I lost $340 on my first dropshipping product.
Not in ad spend. In stock I pre-ordered from a supplier because I was so convinced the product was a winner. Waterproof phone pouches for swimmers. Google Trends showed a seasonal spike. I’d seen a few TikTok videos of people using them at the beach. I thought I’d cracked the code.
The supplier had a 28 day shipping time. By the time my first customer received their order, summer was over in most of the US. I got two refund requests and a passive-aggressive review that said “arrived in October, needed it in July.”
That failure taught me more about product research than any YouTube course I’d ever watched. And when AI tools started getting actually useful, I rebuilt my whole process around them but not in the way most people do.
Here’s what actually works.
The Problem With “Winning Product” Lists
Every week there’s some guru posting a list of “10 winning dropshipping products for this month” and honestly, by the time that list is published, those products are done.
The moment something hits a public “trending” list, a hundred other stores have already launched it. You’re not getting in early you’re getting in late and paying more in ads to fight for scraps.
Real product research isn’t about finding what’s already winning. It’s about finding what’s about to win or finding a better angle on something that’s already selling.
That shift in thinking changed everything for me.
Where AI Actually Helps (And Where It Doesn’t)
I use AI mostly ChatGPT and Perplexity as a thinking partner, not an oracle.
Here’s what I mean.
When I’m exploring a niche, I’ll ask ChatGPT something like: “What are common frustrations people have when working from a home office that products could solve?”
Not “give me winning products.” That kind of prompt gives you garbage. Generic stuff like “standing desks” and “blue light glasses” nothing actionable.
But the frustration-framing question? That gives me things like:
- Cables constantly falling off the desk
- Webcam positioning being awkward for video calls
- Room echoing during Zoom meetings
- No good way to store a laptop vertically when using an external monitor
Those are real problems. Some of them already have solutions on Amazon. Some have weak solutions with terrible reviews. That’s where the opportunity lives.
I also use Perplexity when I want to quickly cross-reference whether a product idea actually has supplier availability, market presence, and what people are saying about it across forums and reviews. It’s faster than jumping between five tabs.
But here’s what AI cannot do for you: it cannot tell you whether a product will actually sell, whether your specific supplier is reliable, or whether you can make the margins work. That part is on you to verify in the real world.
My Actual Research Process, Step by Step
Step 1 Start with a frustration, not a product
I pick a general audience and ask: what problem are they solving today with something clunky, expensive, or outdated?
Remote workers. Parents with toddlers. People who travel for work. Dog owners in apartments. I pick one and go deep.
Step 2 Use AI to brainstorm the gaps
I’ll spend 20 minutes with ChatGPT just exploring. I ask about problems, then I ask what products currently exist, then I ask what complaints people have about those products.
That last part is gold. If the existing solution has consistent complaints too bulky, breaks after a month, doesn’t fit most laptop sizes that’s your product brief.
Step 3 Validate on Amazon before anything else
I go to Amazon and search the product category. Then I sort by “most reviews” and read the 3-star reviews. Not the 1-stars (those are usually anomalies or competitor trolls) and not the 5-stars (obviously biased). The 3-stars are the honest, conflicted customers.
They loved the product enough to keep it but had a specific complaint. That complaint is the exact feature your version needs to improve on.
Step 4 Check Google Trends properly
Most people use Google Trends wrong. They see a spike and call it a winner. You want to look at a 5-year view and check whether interest is growing consistently, seasonal, or declining.
A product with slow, steady growth over 36 months is more valuable than a product with one massive spike last month.
I also check geographic breakdown. If 80% of the interest is from one country and you’re not set up to ship there affordably, that matters.
Step 5 Spend 15 minutes on TikTok
Search the product directly. Not the TikTok shop tab the regular search results. Watch the videos that have 50k-300k views (not millions, because millions means it’s already saturated).
Look for: Are people demonstrating it or reviewing it? Are there comments asking “where did you get this?” That question in the comments is one of the best demand signals I’ve found.
Step 6 Open the Meta Ad Library
Go to Meta’s Ad Library, search the product name, and filter to “active ads.” This shows you what competitors are actively spending money on right now.
If there are zero results, either you’ve found a gap or there’s no market. You need other signals to figure out which.
If there are 30 active ads from different brands, the product sells but the market is competitive. Study what angles they’re using and find a gap.
Step 7 Find your supplier and order a sample
I use AutoDS and AliExpress together. AutoDS for automation and tracking, AliExpress for initially vetting suppliers before committing.
Before listing any product, I order a sample. Every time, without exception.
I know that sounds slow. But I’ve been burned twice by products that looked great in supplier photos and arrived looking like they’d been packaged by someone’s upset teenager. You cannot sell a product you haven’t held in your hands.
Check the actual shipping time too. Not the estimated time on the listing place the order and track it yourself.
The Margin Math Nobody Talks About Honestly
A lot of dropshipping content shows you selling prices and product costs and makes the margins look juicy. What they don’t show you is the full picture.
Here’s how I actually calculate before committing to a product:
Take your selling price. Subtract the product cost. Subtract shipping. Subtract the platform fee (Shopify, Etsy, wherever you’re selling). Subtract payment processing fees (usually 2.9% + 30 cents with Stripe or PayPal). Subtract a return reserve I use 5% of revenue because returns happen.
Now what’s left? That’s your contribution margin before advertising.
If you’re planning to run paid ads, your cost per acquisition needs to come out of that number too.
If what’s left after all that is less than 25-30% of the selling price, the product is a difficult sell unless you can scale volume significantly or sell through organic channels only.
I walked away from three products last year that looked great on paper until I did this math properly. One of them had a 38% gross margin that dropped to 11% after fees and a realistic ad spend estimate. Not worth it.
Mistakes I Made That You Don’t Have To
Chasing viral instead of validated demand. A product going viral on TikTok and a product with sustainable purchasing demand are two different things. Viral gets you a traffic spike. Sustainable demand pays your bills in month four.
Trusting supplier shipping estimates. Order the sample. Track it. Time it yourself. Do not trust what the listing says.
Picking products that need explanation. If you have to spend your ad creative educating someone on why they need the product, your conversion rates will suffer. The best products are immediately understood. You see it, you want it, you know what it does.
Not checking return rates in the category. Some product categories electronics accessories, sizing-dependent items like clothing, anything with complicated setup have naturally higher return rates. Build that into your margin math before you launch.
Launching too many products at once. My profitable runs have always come from going deep on one or two products, not testing fifteen at once. When you spread thin, you learn less from each test and spend more overall.
A Real Example From Last Quarter
I was researching the home office space (again, because it’s consistently strong). Used ChatGPT to brainstorm frustrations, landed on “monitor arm wobble” the problem where monitor arms feel unstable and vibrate when you type.
Searched Amazon. Found that the top-reviewed monitor arms all had 3-star complaints about wobble and flimsy clamp design. Read about 60 of those reviews.
Checked Google Trends: “monitor arm” had consistent 5 year growth, no major dip, strong in the US and UK.
Found a supplier on AliExpress with a heavy-duty version and better clamp reviews. Ordered a sample. It arrived in 18 days (not the listed 10-15, which I expected) and felt genuinely solid.
Checked Meta Ad Library: a few competitors but none specifically positioning on stability. That was my angle.
It’s not my bestselling product ever, but it runs profitably on a small daily ad budget and has a 4.6-star average from real customers. That’s what a good product research process actually looks like methodical and a little boring, not a lightning bolt moment.
One Last Thing
The best product research skill you can build is pattern recognition across many attempts, not perfection on the first one.
You will pick products that flop. It happens to everyone, including people who have been doing this for years and know all the right steps.
What separates people who build something real from people who quit after three months is whether they treat each failed product as tuition or as proof it doesn’t work.
Every bad pick taught me something specific about margins, about suppliers, about what “seasonal” really means when your shipping time is 21 days.
AI tools speed up the research phase significantly. But the judgment you build from actually launching products, reading your own returns data, and talking to your customers? That part still has to be earned the slow way.
Affiliate disclosure: Some links in this article may be referral links. I only mention tools I’ve personally used. This is not financial or business advice results vary significantly based on niche, execution, and market conditions.
Muqit is an author at Insiders Desk, creating simple and practical guides to online business and digital income. He writes about business models, digital tools, SEO, content creation, and monetization methods, with a focus on clear explanations, realistic expectations, and informed decision-making. His content is educational and does not promise guaranteed income or business results.