How We Review Products
Our Research Process, Rating Criteria & Editorial Standards
Our Approach
Every review we publish follows the same structured research process — regardless of product category, price point, or brand. Our goal is a review that answers the questions a real buyer actually has: Does it do what it claims? Is it worth the money? What are the most common problems? How does it compare to the alternatives?
We achieve this by combining four inputs: primary product research, aggregated user data, real-world context, and automated checks. All four apply to every review.
The Four Research Inputs
1. Primary Product Research
We begin with a thorough survey of available product data: manufacturer specifications, official documentation, and retailer listings across both US and Canadian markets. We cross-reference specifications across sources and note discrepancies where found.
2. AI-Assisted User Data Analysis
We use AI tools to process large volumes of owner reviews and identify patterns — common failure points, consistently praised strengths, use-case limitations — that a small editorial team could not surface manually. Since October 2026, the claims an article relies on are checked against saved copies of their sources by a separate automated fact-checker (see How We Check the Facts below); the AI’s summaries are not individually reviewed by a person.
3. Real-World Context
We look for patterns in what owners report in published reviews and owner communities — what works, what fails, and under what conditions. When possible, we also draw on our own hands-on experience with products we’ve purchased ourselves or received as review samples directly from manufacturers — noted directly in the review whenever that’s the case.
4. Automated Checks
No manufacturer, retailer, or advertiser influences our ratings or conclusions. Before publication, automated checks look for testing claims we can’t support, contradictions and filler, and a separate fact-check verifies the article against its sources. Drafts that fail are fixed or not published.
How We Check the Facts
Since October 2026, every article we publish or update goes through these steps before it can go live. Older articles get the same treatment as we update them.
- A saved research trail. Each source the article relies on is saved as a dated copy, so every claim can be traced back to what the source actually said on the day we read it.
- Claims tied to their sources. Each finding we credit to a source, each quote and each point we describe as widely shared is recorded with the passage that supports it.
- Automatic checks. Software confirms that quotes appear word for word in the source they’re credited to, that figures match their source, that no passage copies a source’s wording, that every source we name is listed in the article’s Sources section, and that every product in a comparison is a real, currently sold model.
- A separate fact-checker. A second AI, separate from the one that wrote the article, re-reads the draft against the saved sources. It looks for claims stated more strongly than the evidence, specs that belong to a different model or size, sources that disagree, and any negative claim without a primary source. The article can’t publish until its findings are fixed and it passes.
How we use sources. When several sources agree, we say so in our own words. When a finding comes from one source — its own measurement, its verdict, or a point where sources disagree — we name it. Quotes are short, exact and marked as quotes. Every article lists its sources.
How We Score Products
We rate products on a 5-point scale (1.0 to 5.0). Scores reflect overall value to the target buyer — not a checkbox of features. A $49 air fryer that does exactly what a budget buyer needs can score higher than a $200 model that underdelivers on its premium promise.
Every score comes from our own consistent formula, applied the same way to every product and never picked by feel. It balances three things:
- How well it does the job it’s bought for. Each product is judged on what matters for its specific use (for a robot vacuum, things like cleaning power, navigation and battery life), based on lab and expert tests and long-term owner experience.
- Value for money. What you get for the price, compared with the alternatives.
- The competition. How it stacks up against the other products in the review’s comparison, so a real shortfall on something that matters holds a score down, however well the product does elsewhere.
A product’s US and Canadian reviews carry the same score, because it’s the same product, and our buying guides show the score from each product’s own review. Scores are not permanent: when we revisit a review because of new evidence, a new version or a major price change, we recalculate its score.
Prices
A review’s value score uses the price we found when we researched it. The price tables show the most recent Amazon price we’ve recorded: for US prices on products we’ve reviewed, we re-check regularly, usually daily; Canadian prices are updated when we research or update the review. Prices and availability change often, so always check the current price on the retailer’s site before buying.
US vs. Canadian Reviews
Where a product is sold in Canada, we publish a Canadian edition of its review. Canadian reviews are adapted from the US review of the same product and checked the same way: they use Amazon.ca pricing and availability (or the maker’s official Canadian price, with a note, when Amazon.ca doesn’t sell it) and add Canadian details such as certifications, warranty terms and where to buy. The rating matches the US review, because the product is the same.
Where the Canadian version of a product differs from the US version, the Canadian review says so.
Update Policy
Product reviews have a shelf life. We revisit published reviews when:
- A significant volume of new user data emerges that changes the reliability picture
- The manufacturer releases a new version or a material update
- The product’s price changes enough to affect the value assessment
- A reader submits a credible correction or flags a factual error
Each review shows the date it was reviewed, and an “Updated” date when it has been substantially revised since. That date does not indicate ongoing monitoring. Prices and availability should be verified independently before purchasing.
How We Use AI
AI does most of the work: it gathers and summarises the research, cross-references specifications across sources, and writes the article. Our editors set the research standards, the rating method and the checks every article must pass, and they handle corrections.
Articles publish automatically once they pass those checks; they are not individually reviewed or edited by a person before publication. We think readers deserve to know this.
Every review, guide and First Look shows our affiliate disclosure at the top and explains how it was made at the end. We believe this transparency is a minimum standard for AI-assisted publishing.
Commercial Independence
We earn revenue through affiliate commissions. This is disclosed at the top of every article and on our Affiliate Disclosure page. As an Amazon Associate I earn from qualifying purchases. Affiliate revenue does not influence our ratings. We don’t accept payment for reviews, rankings or placements. When a manufacturer gives us a product sample, the review says so.
Questions or Corrections
If you believe a review contains an error, or if you have a question about our methodology, use our Contact page. We read every message. To report an error, see our corrections policy.