A comparison guide buyers can check starts with the decision, not the winner. State who the guide is for, the criteria that matter for that buyer, and where each product’s evidence comes from. Date every finding, mark what you could not confirm, and say plainly which buyers each product suits and which it does not. If you sell one of the products, say so before the first table.
A ranked list with a crown on your own product is quick to publish and easy to distrust. A buyer cannot tell whether the order reflects testing, a commission or the author’s employer. An AI assistant reading the page has the same problem: it can see that you put yourself first, but not why anyone should agree.
Google’s guidance for review content asks for the reasoning behind the ranking. When something is called the best overall or best for a purpose, it says to explain why, with first-hand supporting evidence. It also asks reviewers to cover comparable options, explain which might be best for certain uses, discuss benefits and drawbacks, and make sure ranked lists are useful on their own. Read Google’s guidance on writing high-quality reviews.
Say who is writing and what they sell
Put the disclosure near the top, in plain words: “We make Harbourline, one of the three tools compared here.” Readers weigh a comparison differently once they know the author has a stake in it, and they should be able to.
In the United States, the FTC’s Endorsement Guides require a connection between an endorser and a seller to be disclosed clearly and conspicuously when it could materially affect how much weight the audience gives the endorsement and the audience would not expect it. Other markets have their own rules, so check the ones that apply to you. Even where no rule applies, an undisclosed stake is the fastest way to lose a careful buyer. Read 16 CFR 255.5 on material connections.
The disclosure is also where you introduce your product honestly. Name it, say what problem it is built for, and point to the evidence later in the page. That gives readers and AI systems a clear statement of who you are, without asking them to take your ranking on trust.
Seenvero publishes this article and sells an AI visibility product. The article does not compare AI visibility tools, and every product in the example below is invented.
Choose the criteria before you look at the products
Write the criteria from the buyer’s situation, not from your feature list. A good test is whether a buyer who has never heard of you would still recognise each criterion as important. If a criterion only exists because your product wins it, drop it or label it as your differentiator.
For each criterion, decide what counts as evidence. “Integrates with our accounting system” can be checked against an integrations page or documentation. “Easy to use” cannot be checked from a vendor site, so either test it yourself and describe the test, or leave it out.
Use a table that shows its work
Every row in a checkable comparison carries the same fields:
| Field | What to record |
|---|---|
| Criterion | The buyer need being compared, in the buyer’s words |
| Finding | What you found for this product, without adjectives |
| Source | The page, document or test that supports the finding |
| Checked on | The date you opened the source or ran the test |
| Plan or region | The plan, tier or market the finding applies to |
| Limitation | What the finding does not cover |
| Status | Verified, vendor-stated, or unknown |
The status column is the one most guides leave out. Verified means you tested it or confirmed it in documentation. Vendor-stated means it comes from marketing copy you could not confirm. Unknown means you looked and found nothing. An honest “unknown” is more useful than a confident guess, because it tells the buyer exactly which question to ask on the sales call.
A worked example with fictional products
The following example is fictional. Harbourline, Waypace and Routekit are invented route-planning tools, and so are their plans, prices and features. Harbourline is the publisher.
The guide is written for a delivery business running about ten vans that wants to replace spreadsheet route planning.
| Criterion | Product | Finding | Source | Checked on | Plan or region | Limitation | Status |
|---|---|---|---|---|---|---|---|
| Driver app works offline | Harbourline | Routes stay available without signal; updates sync on reconnect | Our offline test, method described below | 26 Oct 2026 | All plans | Tested on Android only | Verified |
| Driver app works offline | Waypace | Help centre says routes are cached for 24 hours | Waypace help centre article | 26 Oct 2026 | All plans | Not tested by us | Vendor-stated |
| Driver app works offline | Routekit | No public information found | Routekit docs and pricing page | 26 Oct 2026 | — | Ask the vendor | Unknown |
| Price for ten vans | Harbourline | €49 per van per month | Our pricing page | 26 Oct 2026 | EU, monthly billing | Excludes VAT | Verified |
| Price for ten vans | Waypace | €390 per month for up to ten vans | Waypace pricing page | 26 Oct 2026 | EU, annual billing | Annual commitment; excludes VAT | Verified |
| Customer delivery notifications | Routekit | SMS and email notifications on all plans | Routekit feature page | 26 Oct 2026 | All plans | SMS credits billed separately | Vendor-stated |
In a real guide, the source column holds the URL itself, so a reader can open it. Notice what the table does not do. It does not add up a score, and it does not hide that Waypace is cheaper for this buyer on annual billing. A reader comparing the price rows can see that ten vans on Harbourline cost €490 a month against Waypace’s €390, before VAT and with different billing terms. If your product loses a row, show the loss. A guide that never concedes anything reads as an advert, because it is one.
Describe any test you ran, close to the table. For the offline row: “We loaded a 25-stop route, switched the phone to aeroplane mode, completed ten stops, then reconnected and checked that all ten were recorded.” A method that short still lets a reader judge whether the result applies to them.
Replace the ranking with fit
After the table, write one short section per product that says who should choose it and who should not. This is where a comparison becomes useful to a buyer and to an AI assistant answering “which tool suits a fleet like mine?”
For the fictional guide:
- Choose Waypace if you can commit to annual billing and your drivers mostly work in areas with reliable signal.
- Choose Routekit if customer notifications matter most, and confirm offline behaviour and SMS costs before you sign up.
- Choose Harbourline if drivers regularly lose signal and you want monthly billing. It is not the cheapest option for a ten-van fleet that can commit to a year.
Each recommendation names a condition the reader can check against their own situation. None of them claims a product is best for everyone, and the publisher’s product gets the same treatment as the others, including its weakness.
Google’s guidance on AI features in Search recommends unique, non-commodity content and gives a first-hand review as an example of a perspective that a summary of existing content does not offer. A comparison with your own tests, dates and fit boundaries is that kind of page. A list of product names in your preferred order is not. See Google’s guidance on AI features.
Keep the evidence current
A comparison ages faster than most pages, because three companies can change their pricing and features without telling you. Put a “last checked” date at the top of the table, recheck each row on a schedule your team can keep, and update the date only when you have actually rechecked the sources.
When a competitor changes something, update the row even if it makes your product look worse. A stale row that flatters you is the kind of error a buyer finds on the competitor’s own pricing page, and after that they stop trusting the rest of your table.
Link each claim about your own product to the page that proves it. If that page does not exist yet, our article on building a product page that answers buyers’ proof questions covers what it should contain.
Check how AI answers use your comparison
Publishing the guide is half the job. The other half is seeing what AI assistants say when buyers ask the same comparison question.
Seenvero tracks the questions you choose, such as “Which route-planning tool suits a ten-van fleet?”, across AI assistants. Brand Insights splits each mention of your brand into recommended, listed, mentioned in passing or cautioned against, and shows the sentences behind the split. That separation matters for comparison content: appearing in a list is not the same as being the suggested choice. The Sources view shows which pages the answers cite, so you can see whether assistants are reading your comparison, a competitor’s, or a third-party review.
Treat what you see as an observation. If an answer starts recommending your product for drivers in low-signal areas after the guide goes live, that is worth noting, but it does not prove the guide caused it. Record the questions and answers before publication so you have something comparable afterwards. Our article on telling whether AI is recommending your brand explains how to keep those measurements consistent.
See how recommendations, mentions and cited pages appear in Seenvero’s sample report, which uses an explicitly fictional brand, then draft the criteria list for your own comparison before you open a single competitor’s website.
Common questions
Should we put our own product first?
Only if your evidence shows it fits the buyer the guide is written for better than the alternatives, and the guide explains why. Otherwise organise by fit or alphabetically and let the conditions do the work. Google’s review guidance asks that a “best” label come with the reason and supporting evidence.
Can we name competitors in our comparison?
Most comparison guides do, and buyers expect them to. Keep each claim to what the competitor’s current public pages or your own documented tests show, date it, and link to the source. Where you could not find something, mark it unknown rather than assuming the worst.
How often should we update a comparison guide?
Recheck pricing and plan details whenever you know a vendor has changed them, and review every row on a regular schedule your team can maintain. Change the “last checked” date only after a real recheck.