EyeVerdict
Methodology

How a recommendation gets made

Every quiz answer routes through documented, version-controlled rules — authored by a practicing optometrist, reviewed by a separate licensed eye-care professional, and citable in an FTC audit. No language models, no probabilistic rankers, no proprietary black boxes between your inputs and our recommendation. The full rule set is reviewed every 90 days and versioned in source control.

Last reviewed: July 2026
Medically reviewed by Dr. Michael Brown, OD · Last reviewed July 2026

Rule-based, not machine-learned

Each tool is a deterministic decision tree. The mapping from quiz answers to a category recommendation is a finite set of if/then rules in the source code. The same answers always produce the same result. No language model, no probabilistic ranker, no reinforcement loop sits between the user’s answers and our recommendation.

This is a deliberate compliance choice. Under FTC substantiation standards (16 CFR 255 and the 2024 Quincy / Prevagen verdict), a health recommendation needs a defensible chain of reasoning. A rule-based engine produces that chain by construction; a language model does not.

What goes into a recommendation

Inputs vary by tool. The progressive vs bifocal configurator uses nine variables: age range, near-vision difficulty, primary screen distance, presence of a distance-vision prescription, driving frequency, current eyewear, screen hours per day, budget, and a self-reported emergency-symptom check. The reading-glasses strength estimator uses four. The dry-eye drops comparison uses three filters plus a symptom-severity gate.

No demographic profiling beyond age range. No geolocation fingerprint. No third-party data overlay. Inputs are anonymized at storage.

Citation policy

Medical content on EyeVerdict carries at minimum three peer-reviewed citations per 800 words. Accepted sources:

  • American Academy of Ophthalmology Preferred Practice Patterns
  • National Eye Institute condition pages
  • JAMA Ophthalmology, NEJM, Cochrane systematic reviews
  • Peer-reviewed primary trials on PubMed

Phrases like “studies show” or “research suggests” without a citation are not used.

How retailers and products are selected

Each output category is mapped to a curated set of partner retailers in our database. A retailer enters that set only when:

  • It has an active, public affiliate program we are approved into.
  • Its catalog covers the relevant lens or drug class in the output category.
  • It has no documented FDA or FTC enforcement against its claims (we exclude multi-claim disease-treatment supplements).
  • It ships to the visitor’s region.

Within a matching set, the order of retailers can favor higher-commission partners — but a retailer is never included or excluded based on commission alone. See the affiliate disclosure for the full partner list.

Exclusion criteria

The following brands are excluded from all EyeVerdict tools:

  • Any supplement marketed with compound disease claims spanning AMD, glaucoma, cataracts, and dry eye simultaneously — this claim pattern is at or past the FDA structure-function boundary.
  • Brands with documented FDA warning letters or active FTC enforcement actions.
  • Brands without a verifiable public affiliate program (we cannot map a trackable link).

Hard-route policy

Any quiz answer indicating a potential eye emergency triggers a hard-route overlay that suppresses all product recommendations and directs the user to urgent medical care. The trigger list follows AAO emergency-symptoms guidance:

  • Sudden vision loss
  • New flashes or floaters
  • Sudden double vision
  • Eye pain with nausea
  • Halos around lights with headache
  • Curtain or shadow across vision
  • Painful red eye
  • Vision change after head injury

On the hard-route screen, no affiliate links are shown.

Update cadence

Every medical page carries a last-reviewed date and is re-reviewed at minimum every 90 days. We re-review sooner when relevant clinical evidence changes (a new RCT alters a treatment claim) or when an FDA or FTC enforcement action affects a listed brand.

Related