Part II/III: GEO for Retailers, Brands and Agencies - Optimize the Six Agentic Shopping Engines for Holiday '25 to Drive Maximum Sales

Part II: The 13 Agentic Commerce Pitfalls with detailed real-world examples across various Answer Engines.

Welcome to part two of a three part series:

• Part I - (9/3/25) - What is GEO for Agentic Commerce and What’s the Goal? (here).

• Part II - (9/10/25) - The 13 Agentic Commerce Pitfalls with real-world examples **You are Here**

• Part III - (9/30/25) - The Agentic Commerce Playbook - Easily Implementable Strategies and Tactics to avoid the 13 pitfalls, and optimize the Seven Agentic Shopping Engines to drive maximum sales in Holiday ‘25

Note: In this series we will be showing examples of different brands and retailers in various contexts - both for what’s working well and what is not working well. These have been picked at random as illustrative at the time the screen shot was taken (9/2/25-9/30/25 window of time). This is a moving target and your experience maybe different. It may feel like we are ‘picking on’ a particular retailer or Agentic Shopping Engine, it’s just what we found first and totally randomized. These examples cut across seven engines and thousands of online merchants.

Let’s Connect at RetailClub!

I hope to see a lot of you at RetailClub, come on by our Agentic Oasis and don’t forget to sign up for my session, it has very limited seating (Tuesday at 3:35pm)!

13 Agentic Commerce Pitfalls

In Part I, we introduced GEO and talked about the goal - to ‘Own the Product Card’.

We also introduced a three component framework for evaluating how a set of your top-selling SKUs are performing against the goal on the Agentic Shopping Engines:

Think of the 13 Pitfalls as the most common problems, usually related to the Agentic Shopping engine’s web crawler, product catalog or agent incorrectly pulling data from your product detail page. As we go through these with examples, we’ll categorize them into the three categories of visibility, data quality and AI digital shelf are how we’re going to walk through the most common problems pitfalls we have seen on Agentic Commerce. For each one, when possible we’ll give an example.

Pitfall 1: Visibility: Not showing up at all

Unfortunately this is the most common. We’re at the tail end of back-to-school, but in homage to it let’s start there. Kohl’s has the ‘Jansport cool Student Backpack at $60 in a variety of colors and all in stock. Maybe you are a Kohl’s enthusiast and are swimming in Kohl’s Cash and their other loyalty programs, so you could save $15 on this→

And look when you add to cart there’s a $5 store pick-up Kohl cash bonus→

But you don’t start there, you go to ChatGPT-5:

You see that $60 seems to be the standard price unless you want to go the eBay route. Therefore you throw it in a cart for Walmart or Target depending on your preference there.

Kohl’s didn’t even get a look here, didn’t even get a shot at a zero sum game and thus missed out on the sale. As a Kohl’s regular, if I had seen it there, perhaps I would have explored the Kohl’s cash deals. Maybe I’m going there for Jeans anyway, but they lost out on a $60 transaction loaded with Kohl’s cash.

Why? That’s what we’ll go into in Part III - stay tuned!

Pitfall 2: Data Quality: Wrong SKU mapped to Card (Canonicalization)

Let’s say we have a family of 5, we’re going camping and some of our kids are pretty dang tall so we’re not squeezing into 4 person any more. We need a 6 person tent. We go to ChatGPT and find this amazing North Face Wawona 6 tent - the 6 is for number of people that it holds.

Holy cow, tents are expensive $550, but this is the Taj Mahal of tents and everyone likes glamping > camping so we’ll splurge….wait, look there’s one for $400 - awesome. I don’t know Jesse Brown, but heck for $150 I’ll do some research…

If I was in a hurry or this was agentic, I’d end up thinking I was buying the 6 person and getting a 4 person. My glamping vision just turned into an 8-hr drive explaining to my wife how I ended up with a 4 person tent to save $100 by ‘accident’.

This we call mis-canonicalization. We’ll explain in more in post III. Your reaction maybe -well this is ChatGPT’s fault. Sure you could say that, but I’d say it’s Jesse Brown’s fault.

This is bad for them for 2 reasons:

• They may get higher returns here

• If they have the 6 person, they missed out on a possible sale here.

Pitfall 3: Data Quality: Wrong Title/Title very short under utilizing product card title real estate

I’m looking for a new mountain bike, I have gone through the Research and into the Find part of my shopping journey:

Anatomy of a Product Card

Notice the product cards have several elements, let’s get nerdy on them:

• Image (ex: the bike you see above)

• Image overlay text (Product Card title) (ex” “Sale at Mike’s Bikes (~$949)”)

• Product Title (es: “Cannondale Habit HT3” - in bold)

• Price (ex: $948.95)

• Merchant: (ex: “Mike’s Bikes + others”)

• Product Card owner: Mike’s bikes owns this product card (congrats Mike!)

Back to Pitfall 3

Ok, now that we have the vocab for hit, here’s the pitfall, The card on the right.

Notice that the image overlay text says ‘standard price’

The Product Title is 8 characters and doesn’t mention the brand: “Habit HT 3”

Brooks lost to Mike for several reasons, one of them is Pitfall 3 - they have been gifted for free extremely-valuable digital real-estate here on the ChatGPT-5 product card and they failed to optimize it.

Pitfall 4: Data Quality: Variation Problems

Variations are a great example in e-commerce where every time we improve the buyer experience from a 8 to a 1 (1 is best), we make the back-end product complexity go from a 1 (simplest) to a 10 (hardest). I’ve probably spent a cumulative 2yrs of my life working, worrying about, messing up and fixing variations.

We’re back to our back-to-school example and my daughter is really excited to get a new backpack, but she’s very…..selective… let’s say and the ONLY color she can possibly live with is Lavendar Ash - she saw it on an influencer, it’s perfect with her FDOC dress and “ ‘hear me out’ if I don’t get it the entire year is ruined!!! 😢😢😢😢😢😢” (see I get it from both sides with variations!)

I got this, I fire up Perplexity:

You can see there’s two variations: size - which is a faux variation as there’s only one size (go figure), and then under ‘Option’ is a color picker! Yes we are one click from Lavender ash, saving the school year and winning the best Dad….

I’ve scanned this 100x and guess what, no Lavender Ash NOOOO!!

But wait - if you’ll notice Perplexity has found this base model at Office Depot. Let’s go there and just double check:

Yep there it is. But if I didn’t take that extra step, I would have missed it, the school year would have been RUINED. You may argue that this is Perplexity’s fault, but what if it’s Office Depot’s fault? They would have missed out on a sale.

Pitfall 5: Data Quality: Language/Currency confusion

Let’s say I have a 14 year old Niece with a birthday coming up, Microsoft Copilot helps me by teaching me that fun travel size teen-friendly beauty sets are the way to go (who knew? What happened to Barbies!? I digress…). I quickly narrowed in on this Sol De Janero Jet Set Travel Kit - fun!

Wait, what’s McGrocer and why is it listed in pounds? As you shop on answer engines, you’ll see a lot of of Euros, Pounds, and the good old Canadian dollar popping up. What’s up? The Agentic Shopping Engine knows where I am, why is it showing me this?

Let’s go look at McGrocer and see if we learn anything:

Wait, what is going on here -that’s USD not British Pounds. Part III - stay tuned - we’ll have a proper cup of tea 🫖

Pitfall 6: Data Quality: Missing/incorrect Attributes

Back in Pitfall 3 with the awesome Cannondale HT 3 bicycle, we showed how some brands and retailers aren’t utilizing the full space available in cards. The same bike, now on Perplexity is perfect for showing Pitfall 6.

Here Pereplexity’s product cards have these little ‘highlights’ call outs. Some of them are related to positives and negatives from reviews, but you’ll attributes in there too. Or the Answer engine will pull the attributes from the PDP and use those to learn more about how people in reviews feel about the ‘components’, forks, brakes, weight, suspension, dropper post, etc.

This is another missed opportunity for Brooks - the information Perplexity has isn’t enough for them to fill that out.

This is currently more common than ‘wrong’ attributes, but we do see the engines adding more and more data up at the product card level in the near term which will increase the likelihood of incorrect attributes being picked up.

Pitfall 7: Data Quality:Product Reviews Missing/Wrong

Here’s a great brand Kendra Scott. Their customers love the brand and cleverly, Kendra Scott has reviews on their PDP. This bestseller has 11,267 reviews. LLMs love context - you can’t ask for more than that!

Surely ChatGPT has pulled this in and referred to it….

Nope. Welcome to Pitfall 7, Elisa.

Pitfall 8: Data Quality: Wrong, missing images

Let’s say I’m in the market for a fancy e-watch from Garmin - the fenix 8 AMOLED - I want a really bright screen and this AMOLED delivers based on all I’ve researched.

I fire up Perplexity and here’s the hero image for the Sapphire color I want:

I’m no expert merchandiser, but there seems to be a disconnect here 🤔. You can see in the description: ‘bright AMOLED display’ and I’m getting a hero image with….no display? Down at the Balardi site we have the much better:

That’s better. What went wrong here? Part III friends.

Pitfall 9: Data Quality:Product out of stock

I need some new Merrell Moab 3’s and I like the exciting vibrant Olive/Gum color:

You can see I have explicitly asked for the men’s size 11.5 in this prompt. When I go to BigRay’s→

Womp Womp - it’s out of stock as denoted by the strike through on the 11.5 size. Darn you Pitfall 9 - foiled again!

Pitfall 10: Data Quality: Incorrect PDP/product linked to for checkout

Unfortunately due to the ethical issue on this one that I’ll explain, I am unable to show examples, but you’re going to have to trust me that we do find a lot of these and we always prioritize them as the first thing we fix. The most common example is there is a product that has two sizes: travel (3oz) and standard (30oz). The travel size is $4 and the standard size is $36.

If the wires get crossed you have two situations:

• Buy the standard size (30 oz) for $36, get the travel size (3 oz) - Bad for customer who paid $32 too much. This is initially great for the retailer until the customer is angry , churns for life, then it’s lose/lose.

• Buy the travel size (3oz) for $4, get the standard (30oz) for $4 - Here the customer is thrilled - they paid $4 and got the $36 product! This is obviously bad for the retailer and they are definitely up-side-down on this product.

Now many times it’s the Agentic layer that is eating the problem here, but in any chase it would be unethical to highlight one of these because it’s obviously an error and we don’t want these to escape into the wild.

You’ll have to trust me that there are a lot of these.

Pitfall 11: Data Quality: Incorrect Price

My GoPro is broken and it’s time for a new one. I fire up CoPilot and quickly narrow it down to the Hero13 with the Ultra-Wide for some really amazing panoramic shots:

I click in and BestBuy owns the Product Card, but if you look closely….

Wait - it says $400, but also says $379.99? Which is it?

It’s clearly $379.99, where did that $400 come from?! (Part III, Part III!)

In this example, it's a ‘light’ pricing error, there are plenty where the $400 takes over, the product is passed over and the retailer loses a shot, or it can go the other way where you see a crazy low price at the Agentic Shopping Engine level and it’s higher when you land.

Pitfall 12: Not Owning the Product Card

Let’s say you roll up your sleeves and do all the work to avoid, detect, fix all the Pitfall 1-11 examples. As frustrating as it is, you can still fail to Own the Product Card. A frequent example you see here is the Brand DTC vs. Retailers. Let’s say you are a rabid Taylor Swift fan and want to celebrate the Big Engagement and the big October 3 album release, and you already have 10 Travis Kelce Jerseys, so you’re going to add a Jason Kelce Jersey to your collection to round it out:

I’m a big TS fan, but I’ll take the $130 option over the $325→

Here we have Nike which is the DTC competitor and then the Eagles Shop, the Fanatics site and the NFL Shop are all “Fanatics experiences” - so they get have 75% of the slots here - but note the different shipping and return policies enhance the feeling they are ‘different’ - very clever move by Fanatics - they are crowding out the AI Digital Shelf. But despite all that, they are getting beat by Nike that has $8 delivery vs. Fanatics at $6 and they have the tightest return policy at 60-days with NFL shop at 90-day and Eagles/Fanatics at 365.

I also checked and Dicks has this SKU but hit Pitfall 1 on this one - they have free shipping, great content/reviews and a 90-day return policy. They could totally be in this game, but fell smack into Pitfall 1.

Pitfall 13: Winning a Product Card, losing the Competitive Mosaic

In Pitfalls 1-12, we’ve focused at a SKU level because it’s easiest to understand. But most brands and retailers have hundreds, thousands, tens of thousands and in the world of Amazon vs. Walmart, millions and hundreds of millions of SKUs.

The US retail landscape is full of two or three large retailers battling it out all the time:

• Sephora vs. Ulta

• Target vs. Walmart vs. Amazon

• Dick’s Sporting Goods vs. Academy

• REI vs. Cabellas vs. Backcountry

• Lowes vs. Home Depot

• Macys vs. Lowes

• Best Buy vs. Circuit City (ok they won there)

• Costco vs. Sams vs. BJs

• Walgreens vs. CVS

• Kroger vs. Albertsons vs. Publix

• Dollar General vs. Dollar Tree

and so on…

My point is as your Agentic Shopping maturity increases, retailer and brands will look at a broader mosaic of their SKUs that we call the AI digital shelf.

Let’s pick on REI and Backcountry just as an example.

They each have about 50k SKUs. You can imagine the Ven diagram between those breaks down into these buckets:

• REI exclusive SKUs

• Backcountry exclusive SKUs

• Overlapping SKUs that REI wins the AI Digital Shelf

• Overlapping SKUs that Backcountry wins the AI Digital Shelf

• Overlapping SKUs where neither owns the product card

That’s the mosaic of the AI digital shelf. If I’m REI in this example, I’m going to want to focus in on ‘overlap backcountry wins’ and ‘overlap someone else wins’. Let’s say that’s 6,000 SKUs - 3k in each bucket. That’s a lot, but if I can now prioritize them by sales rank and category, there’s probably 600 products in there that ‘matter most’.

In our experience, even though it’s early, you can definitely significantly move the needle. The pitfall here is failure to see the forest for the Trees - pitfalls 1-12 are Trees and pitfall 13 is here to remind you that the Forest is important and don’t forget it.

Part III Coming Up!

September 30th we’ll wrap up this series with Part III where we’ll provide a complete Agentic Commerce Playbook. This is a set of easy-to-implement strategies and tactics that will help you get out of these 13 pitfalls. If you jump on this project in October/November, you can be ready for what we believe is going to be a robust Holiday 25 Agentic Shopping season and lay the groundwork for 2026 when things really go crazy 🚀