Latest / Drop Ship Lifestyle Podcast / I Tried Using AI for Dropshipping… Here’s What It Got Wrong
Transcript
- 0:00AI is one of the most powerful tools I've ever seen for
- 0:03e-commerce research, content outlines, ad creation, product
- 0:07page optimizations, and even creating complex automations.
- 0:11But there is a major problem and that is if you trusted blindly
- 0:15it will confidently give you answers that are completely
- 0:18wrong. I've given UR LS and asked is
- 0:21this a drop shipping store and the answer is just wrong.
- 0:24I've asked it how many brands or suppliers a certain store
- 0:27carries and the count is completely off when I compare it
- 0:31to what I see manually. And if you're using that
- 0:33information to choose a niche to build your store, to contact
- 0:36suppliers or spend a dollar on ads, those wrong assumptions can
- 0:41turn into a very expensive mistake.
- 0:43And trust me I am not anti AI. Myself and our entire team use
- 0:47it every single day. The problem is when you use it
- 0:50as a judge instead of as an assistant.
- 0:53AI is great for helping you move faster, but it is not always
- 0:56great at helping you confirm what is true.
- 0:59O here is a problem that I've seen first hand many times.
- 1:03This is a chat I had where I said is this website drop
- 1:06shipping and I gave it a URL and thought for a couple seconds and
- 1:10told me no. Well that answer is wrong
- 1:13because when I go to that website and I manually go to
- 1:16their FAQs it says they have direct ship items that ship
- 1:19directly from the manufacturer. Those are drop shipping.
- 1:22So it confidently will give you wrong answers that again can
- 1:26completely change the trajectory of your business.
- 1:29So then I asked it another question.
- 1:31How many router tables on the same store are drop shipped?
- 1:35And then it actually went to the product pages, did some further
- 1:38research and it found nine drop shipped router table products on
- 1:42that website that shipped from the manufacturer.
- 1:45So again, it found something to use the drop ship marker, but
- 1:48only after a follow up question. Only after I manually verified
- 1:52that this website actually is drop shipping.
- 1:55So the problem isn't AI. Like I said earlier I love AI
- 1:59but the problem is that people will trust it blindly.
- 2:03And I think this happens for a couple reasons. 1 is out of pure
- 2:07excitement that they found this tool that just gets everything
- 2:10right the first time because they don't manually check the
- 2:12work and they just go for it. The other reason is probably
- 2:16laziness. But you have to overcome that if
- 2:18you want to actually use AI correctly.
- 2:21And in my opinion, the way to do that is to utilize AI for
- 2:24drafting, for organizing data, which is great at and the other
- 2:28is summarizing data from somewhere that already exists.
- 2:32That is a source of truth and we'll talk more about that
- 2:36later. Do not use AI for verification,
- 2:39do not use it for exact counts, and definitely don't use it for
- 2:44product specs. So I'm going to take you through
- 2:46a few blind spots that I keep trying to find ways around, but
- 2:50they keep coming up. And the first one like I just
- 2:52showed you is that AI guesses whether a store is drop
- 2:56shipping. Sometimes it just sees multiple
- 2:58brands on a store and it'll tell you that they're drop shipping
- 3:01even though they have a warehouse with all of these
- 3:03products. Sometimes it just thinks all
- 3:05Shopify stores are drop shipping, which is obviously not
- 3:08true. But AI doesn't always know that.
- 3:10And when you see or when AI sees ships direct on a website, that
- 3:14doesn't automatically confirm the business model.
- 3:17Again, further research is required.
- 3:19And at this point, in May of 2026, that further research
- 3:23still can and should be done manually.
- 3:25Blind spot #2 is AI will miscount brands and suppliers.
- 3:30So let's just say even in step number one, you manually verify
- 3:34that a website is drop shipping and then you ask AI, please give
- 3:37me a list of all the brands being sold on store xyz.com.
- 3:42Well, those brand counts from my experience are almost never
- 3:45correct. It can miss filters in the
- 3:48collections. It can miss pagination where
- 3:50there's different brands on multiple pages through a
- 3:52collection. Sometimes there's also duplicate
- 3:55brands and a big one is based on how the website it is reviewing
- 3:59is set up. It cannot find all the brands
- 4:02because sometimes there's a parent company that has 5 or 10
- 4:04brands under it and it will only pick up one of them.
- 4:07So again, this is a huge blind spot and it can literally cost
- 4:11you possibly half your business or more in potential sales If
- 4:15you're only using AI and not verifying what it gives you when
- 4:18you're building your high ticket store, you might have realized
- 4:21this yourself already, but AI loves to tell you you're
- 4:24absolutely correct. So if you say something like I'm
- 4:26thinking of selling Persian rugs, do you think this is a
- 4:29good niche for my store? It'll tell you you're absolutely
- 4:31correct and it'll give you a bunch of reasons to validate
- 4:34your belief. But in reality, there are
- 4:37multiple things that we actually need to verify.
- 4:40Is the average order value high enough?
- 4:41Is there enough monthly demand? Are there enough suppliers?
- 4:45At least 20? Is there little to no brand
- 4:47loyalty? Is it not too seasonal where
- 4:49it's not going to be worth our time?
- 4:51And is the niche not dominated by a few big retailers?
- 4:54And again, if you just ask it, is this a good niche, it's going
- 4:57to look for reasons why it is rather than actually checking
- 5:00all of the data. And like I showed you in the
- 5:03previous points, even even if it checks the data, it often gets
- 5:06that wrong. So that takes us to the 4th
- 5:09blind spot, which is that AI doesn't know supplier reality.
- 5:14For example, they don't know what the dealer approval process
- 5:17looks like. They don't know if map is
- 5:19enforced. Again, you can ask it this, but
- 5:21it's not going to give you correct answers with this type
- 5:24of search. It doesn't know if products are
- 5:26being drop shipped, but they're being private labels and it
- 5:29doesn't know if the products are even eligible for Google
- 5:32shopping. These are all things that you
- 5:34can and should do manually. And again, I am not an AI hater,
- 5:38so I'm going to talk about how to do things right with it.
- 5:40But first I want to put this out there as APSA so that you don't
- 5:44waste your time and build the wrong business.
- 5:46Another big one for blind spot 5 is that AI can and will invent
- 5:51product details. So for example, if you just ask
- 5:54it to write a product description for product ABC, it
- 5:58might give the wrong dimensions, it might list to the wrong
- 6:01materials, it might add a warranty that doesn't exist.
- 6:05It might put a wrong shipping timeline on it.
- 6:07It might say it's compatible with things it's not.
- 6:09And it might say accessories included that don't even exist.
- 6:13And this is a huge problem. And it actually reminds me of
- 6:16like issues we used to have pre AI when we would have huge CSVS
- 6:21of products from suppliers. And occasionally, you know, when
- 6:24sorting these CSVS, items and columns would become mismatched
- 6:28and maybe an incorrect price would be attached to the wrong
- 6:31product and it would just cause issues, right?
- 6:34So that's something that obviously, you know, is
- 6:36avoidable, but this is almost an automated way that those things
- 6:40can go wrong at larger scale. If you have AI just rewrite all
- 6:44of your product descriptions without double checking them,
- 6:47you could find yourself in a situation where all of a sudden
- 6:49customers are not happy because the product page says something
- 6:53that's actually not true. And that goes back to what I
- 6:56said earlier with having that source of truth that it works
- 6:59off of. So the way you could think about
- 7:01this is that every AI answer can fall into one of three buckets.
- 7:07The first is confirmed facts. And these are things that you
- 7:10verify yourself on site, whether that be yours or a future
- 7:13competitors or even a suppliers. Anything that is inside a
- 7:18supplier document that you give it as a starting point, that is
- 7:21also a confirmed fact. Anything in actual data that you
- 7:25feed to it. This might be a CSV.
- 7:27Again, that is a confirmed fact and anything that you tell it in
- 7:31real communication. For example, I want a new
- 7:33product description for product page ABC.
- 7:36Please base it off of this URL that's on my store.
- 7:39Please do not add anything or change anything in terms of
- 7:42features and specifications. Those are confirmed facts.
- 7:46The next things are educated guesses.
- 7:48And this goes into the first few blind spots that I referenced
- 7:51earlier where it is looking for patterns and they might be true
- 7:55but they might not. These need confirmation.
- 7:58And in my opinion you're better off not even starting with nits
- 8:01research with AI because the manual method is just so much
- 8:05better because you're looking at real data rather than having AI
- 8:09summarize it. And let's just say it tells you
- 8:11no, this store is not drop shipping when it is.
- 8:14You could be missing out on a massive the future competitor
- 8:17that you can source suppliers from.
- 8:19Now the unknowns are things that AI might tell you that you
- 8:22simply can't confirm. You can't find any sources
- 8:25online that validate that information.
- 8:27You go out, you even do manual research and you still can't
- 8:30follow it. For the things like this, if you
- 8:32want to, you could reach out to your suppliers and ask them to
- 8:34confirm, but most of the time this is when AI is hallucinating
- 8:38and you shouldn't use anything it is giving you when you cannot
- 8:41confirm it off of your real data or what you find on your future
- 8:45competitors or supplier stores. So with all that being said, how
- 8:49do we use AI the right way? Well, for generating ideas for
- 8:53example, different blog posts we could write, different content
- 8:56pages we could build. I love it for that.
- 8:58But I start with that list and again, I don't just take it at
- 9:02face value. I then disqualify aggressively
- 9:06by going out and gathering real world evidence, seeing if I
- 9:09could find what it's telling me is actually true.
- 9:12This is when I'm verifying manually.
- 9:14From there I'm using AI for execution again, once I have
- 9:18that source of truth to base things off of.
- 9:20And even then, when it creates content or assist in it, I audit
- 9:24everything before scaling and starting to push this
- 9:27information to the world. Or else what do we do?
- 9:30We push a bunch of untrue, unverified information out.
- 9:34And even if it gets results, I'm telling you these are going to
- 9:37be results that lead to problems down the line because your
- 9:40customers and audience will notice.
- 9:42And not only that, if it doesn't get, it's probably because
- 9:45things are incorrect and people aren't even finding what it is
- 9:48you have to offer. So going back to how I started
- 9:51this video, you should use AI as an assistant, but never as the
- 9:56final judge. It can help you write faster.
- 9:59It can help you organize faster. It can help you launch and scale
- 10:01faster. But verification still matters
- 10:04and it matters now more than ever.
- 10:07So if you have any questions, let me know in the comments.
- 10:09I'm really curious to hear your thoughts on this.
- 10:11And again, I'm putting this out there as APSA because I audit a
- 10:15lot of stores and a lot of people, I'm telling you, they
- 10:18rely too heavily on AI while using everything it gives them
- 10:22at face value, making a bunch of changes, doing a bunch of work
- 10:26that easily could have been avoided because it's not for the
- 10:30better. So hope you guys got value from
- 10:32this one. I'll be back with the next video
- 10:34right here next week on Drop Ship Lifestyle.
- 10:37See you there.