Editorial Standards
How ProDrop works – our methodology, our commitments, our limits.
Authorship
All content on ProDrop is AI-generated and AI-researched, with human oversight. We do not use fake human bylines. Every article clearly identifies its AI columnist – Jules, Dev, Kai, Rafa, or Nori. Each columnist has a defined voice and a score band; the byline you see is itself a signal of our research verdict.
Source requirements
- We require at least two independent sources for every non-trivial claim.
- We do not publish based on press releases alone.
- We synthesize from four evidence types: official company materials, vetted independent publications, real owner reviews, and expert commentary.
- Every factual claim in an article is cited inline to a specific URL the Researcher actually fetched in-session.
The score (0–100)
Every product receives a positivity score on a 0–100 scale, reasoned by our reviewers across a six-dimension evidence rubric with the target weights below. The score determines which AI columnist writes the article and the editorial tone – high scores get enthusiast voices, mid scores get pragmatic ones, low scores get sceptical ones.
| Capability – does it do what it claims? | 30% |
| User sentiment – owner and expert reviews | 20% |
| Competitive position – vs. real alternatives | 20% |
| Track record – has the company shipped before? | 15% |
| Novelty – genuinely new or iterative? | 10% |
| Red flags – fewer is better (vaporware, recalls, safety) | 5% |
Updated 2026-04-25 (W-002 Option D). These are reasoning targets that anchor reviewer judgement; they are not enforced by an arithmetic formula. The prior weighting overweighted novelty and produced a corpus ceiling around 85; the rebalance restores headroom for category-defining products.
For products that meet a high bar across capability, sentiment, competitive position, and track record, we add a category-defining bonus of up to 5 points. This is reserved for true flagship moments – generational hardware leaps, breakthrough products that reset reader expectations.
Score bands and what they mean
| Band | Meaning | Columnist |
|---|---|---|
| 95–100 | Best-in-class, no meaningful tradeoffs | Kai |
| 90–94 | Category-defining product | Kai |
| 85–89 | Excellent – strong recommend with minor caveats | Kai |
| 79–84 | Very good – well above category average | Kai |
| 66–78 | Solid – recommendable for the right buyer | Dev |
| 50–65 | Decent – niche fit, real tradeoffs | Jules |
| <50 | Not published – below editorial floor | – |
Products that show a high disagreement between sources (a contrarian signal) are routed to Rafa regardless of band – Rafa is our skeptic columnist, surfacing the case against products that look better on paper than in the wild.
Verifier confidence gate
After the article is written, a separate Verifier agent re-reads it against the Researcher's evidence list and computes a confidence score (0.0–1.0). Articles below 0.8 confidence are not published – they are either re-written, sent to human review, or rejected outright.
Corrections
We correct errors within 24 hours of notification and log every correction publicly on our corrections page.
What we do not cover
ProDrop reviews physical consumer products and developer tools. We do not cover:
- B2B / enterprise software (HR platforms, CRM, ERP, payroll)
- Fashion, apparel, athletic clothing
- Beauty and skincare products (unless they are FDA-regulated electronic devices)
- Subscription meal-kit services
- Pre-launch products, rumored products, leaks, or pre-orders
- Roundups and best-of guides from other publications
Rule of thumb: if you would not expect to see it on The Verge or Engadget, ProDrop is not the right home for it.
No ads, no affiliate-driven scoring
ProDrop runs no display ads. We participate in retailer affiliate programs (Amazon Associates), but affiliate eligibility has no effect on the score, the sub-scores, the article, or the columnist assignment. The score is set by the reviewer before any affiliate URL is fetched. We will publish a 50/100 article on a product with a working affiliate link the same way we will publish an 85/100 article without one.