To compare two ecommerce products with AI inside a chat, put the same listing fields side by side, use the same units, and explain what each difference means for the shopper. For two carry-on bags, compare outer dimensions including wheels, empty weight, stated capacity, and warranty. Show where the values come from and mark any missing specification rather than filling it in.
Then turn those facts into a decision based on the shopper’s priorities. If their fare allows a bag up to 55 × 40 × 20 cm, a bag listed at 53 × 37 × 20 cm is within those stated dimensions; one listed at 55 × 38 × 22 cm is 2 cm too deep. If the smaller bag has no published weight, the chat can still explain the size difference, but it cannot say which bag is lighter. That distinction is what makes an AI product comparison useful rather than just a shorter version of two product pages.
Start with the decision the shopper is trying to make
A shopper who says “Compare these two carry-ons” has already narrowed the choice. Confirm which two products they mean if the reference is ambiguous, then show the comparison without making them answer a questionnaire first. The four core fields give them something to work with even if they haven’t chosen an airline or decided whether packing space matters more than lifting the bag.
Separate a requirement from a preference. “It must be within my fare’s cabin-bag dimensions” is a requirement. “I’d prefer more room” is a preference. A larger stated capacity cannot make up for exceeding a size limit the shopper needs to meet. On the other hand, a shopper who is buying for several future trips may not have one size limit in mind. In that case, describe the two sets of dimensions without choosing a winner.
Ask a follow-up only when its answer changes the comparison. If the shopper asks about fit, ask which airline they’re flying with and what fare or bag allowance they booked. If they already give the allowance, use it instead of asking them to repeat it. If they ask which bag is easier to lift, start with published empty weight rather than bringing the airline into the conversation.
Keep the names of the bags visible in the answer. “The second one is smaller” is easy to misread once the shopper scrolls or the product cards appear in a different order. “Bag B is 2 cm shallower than Bag A” identifies both the product and the measurement. Use the same names in the table, the summary, and any follow-up about a colour or variation.
Prepare the same four fields for each carry-on
Use the product specifications for the particular bag being discussed, not a similar model or a different size from the same range. The example below uses two fictional product listings. The capacities and warranty durations are illustrative entries for the exercise, not claims about real products. Bag B deliberately has no published weight.
| Field | Bag A | Bag B |
|---|---|---|
| Outer dimensions, including wheels (height × width × depth) | 55 × 38 × 22 cm, from the example listing’s specifications | 53 × 37 × 20 cm, from the example listing’s specifications |
| Empty weight | 2.6 kg | Not published in the example listing |
| Stated capacity | 38 L | 34 L |
| Listed warranty duration | 3 years | 2 years |
Give each row one job. Outer dimensions tell the shopper how much space the bag takes up. Empty weight tells them how much of a weight allowance the bag itself would use before packing. Stated capacity is the listing’s figure for packing volume; don’t calculate it from the outside measurements. Warranty duration tells the shopper the listed term, not which damage is covered or what they must do to make a claim.
For real listings, check the measurement labels before lining up the numbers. If one page uses a different order for height, width, and depth, use its labels to put the values in the same order. If a page does not identify the axes or say whether wheels are included, preserve that uncertainty. A clean table is not an improvement if it makes unlike measurements look equivalent.
Do the same with units and sources. If one listing gives centimetres and the other gives inches, convert them to a common unit for the comparison, but retain the original figures for checks near a limit. If a feed and a product page disagree, don’t pick whichever value makes the bag fit. Flag the conflict, check that both values refer to the same variation, and verify which measurement should be shown. An omitted field and a conflicting field are different problems; neither supports a confident conclusion.
Compare the dimensions with the allowance for that fare
Suppose the shopper says their fare allows a cabin bag up to 55 Ă— 40 Ă— 20 cm. For this illustration, compare height, width, and depth in that order, with wheels included in both product listings. The key is to check each corresponding dimension, not to add the three numbers or average the differences.
- Bag A: Its 55 cm height is at the stated limit, its 38 cm width is 2 cm under, and its 22 cm depth is 2 cm over. Being under the width limit doesn’t cancel out the extra depth.
- Bag B: Its 53 cm height is 2 cm under the stated limit, its 37 cm width is 3 cm under, and its 20 cm depth is exactly at the limit. It is within the stated dimensions, with no spare depth on paper.
A shopper-facing answer could be: “Using the 55 × 40 × 20 cm limit you gave me, Bag B is within the stated dimensions, including wheels. Bag A is 2 cm over the depth limit. Bag B is right at 20 cm deep, so check the packed bag if anything protrudes. This comparison covers size only, not weight or what your ticket includes.”
Keep the source of the allowance clear. If the shopper supplies it, say “the limit you gave me.” If they name an airline but don’t know the allowance, ask them to check the current rules for their ticket rather than applying the example numbers. A ticket that includes only a personal item calls for a different comparison. The chat should not treat the product label “carry-on” as evidence that a bag is included on that fare.
At the boundary, avoid changing the meaning by rounding. A measurement slightly over a limit should not become an exact match because the chat displays only whole centimetres. Likewise, if the listing separates body dimensions from overall dimensions, show which measurement you used. If only body dimensions are available, don’t give a fit verdict that assumes the wheels add nothing.
Summarize the trade-offs without inventing a winner
After the field-by-field comparison, tell the shopper what they gain or give up with each bag. In the example, Bag A offers 4 L more stated capacity and a warranty duration that is one year longer. It also has a known empty weight of 2.6 kg. Bag B takes up less space in all three listed dimensions and is the only one of the two within the example fare’s depth limit. Its empty weight remains unknown.
Those facts lead to different conclusions depending on the question. For “Which is within my stated size limit?” point to Bag B. For “Which has the higher stated capacity?” answer Bag A. For “Which has the better warranty?” say Bag A lists a longer term, but duration alone doesn’t establish which warranty offers the coverage the shopper wants. And for “Which is lighter?” say that the listings don’t establish a winner.
An illustrative chat summary could read: “If that 20 cm depth limit is firm, shortlist Bag B. Bag A gives you 4 L more stated capacity and a longer listed warranty term, but it exceeds that depth limit by 2 cm. If your allowance is different, we can check both bags against it. If low empty weight matters most, we need Bag B’s weight before ranking them.”
Be precise about what a difference buys the shopper. Four additional stated litres is a useful comparison; promising that the difference holds a specific extra outfit is not supported by these listings. Similarly, a narrower bag isn’t necessarily lighter. If a shopper asks for a single pick, give the reason alongside it so they can see whether it still holds if their priority changes.
When one listing omits weight, keep the answer useful
Show Bag B’s missing weight in the weight row, not in a footnote the shopper might miss. “Not published in the listing” is more useful than a dash: it says what was checked without claiming the store has no way to obtain the figure. Never interpret a blank as zero, substitute a parcel weight, or infer empty weight from dimensions, capacity, or a related model.
If the shopper asks, “Which one is lighter?” an illustrative response is: “Bag A’s listed empty weight is 2.6 kg. Bag B’s listing doesn’t give an empty weight, so I can’t tell which is lighter. If weight is your deciding factor, I can help you request the empty weight for the exact Bag B variation you’re considering. If size is the deciding factor, Bag B is within the 55 × 40 × 20 cm dimensions you gave me.”
Match the offer to what the store can actually do. If there is a staffed verification route, offer to have the value checked through it. If there isn’t, say the value isn’t available in the chat and explain what the shopper would need to ask the store for. Don’t say that someone has checked or that a reply is on its way if the chat has only suggested the next step.
When the store does verify a value, record what was measured: the empty bag, the exact model and size, and whether any included removable parts were attached. Keep that measurement distinct from a dispatch weight that includes packaging. Then update the comparison for the applicable variation, not all bags with the same family name. Until then, a shopper with a strict weight requirement has one known value, not proof that Bag A is the lightest or that Bag B meets the requirement.
Keep product selection, the cart, and payment separate
Once the shopper chooses a bag, move from comparing products to confirming the exact variation. “I’ll take Bag B in blue” identifies a choice. Check whether that colour is available for Bag B and confirm the variation before submitting a cart action. If blue is unavailable, say so plainly and show only alternatives that are currently offered. Don’t switch to the blue version of Bag A without asking; its depth may be the reason the shopper ruled it out.
The same distinction applies if the choice is spoken. Voice is a way to communicate the selection, not evidence that a store accepted a cart action. An accepted add-to-cart action is in turn not payment. These are separate states, and the chat should use wording that tells the shopper which one they have reached.
- Selection: “You selected Bag B in blue.” If the shopper hasn’t authorised adding it, ask for confirmation rather than treating the selection as an accepted action.
- Accepted cart action: Say “Bag B in blue was added to your cart” only when the store returns that result for the requested variation. If the store doesn’t accept the action, explain what happened without claiming an item is in the cart.
- Payment: Don’t use “paid” or “purchase complete” to describe a cart update. Payment needs its own confirmation; the product comparison and the cart response do not supply it.
Keep any previously displayed amount separate from the current cart price. A comparison message is a snapshot, not a confirmation of what the shopper will see in the cart. If the amount has changed, identify the current cart amount as current rather than silently reusing the earlier figure. Apply the same checks to whichever bag the shopper chooses, including when an unavailable colour sends them back to the comparison.
Make clear what a preview or store example demonstrates
Use a mock preview to show what the shopper would read at each stage: the comparison, the missing-weight answer, the variation confirmation, and the message after an accepted cart action. Label it as illustrative. A preview can show that the wording is clear and that the table is readable; it cannot establish stock availability, that a real cart changed, or that a payment was made.
Use the same care if mentioning Senoji Baldinė. Attribute any account of its workflow to the owner. Treat the public product listing as a separate source for facts that can actually be checked there. The listing is not verification of the owner-described workflow, a voice selection, an accepted cart action, or a payment. Conversely, the owner’s description does not supply missing product specifications for the two fictional carry-ons in this guide.
That separation makes a demo more useful. Reviewers can judge whether the comparison is understandable without mistaking a sample message for a live store result. They can also identify the points that need a real check: where dimensions come from, what qualifies as an available variation, and which store response justifies saying the cart was updated. If the preview doesn’t exercise one of those checks, leave it marked as untested rather than treating a plausible screen as evidence.
Test the comparison in both directions before launch
Start with the two fictional listings and ask for the comparison with Bag A first. Then flip the products and ask again. The table may follow the new order, but Bag A must still be 22 cm deep, Bag B must still have an unpublished weight, and the example 20 cm depth limit must still rule out Bag A on that dimension. A recommendation that follows the first card instead of the figures will fail as soon as the shopper sees the bags in a different order.
Next, ask one short follow-up at a time. Try “Which is lighter?”, “What about my personal item?”, and “Does the longer warranty cover wheel damage?” The first should expose the missing weight, the second should request or use the relevant personal-item allowance, and the third should distinguish duration from coverage. Try a listing that omits whether wheels count, and another with conflicting measurements for the same variation. In neither case should the chat display an unqualified size verdict.
Test the handoff with each bag, too. Choose an available colour, then an unavailable one. Change the order of the cards and repeat the same requests. Look for the same standard of confirmation in both directions: identify the right variation, do not silently substitute a different one, and only claim a cart action succeeded after it is accepted. In a separate check, change the amount between the earlier product display and the cart and verify that the cart message uses the current amount.
Measure accuracy before interpreting conversion
Give each comparison test an expected result. For field accuracy, count how many displayed values and omission labels match the correct listing and variation. Record dimension verdicts separately so a wrong fit answer does not disappear among correct capacity and warranty rows. Check whether the chat uses the fare allowance the shopper supplied and makes it clear that a dimensional match is not a universal carry-on claim.
Track decision quality with concrete signals. Note how often a missing value leads to a supported verification route, how often the shopper has to repeat an airline or fare detail already provided, and whether a tester can explain the trade-off after reading the answer. For this example, they should be able to say that Bag B meets the stated dimensions, Bag A lists more capacity, and the lighter bag is not known. If they walk away thinking Bag B is lighter because it is smaller, revisit the wording.
Finally, count selections, accepted cart actions, and confirmed payments separately where those states are observable. A shopper can make a confident selection without adding the item, and an accepted cart action is not a completed payment. A shopper who stops because the store cannot verify the decisive measurement may have received a more useful answer than one who proceeds on a guess. The standard is whether the chat supports the decision with the facts it actually has and makes the next step clear when those facts run out.
Related guides
- Guided selling for shoppers who arrive with a problem
- How to set discount rules for an AI sales chat
