methodology, and reference PDFs into one GPT that will write a product title and description, create a customer avatar, generate an audience segment report, and produce a long-tail keyword report. The build was pure natural language conversation with ChatGPT’s GPT Builder.

Step 1: Describe the app you want in plain English

Open the GPT Builder and describe what you want the GPT to do (“I want an AI bot that writes SEO-rich product titles and descriptions for ecommerce products”). The builder configures the GPT in the background and shows a live preview panel on the right.

The left panel is where you keep talking to it, and the right panel is where you test what you just built.

Step 2: Name it, brand it, and set the voice

The builder suggests a name and generates an icon using DALL-E 3 inside ChatGPT. It then asks about the voice you want (formal, informative, casual) and how detailed responses should be.

You answer conversationally: “make it professional and informative”, “always output five to seven bullet points”, “keep product titles under 200 characters”. Every answer gets written into the app’s instructions automatically.

Step 3: Upload your knowledge base as PDFs, docs and CSVs

The knowledge section is where the specialization actually lives. Upload PDFs of your best prompts, brand guidelines, previous top-converting copy, a CSV of your SKUs, or docs describing your buyer personas.

The GPT references those files when it responds. This is how John’s GPT knows his methodology instead of relying on generic ChatGPT defaults.

Step 4 (optional): Add API actions via Zapier

Custom GPTs can call APIs. You can wire yours to a Zapier action that reads a Google Sheet of live inventory, updates a row, or triggers a downstream automation.

You do not need to know how to write the API call yourself: give the GPT the URL of the API documentation and ask it to build the call, and it writes the code for you.

How to use AI to mine competitor product reviews for product-improvement ideas

The single sharpest AI trick John shared is competitor review mining. Give ChatGPT a competitor product’s SKU or ASIN, have it pull the reviews, throw out the top and bottom scores (the extremes rarely carry useful information), and analyze the middle-range reviews for concrete suggestions and complaints.

Those middle reviews are where you find sentences like “I love this product but it would be better if it did X”. Every “I love this but” is a product-improvement hypothesis.

Do this across the top three or four competitor products in your category and you have a ranked list of feature gaps to build against. John used this exact loop to identify the shoelace category features worth investing in. Same play works for any physical-products category on Amazon.

ChatGPT 3.5 vs GPT-4: which one should ecommerce sellers pay for?

ChatGPT 3.5 handles almost every text task an ecommerce seller needs (titles, descriptions, avatars, keyword lists) if the prompting is good. GPT-4 is where you pay for image generation with DALL-E 3, web browsing (knowledge cutoff is now current to April 2023 with browsing), plugins, and the ability to build and use custom GPTs.

If you want to package your workflow into a reusable app or generate product imagery, upgrade to GPT-4. If you are just writing copy, GPT-3.5 still gets it done, and John predicts GPT-4 becomes the free baseline within six to eight months.

Are the AI-copy startups going to survive as ChatGPT adds features?

Most of the standalone AI-copy startups are not going to survive. The historical parallel is Microsoft Word: third-party spell checkers were their own businesses in the 1990s, then Microsoft folded spell check into Word as a feature and the market disappeared overnight.

ChatGPT is doing the same thing with the GPT Builder and its expanding native features. If your product is a thin wrapper around an OpenAI API call, that feature is going to arrive inside ChatGPT itself.

The businesses that survive move up the value stack: done-with-you (guided workflows), done-for-you (agency services), or vertical specialization deep enough that a horizontal platform will not build it. General-purpose “write my product description” tools do not clear that bar.

Can I run AI on my own devices instead of sending my data to a big platform?

Yes, and the timeline is short. Models compressed enough to run on a laptop with roughly GPT-3.5 capability already exist in developer communities.

John’s projection is that within about a year (from early 2024), you will have a ChatGPT 3.5-class model running natively on your phone in its own environment, with no calls out to a cloud provider. That solves the privacy question for sellers who are hesitant to feed their business data (customer lists, product roadmap, financials) into a third party’s training pipeline.

The interim step for privacy-sensitive sellers: use the ChatGPT API with data-sharing disabled, use enterprise ChatGPT (which does not train on your data), or run a local model on your own hardware today. All three are workable in 2024.

How AI is going to change search and SEO for ecommerce

AI is going to reshape SEO by moving the top of the search results into an AI-generated answer, and the sites that keep winning will be the ones that produce content people actually engage with. John’s KPI has already shifted from “how does this rank” to “does the reader take action on this”. Keyword-stuffed pages that used to rank at the top are going to lose that spot to AI-curated summaries anyway, so the play is content that resonates deeply with a specific audience.

The bigger shift is discovery vs search. TikTok has trained users to discover rather than search, and that pattern is spreading. Sellers who build a following on a platform where their audience discovers new products are less exposed to whatever Google does to its results page next.

Content is king, context is queen: how to make AI copy actually convert

Content is king, context is queen. John wrote that line in his 2014 book “Kick Ass Social Commerce for E-preneurs”, and it is more true now than then.

Great content shown to the wrong person at the wrong moment converts at zero. Great content shown at the moment the person is deciding what car to buy converts at a much higher rate.

AI is what makes context practical at scale. Instead of writing one generic product description and running it in every ad, generate five variants keyed to five audience segments (from your GPT-built segment report) and match each variant to the placement where that segment lives. That is the difference between throwing spaghetti at the wall and serving each buyer what they were already looking for.

Which platform should an ecommerce seller build content on first in 2024?

Pick one platform, exhaust it, then expand. Multiplication by zero is still zero.

If nobody is engaging on your Facebook page, moving to Instagram and Pinterest and TikTok all at once does not fix that. Fix the first platform or move to the one where your audience actually lives.

If John were starting over today, he would start on TikTok. The pattern in every prior platform cycle (Facebook ads circa 2013, Instagram circa 2016) is that the emerging platform is where the arbitrage lives.

TikTok Shop is opening up right now and the opportunity is early. The China political risk is real, but the platform is not going away in the near term.

Where the future of ecommerce is heading (5-year view)

Nobody knows for sure, but the base-case bet is that something will eventually displace Amazon. Yahoo got displaced by Google, eBay got displaced by Amazon, and that cycle does not stop.

TikTok Shop and YouTube Shopping are the two most credible candidates for the next platform, and both are inching toward serious ecommerce ambition.

Google has always had the raw ability to be a shopping destination and has always been reluctant to compete too hard with its own SEO business. That reluctance may end.

The safer bet for individual sellers: build brand and audience on the platforms where your buyers actually discover new products, and treat every marketplace as a distribution channel rather than a home. Marketplaces come and go. Your audience is your audience.

Frequently asked questions

Do I need to pay for ChatGPT Plus to use AI in my ecommerce business?

Not for the basics. GPT-3.5 (free) handles product titles, descriptions, keyword lists and audience research well if the prompts are good. You need ChatGPT Plus ($20/month) to use custom GPTs, generate images with DALL-E 3, browse the web, and access plugins, and if you are packaging your workflow into a reusable custom GPT, the upgrade pays for itself quickly.

Will Amazon or Google penalize AI-generated product listings?

Amazon has started asking sellers whether they used AI to write their listings, and there are already Amazon listings with titles that read “I cannot fulfill this request. It goes against OpenAI use policy”, proof that sellers are bulk-publishing raw AI output without review. The safe pattern is to use AI as a draft and edit for accuracy, brand voice and compliance before publishing; raw output at scale is a policy risk.

What is a custom GPT and do I need to code to build one?

A custom GPT is a specialized app that runs inside ChatGPT. It packages your prompts, methodology and reference docs into a single interface that always behaves the way you configured it, and you build one entirely in natural language through ChatGPT’s GPT Builder (no code required). Optional API actions via Zapier let a GPT read and write to external systems like Google Sheets.

How do I use ChatGPT to research competitor products?

Feed ChatGPT the competitor product’s URL, ASIN or a copy of the top reviews. Ask it to summarize the recurring complaints and suggestions in the middle-range reviews (three and four star reviews carry the most useful signal), then list features the product is missing that reviewers requested. Repeat across your top three or four competitors and you have a ranked list of feature gaps for your own product.

Which AI tool should I focus on if I only pick one?

ChatGPT. Most other tools use it as their backend anyway, so investing in prompting skills for ChatGPT transfers to the rest of the ecosystem. Google’s Gemini and Microsoft’s Copilot are worth watching, and Google is the safer long-term bet because of its data advantage, but ChatGPT is the tool to master in 2024.

Ready To Get Serious About Starting An Online Business?

If you are really considering starting your own online business, then you have to check out my free mini course on How To Create A Niche Online Store In 5 Easy Steps.

In this 6 day mini course, I reveal the steps that my wife and I took to earn 100 thousand dollars in the span of just a year. Best of all, it's absolutely free!