The short answer
A useful comparison explains which option fits which buyer, using the same questions and evidence standards for every product. Identify the exact models, date your sources, distinguish specifications from tests and say what remains unknown. Your product does not need to win every row for the page to help someone make a confident choice.
A comparison page often starts with the wrong assignment: prove that our product is best. That leads to selective criteria, unsupported checkmarks and conclusions that do not survive a shopper opening another tab. A stronger assignment is narrower: help a particular buyer decide between plausible alternatives.
This is an editorial framework for Shopify and DTC teams, not a proprietary testing methodology or a claim about client results. It extends our Shopify AI Search/GEO guide into a practical comparison-page template. The worked example is entirely fictional.
1. Write the buyer decision before choosing the products
“Which commuter backpack fits my laptop and is comfortable to carry on a train?” gives you a useful starting point. “Why our backpack beats everyone else” does not. Define the shopper, intended use, market and constraints before selecting alternatives. Record any non-negotiable requirement separately from preferences.
A sleeve that cannot hold the shopper’s device fails a fit requirement. A bag that weighs slightly more might still be acceptable if it offers the access or storage the shopper values. Do not bury a failed requirement inside an average score.
Select alternatives a shopper could reasonably consider for that decision. Compare current models available in the relevant market. If you include a cheaper, larger or differently designed option, explain why it belongs. Avoid choosing an outdated or unsuitable rival merely because it makes your own product look stronger.
2. Build a claim record before building the table
For each product, record its model, variant, source URL, date checked and the exact fact the source supports. Keep a distinction between these evidence types:
Attribute the claim
A manufacturer’s listed weight is a published specification. Label it accordingly; it is not your own measurement.
Describe the method
If you measured or tested a product, identify the sample, conditions and limits. Keep raw evidence available for review.
Make the reasoning visible
Explain why a feature matters for this use case. Separate that interpretation from the underlying fact.
Leave the uncertainty intact
“Not published in the source checked” is different from “does not have it.” Missing evidence is not proof of absence.
For changing details such as price, stock or return conditions, identify the country, currency, variant and date. Do not compare a temporary promotion with another product’s regular price without making the difference visible. If you cannot maintain a price table, link to current pricing and compare the more durable decision criteria.
Use public product information to check competitor facts. If sources conflict, investigate the exact variant and revision before publishing a verdict. Do not turn an AI-generated summary into a primary evidence source. For gaps in your own records, use our product-page facts checklist.
3. Apply the same criteria to every option
Choose criteria because they change the buying decision. For commuter bags, usable sleeve dimensions, empty weight, opening style and weather limitations might be useful. A proprietary feature name may mean little unless you explain what it does and whether another design serves the same need.
Use matching units and definitions. External bag dimensions cannot stand in for usable laptop-sleeve space. A stated water-resistant fabric is not equivalent to a completed waterproof test. Compare a published specification with another published specification, or clearly disclose the difference in evidence quality.
Google’s review guidance encourages decision-relevant measurements, original evidence, alternatives, benefits and drawbacks. It also asks for supporting first-hand evidence when recommending a product as best for a purpose. Use that as a quality check, not as a promise that a comparison will rank or earn an AI citation. Source: Google’s review guidance.
Avoid an unexplained overall score. If you use weights, publish them and explain whose preferences they represent. Show what happens when the buyer’s priority changes. Often a short “choose this if” conclusion is more useful than a precise-looking number built from subjective inputs.
4. A fictional comparison that can change its recommendation
Illustrative example only. Harbor 18 and Metro 20, all specifications and the buyer scenario below are invented. They are not real products, competitor claims, client data or results of a product test.
Imagine a commuter with a laptop and case measuring 32 × 22 × 1.8 cm. The first requirement is enough usable sleeve space. The buyer then cares about weight and access on a crowded train.
| Decision criterion | Harbor 18 | Metro 20 | What it means |
|---|---|---|---|
| Usable sleeve dimensions | 33 × 23 × 2 cm | 35 × 25 × 3 cm | Both exceed the example device dimensions; check opening and fit before purchase. |
| Empty weight | 720 g | 910 g | Harbor is 190 g lighter in this invented specification set. |
| Main opening | Top opening | Clamshell opening | Preference depends on packing and access needs; neither is universally better. |
| Weather information | Water-resistant fabric; unsealed seams | No weather claim specified | Neither has evidence here supporting a waterproof recommendation. |
For this buyer, Harbor could be the more suitable option if low empty weight is the main preference after fit. Metro could be more suitable if the buyer values the larger sleeve and clamshell packing access. The table cannot establish comfort: dimensions and weight alone do not show how a loaded bag feels on someone’s body.
Now change the laptop-and-case dimensions to 34 × 24 × 2.5 cm. Harbor no longer satisfies the stated sleeve-space requirement; Metro remains a candidate. The conclusion changes because the buyer’s constraint changed. That is the point of a useful comparison.
On a real page, replace every invented fact with verified evidence and put source links near the relevant rows. Do not publish these fictional numbers as a ready-made comparison of actual products.
5. A reusable page outline
- Decision and scope: who the comparison serves, market, exact models and date checked.
- Relationship disclosure: identify the brand publishing the page and any relevant commercial relationships. Do not imply independent editorial testing.
- Short answer: explain which option suits which need, with the conditions that could change the answer.
- Comparison table: consistent criteria, units, evidence labels and source links.
- Tradeoffs: explain the practical differences and limitations that the table cannot capture.
- Method and unknowns: separate desk research from hands-on testing and state unanswered questions.
- Next decision: link to exact products, applicable policies and deeper evidence; offer a way to report factual corrections.
Answer the buying question in visible page text, not only in an infographic. On mobile, keep row labels and product names understandable when the comparison table scrolls. Test links to the exact variants rather than assuming a category link preserves the context.
When there is no genuine difference in buyer intent, improve the existing comparison instead of making multiple near-identical “best,” “versus” and “alternatives” pages. A separate page should resolve a distinct decision, not simply rearrange the same claims.
6. Review, publish and measure without promising an AI recommendation
Assign a reviewer to challenge the strongest claims: would the statement still be fair if the competitor published it about you? Check that each conclusion follows from the evidence, that unknowns remain unknown, and that your own product’s limitations are as easy to find as its benefits.
Record who owns the page and what triggers a review: a new model, changed dimensions, discontinued variant, promotion or policy revision. Update the factual record first and change the visible review date only when a substantive check has occurred.
Google says a supporting page in AI Overviews or AI Mode must be indexed and eligible for a Search snippet; there is no special AI markup requirement. Eligibility does not guarantee inclusion. These are Google-specific rules, not a promise about every AI engine. Source: Google’s AI features documentation.
After release, record the URL and actual publication date. Monitor a consistent set of buyer questions and separate page citations, brand mentions and identifiable referral outcomes using the GEO measurement scorecard. A citation does not establish endorsement, and a before-and-after change does not establish causality.
If independent comparison pages appear in answers, investigate them through the citation-gap playbook. Your owned comparison and a publisher’s independent review have different roles. Neither should pretend to be the other.
YOUR NEXT STEP
Bring one buying decision and the evidence behind it.
Start with two plausible alternatives, a small set of meaningful criteria and a clear record of what you can verify.
Discuss your AI search content