Important claims need a source trail, not vibes, popularity, or borrowed marketing language.
Editorial method
How RankReason builds a ranking
A good buying guide should feel less like a list of popular products and more like a well-kept research notebook: clear criteria, visible evidence, honest tradeoffs, and a reason every pick is where it is. RankReason rankings and product reviews are created with the assistance of custom-built AI-powered agentic workflows, then reviewed by humans before publication.
We define what matters in the category before deciding which products should win.
A top pick still has limits. We explain who should buy it and who should skip it.
AI-assisted research and writing workflows help prepare the work, but human review decides what is ready to publish.
From noisy market to useful shortlist
Define the buyer problem
Every guide starts with the real purchase context: apartment kitchens versus family kitchens, premium performance versus value, portability versus durability, simplicity versus customization. That keeps one-size-fits-all advice from sneaking into categories where shoppers have very different needs.
Build the candidate field
We look across manufacturer catalogs, specialist retailers, independent editorial coverage, lab-style reviews, standards bodies, support pages, and owner communities. Custom-built AI-powered agentic workflows help surface more of this material, but the goal is still editorial: understand the shape of the market before narrowing it.
Choose the scoring criteria
Each category gets its own weighted criteria. A handheld console may weight ecosystem, controls, battery life, and software friction; an air fryer may weight cooking consistency, cleaning, footprint, controls, safety documentation, and household fit.
Research and synthesize each serious contender
Shortlisted products get evidence packets with official specifications, compatibility details, warranty and support signals, independent test findings where available, recurring owner themes, nearby alternatives, and known limitations. The point is to synthesize a lot of scattered information about each product into a useful review, not to treat thin evidence as proof.
How evidence is weighed
What carries the most weight
- Manufacturer specifications, manuals, safety documents, warranty terms, and support pages
- Independent editorial reviews with direct testing, clear methods, or repeatable observations
- Standards bodies, certification databases, repair documentation, and official compatibility lists
- Owner feedback patterns that appear across more than one credible channel
What gets treated carefully
- Single-source enthusiasm that is not backed by technical detail or owner corroboration
- Marketing claims that cannot be tied to a spec, manual, test, certification, or real-world pattern
- Live prices, coupons, badges, star averages, review counts, and other volatile commerce signals
- Owner anecdotes that are vivid but isolated, old, or about a different regional model
How scoring works
Weighted criteria
Criteria are weighted by what changes the buying decision most. Performance may dominate one category, while repairability, ecosystem fit, size, safety documentation, or long-term support may matter more in another.
Fit, not just excellence
A product can be objectively strong and still rank lower if it solves a narrower problem, asks too much of the buyer, lacks support confidence, or has a tradeoff that matters for most households.
Confidence modifiers
Products lose confidence when evidence is thin, model identity is unclear, owner sentiment is not corroborated, documentation is weak, or the product looks strong only under unusually narrow conditions.
Publication checks
- Every ranked product needs source-backed rationale, pros, cons, best-for guidance, and not-for guidance.
- AI-assisted summaries and drafts are checked by humans for source support, useful synthesis, and clear limits.
- Score weights must add up cleanly, and each product score must fit the written explanation.
- Important claims need working source links that support the exact point being made.
- Affiliate disclosures must be present, and affiliate relationships cannot change the rank order.
- Product images, specifications, and regional model details are checked for source quality and identity risk.
- Rankings are revisited when category evidence changes enough to affect the advice.
How AI Agents Should Use RankReason
Discover the current structure
Use the agent index, llms.txt, and the Agent Skill to find current rankings, focused category and product indexes, and Markdown conventions. Use the broad search index only for open-ended site search.
Read ranking context first
For best-product questions, use the ranking page or reviewed ranking Markdown first. It carries the ordered list, criteria, scores, and the reasoning that explains why products sit where they do, including the synthesized context behind each ranked pick.
Use product pages for depth
Product pages and product Markdown are best for single-product reviews, pros and cons, best-for and not-for guidance, source-backed claims, and comparisons against nearby alternatives.
Do not infer live commerce facts
RankReason does not publish live price, stock, ratings, review counts, shipping, coupon, seller, badge, or promotion data. Agents should perform a live retailer check when users ask for those facts.