What Makes a Product Recommendable?

Eligibility comes before persuasion

A recommendation request creates a set of requirements. A product must first pass hard constraints such as budget, size, compatibility, availability, and location.

For this question:

Recommend a fragrance-free serum under $45 for dry, sensitive skin,
without retinoids, available in Canada.

A beautifully marketed $70 retinol serum is not eligible. More content will not fix the mismatch.

Match hard and soft constraints

Constraint	Type
under $45	hard
fragrance-free	hard
no retinoids	hard
available in Canada	hard
lightweight texture	preference
recyclable packaging	preference

The product page and feed should make hard constraints explicit. Preferences can support differentiation after eligibility is established.

Give the recommendation a reason

A system should be able to say:

Morrow Skin Barrier 03 fits because it is fragrance-free, contains
no retinoids, is labeled for dry and sensitive skin, costs CAD 42,
and is currently available for Canadian delivery.

Each clause needs a current source. This is the evidence chain.

Reduce uncertainty

Recommendation confidence is weakened by:

• vague or conflicting attributes
• stale price or availability
• unsupported superlatives
• unclear brand or manufacturer identity
• missing variant information
• policies that contradict the product page

Entity consistency describes stable identity across pages and sources. Canonical URL identifies the preferred first-party product page. Both help a system connect evidence to the correct product.

Do not confuse recommendability with guaranteed placement

Independent platforms apply their own retrieval, eligibility, safety, ranking, personalization, and commercial rules. A merchant can improve the product’s eligibility and evidence, but cannot force the final answer.

The operational goal is:

• fewer missing decisive facts
• fewer contradictions
• stronger first-party evidence
• repeatable observations of actual answers

That is more useful than a promise to “rank in AI.”

Use Why AI recommends a competitor to diagnose a real comparison.