OpenAI recently rolled out ChatGPT shopping integrations with Instacart, Target, and DoorDash, promising to turn the AI assistant into your personal shopping helper. The pitch sounds compelling – compare prices across stores, build grocery carts, streamline your weekly shopping routine. I connected all three to put them through their paces. What I found was a masterclass in AI overpromising and underdelivering – especially when compared to what I already use for grocery shopping.
Testing the Instacart Integration
I started with a simple test. I gave ChatGPT my standard weekly grocery list with the prices I normally pay at Ralphs for pickup. Nothing complicated – chicken thighs, bread, frozen pizza, fruit cups, lactose-free milk, bagels, and a few other basics. I asked ChatGPT to compare prices across Walmart, Ralphs, Food 4 Less, and Costco using the Instacart integration I had just connected.
The response was confident. ChatGPT told me it would “pull your weekly list into Instacart and compare delivery pricing across multiple stores.” It promised to “factor in brand substitutions” and “show estimates by store with delivery fees included.” This sounded exactly like what I wanted.
The Reality Check
When I asked for the actual price comparison, things fell apart quickly. Instead of real numbers, I got a table full of vague descriptors like “lowest” and “higher” and “competitive.” When I pushed back and asked for actual pricing differences, ChatGPT finally admitted the truth: it cannot see real prices from any of these stores. It cannot calculate item-by-item price differences. It cannot actually build or modify carts.
I called it out for giving me the runaround, and to its credit, ChatGPT came clean with a more honest explanation of what the integration actually does versus what it claimed it could do.
What ChatGPT’s Shopping Integrations Actually Do
After pressing for straight answers on all three integrations, here is the reality.
The Instacart connection lets ChatGPT help you plan what to buy and suggest where items are “usually” cheapest based on general pricing patterns. It can recommend substitutions and help structure a shopping list. What it cannot do is see your local store’s actual prices, check real-time availability, or build a cart that you can then checkout.
Target and DoorDash work the same way. ChatGPT can browse catalogs conceptually and help you make decisions about what to order. It cannot see live pricing, current promotions, Circle deals, or RedCard discounts. It cannot place orders or track deliveries.
In ChatGPT’s own words: “These integrations make me a decision and planning assistant, not a transactional agent.”
That is a significant downgrade from what the marketing implies.
Meanwhile, Alexa+ Actually Does the Thing
Here is where the contrast gets stark. I use Alexa+ for my grocery shopping with Amazon Fresh and Whole Foods, and it does what ChatGPT’s integrations only pretend to do.
Alexa+ can order groceries from Amazon Fresh and Whole Foods Market, track purchase history, and remember recipes and dietary preferences. When I tell Alexa to add lactose-free milk to my Fresh cart, it adds the actual product at the actual price to an actual cart I can check out. No hypotheticals, no “typical pricing patterns” – real transactions.
A Real-World Example
Let me give you a real example of why this matters. I recently moved, and my Echo Show 15 was one of the first things I set up in the new kitchen. Because we knew the move was coming, we had intentionally let the kitchen run low on staples rather than pack and haul things we would just replace anyway. So there I was, unpacking boxes and putting away what little we had brought, realizing just how much we were missing.
For the next few hours, it was a constant stream of “Hey Alexa, add yeast to my grocery list.” Then “Hey Alexa, add butter.” Then BBQ sauce. Then olive oil. Then trash bags. Over and over while my hands were full of dishes and pantry items and packing paper. I did not have to stop what I was doing, wash my hands, find my phone, open an app, and type. I just spoke and kept working.
That is what a real integration looks like. Hands-free, frictionless, actually connected to a real cart with real prices that I could review and checkout when I was done unpacking.
And it goes beyond groceries. Partway through unpacking, I discovered the movers had broken my vacuum. “Hey Alexa, I need a new vacuum” – and suddenly I am browsing options on the Echo Show screen, reading reviews, and ordering a replacement without breaking stride. Same ecosystem, same voice, same seamless experience whether I need bread or a Bissell.
The Technical Difference
Instead of simply listing out ingredients you need by voice, you can chat with Alexa as she helps you build your shopping list. I can say something like “get everything I need for chicken stir fry” and Alexa understands the natural language request and turns it into grocery list items. Alexa can adjust quantities on the fly when asked to “make it two and a half gallons instead of a gallon” of milk.
The difference is fundamental. Alexa+ has what Amazon calls “agentic capabilities, which will enable Alexa to navigate the internet in a self-directed way to complete tasks on your behalf, behind the scenes.” It can authenticate, execute transactions, and confirm when things are done. ChatGPT’s integrations cannot do any of that.
Users can also ask Alexa+ “to remember things that will make the experience more useful” such as recipes and dietary preferences. My Alexa knows I need lactose-free products and adjusts recommendations accordingly. ChatGPT required me to specify this preference in every conversation, and even then it could not apply that knowledge to real product selection because it cannot see real products.
Try doing any of that with ChatGPT’s Instacart integration. “Hey ChatGPT, add butter to my cart” would get you a friendly explanation of how butter is typically priced at various stores and an offer to help you think through your dairy needs. Meanwhile, you are still standing there with your hands full of packing paper.
Why This Matters for IT Leaders
The disconnect between initial promises and actual capabilities is not just annoying – it is a trust issue. When an AI confidently tells you it will compare prices across four stores and build carts for you, that creates an expectation. When it turns out the AI was essentially making up capabilities it does not have, users rightfully feel misled.
This is a pattern worth watching as AI assistants add more integrations. The marketing suggests these connections unlock powerful new workflows. The reality is often much more limited – essentially a slightly smarter way to organize information you still have to verify and act on yourself.
Amazon figured this out years ago with Alexa. The difference between “helping you think about what to buy” and “actually buying things for you” is enormous. One is a planning tool. The other is a shopping assistant. ChatGPT marketed the second but delivered the first.
For those of us advising clients on AI adoption, this is an important distinction to communicate. Not all integrations are created equal. Some connect to real data and execute real actions. Others are essentially sophisticated prompt wrappers that sound impressive but add minimal functional value.
What Would Make These Integrations Actually Useful
To be fair to OpenAI, building real transactional integrations is hard. Here is what ChatGPT’s shopping integrations would need to become genuinely useful:
Real-time pricing access. The integration needs API access to actual store prices in the user’s location, not general knowledge about which stores are “usually” cheaper.
Cart manipulation. Users should be able to add items to a real cart that persists and can be checked out, not just get a conceptual list they have to manually recreate.
Transaction capability. Even if ChatGPT cannot complete purchases autonomously, it should be able to hand off a populated cart to the user for final confirmation and checkout.
Preference memory. Dietary restrictions, brand preferences, and purchase history should carry across sessions automatically, not require re-specification every time.
Until these capabilities exist, the integrations remain more demo than tool.
The Bottom Line
If you are expecting ChatGPT’s shopping integrations to automate your grocery routine or genuinely save you money through real price comparisons, temper those expectations. What you are getting is a planning assistant that can help you think through decisions, not an agent that can execute transactions or access real-time data.
For actual grocery automation, I am sticking with Alexa+ and Amazon Fresh. It knows my preferences, tracks my purchase history, and can actually complete transactions. That is what a real shopping integration looks like.
For now, if you want actual price comparisons across non-Amazon stores, you still need to open multiple browser tabs and check each store yourself. ChatGPT can help you organize your list before you do that manual work, but it cannot replace it.
I will keep testing these integrations as OpenAI improves them. But for today, the gap between promise and delivery is wide enough that I would not change my shopping workflow based on them.
]]>Look, I am a busy woman, mom, CIO, etc. My days and nights fill up quickly between work, kiddo, hobbies, etc. So when OpenAI recently rolled out ChatGPT shopping integrations with Instacart, Target, and DoorDash, I couldn’t wait to try it. The pitch sounds compelling, compare prices across stores, build grocery carts, streamline your weekly shopping routine. I connected all three to put them through their paces. What I found was a masterclass in AI overpromising and underdelivering, especially when compared to what I already use for grocery shopping.
Testing the Instacart Integration
I started with a simple test. I gave ChatGPT my standard weekly grocery list with the prices I normally pay at Ralphs for pickup. Nothing complicated: chicken thighs, bread, frozen pizza, fruit cups, lactose-free milk, bagels, and a few other basics. I asked ChatGPT to compare prices across Walmart, Ralphs, Food 4 Less, and Costco using the Instacart integration I had just connected.
The response was confident. ChatGPT told me it would “pull your weekly list into Instacart and compare delivery pricing across multiple stores.” It promised to “factor in brand substitutions” and “show estimates by store with delivery fees included.” This sounded exactly like what I wanted.
The Reality Check
When I asked for the actual price comparison, things fell apart quickly. Instead of real numbers, I got a table full of vague descriptors like “lowest” and “higher” and “competitive.” When I pushed back and asked for actual pricing differences, ChatGPT finally admitted the truth: it cannot see real prices from any of these stores. It cannot calculate item-by-item price differences. It cannot actually build or modify carts.
I called it out for giving me the runaround, and to its credit, ChatGPT came clean with a more honest explanation of what the integration actually does versus what it claimed it could do.
What ChatGPT’s Shopping Integrations Actually Do
After pressing for straight answers on all three integrations, here is the reality.
The Instacart connection lets ChatGPT help you plan what to buy and suggest where items are “usually” cheapest based on general pricing patterns. It can recommend substitutions and help structure a shopping list. What it cannot do is see your local store’s actual prices, check real-time availability, or build a cart that you can then checkout.
Target and DoorDash work the same way. ChatGPT can browse catalogs conceptually and help you make decisions about what to order. It cannot see live pricing, current promotions, Circle deals, or RedCard discounts. It cannot place orders or track deliveries.
In ChatGPT’s own words: “These integrations make me a decision and planning assistant, not a transactional agent.”
That is a significant downgrade from what the marketing implies.
Meanwhile, Alexa+ Actually Does the Thing
Here is where the contrast gets stark. I use Alexa+ for my grocery shopping with Amazon Fresh and Whole Foods, and it does what ChatGPT’s integrations only pretend to do.
Alexa+ can order groceries from Amazon Fresh and Whole Foods Market, track purchase history, and remember recipes and dietary preferences. When I tell Alexa to add lactose-free milk to my Fresh cart, it adds the actual product at the actual price to an actual cart I can check out. No hypotheticals, no “typical pricing patterns”, real transactions.
A Real-World Example
Let me give you a real example of why this matters. I recently moved, and my Echo Show 15 was one of the first things I set up in the new kitchen. Because we knew the move was coming, we had intentionally let the kitchen run low on staples rather than pack and haul things we would just replace anyway. So there I was, unpacking boxes and putting away what little we had brought, realizing just how much we were missing.
For the next few hours, it was a constant stream of “Hey Alexa, add yeast to my grocery list.” Then “Hey Alexa, add butter.” Then BBQ sauce. Then olive oil. Then trash bags. Over and over while my hands were full of dishes and pantry items and packing paper. I did not have to stop what I was doing, wash my hands, find my phone, open an app, and type. I just spoke and kept working.
That is what a real integration looks like. Hands-free, frictionless, actually connected to a real cart with real prices that I could review and checkout when I was done unpacking.
And it goes beyond groceries. Partway through unpacking, I discovered the movers had broken my vacuum. “Hey Alexa, I need a new vacuum”, and suddenly I am browsing options on the Echo Show screen, reading reviews, and ordering a replacement without breaking stride. Same ecosystem, same voice, same seamless experience whether I need bread or a Bissell.
The Technical Difference
Instead of simply listing out ingredients you need by voice, Amazon showed how you could also chat with Alexa as she helps you build your shopping list. I can say something like “get everything I need for chicken stir fry” and Alexa understands the natural language request and turns it into grocery list items. Alexa was able to adjust quantities on the fly when asked to “make it two and a half gallons instead of a gallon” of milk. (TechCrunch)
The difference is fundamental. Alexa+ has what Amazon calls “agentic capabilities, which will enable Alexa to navigate the internet in a self-directed way to complete tasks on your behalf, behind the scenes.” (Amazon) It can authenticate, execute transactions, and confirm when things are done. ChatGPT’s integrations cannot do any of that. (PS, you can also turn off the automatic checkout in Alexa or even make it where it will only check out for those who’s voices it recognizes! Hugely helpful for those with little toddlers who think Alexa is their besties)
Users can also ask Alexa+ “to remember things that will make the experience more useful” such as recipes and dietary preferences. My Alexa knows I need lactose-free products and adjusts recommendations accordingly. ChatGPT required me to specify this preference in every conversation, and even then, it could not apply that knowledge to real product selection because it cannot see real products.
Try doing any of that with ChatGPT’s Instacart integration. “Hey ChatGPT, add butter to my cart” would get you a friendly explanation of how butter is typically priced at various stores and an offer to help you think through your dairy needs. Meanwhile, you are still standing there with your hands full of packing paper.
Why This Matters for IT Leaders
The disconnect between initial promises and actual capabilities is not just annoying… it is a trust issue. When an AI confidently tells you it will compare prices across four stores and build carts for you, that creates an expectation. When it turns out the AI was essentially making up capabilities it does not have, users rightfully feel misled.
This is a pattern worth watching as AI assistants add more integrations. The marketing suggests these connections unlock powerful new workflows. The reality is often much more limited. Essentially a slightly smarter way to organize information you still have to verify and act on yourself.
Amazon figured this out years ago with Alexa. The difference between “helping you think about what to buy” and “actually buying things for you” is enormous. One is a planning tool. The other is a shopping assistant. ChatGPT marketed the second but delivered the first.
For those of us advising clients on AI adoption, this is an important distinction to communicate. Not all integrations are created equal. Some connect to real data and execute real actions. Others are essentially sophisticated prompt wrappers that sound impressive but add minimal functional value.
What Would Make These Integrations Actually Useful
To be fair to OpenAI, building real transactional integrations is hard. Here is what ChatGPT’s shopping integrations would need to become genuinely useful:
Real-time pricing access. The integration needs API access to actual store prices in the user’s location, not general knowledge about which stores are “usually” cheaper.
Cart manipulation. Users should be able to add items to a real cart that persists and can be checked out, not just get a conceptual list they have to manually recreate.
Transaction capability. Even if ChatGPT cannot complete purchases autonomously, it should be able to hand off a populated cart to the user for final confirmation and checkout.
Preference memory. Dietary restrictions, brand preferences, and purchase history should carry across sessions automatically, not require re-specification every time.
Until these capabilities exist, the integrations remain more demo than tool.
The Bottom Line
If you are expecting ChatGPT’s shopping integrations to automate your grocery routine or genuinely save you money through real price comparisons, temper those expectations. What you are getting is a planning assistant that can help you think through decisions, not an agent that can execute transactions or access real-time data.
For actual grocery automation, I am sticking with Alexa+ and Amazon Fresh. It knows my preferences, tracks my purchase history, and can actually complete transactions. That is what a real shopping integration looks like.
For now, if you want actual price comparisons across non-Amazon stores, you still need to open multiple browser tabs and check each store yourself. ChatGPT can help you organize your list before you do that manual work, but it cannot replace it.
I will keep testing these integrations as OpenAI improves them. But for today, the gap between promise and delivery is wide enough that I would not change my shopping workflow based on them.