Controls for AI-Assisted Packaging Referrals

What counts as an AI-assisted packaging referral

An AI-assisted referral is any recommendation surfaced by a model, chatbot, comparison widget or generated buying guide that points a buyer toward a supplier, material or format, whether or not money changes hands. The control problem is not the model itself. It is that the recommendation reads like neutral advice while the operator may hold a commercial or personal interest in the outcome.

Three failure points recur in these setups. The disclosure sits somewhere the reader does not encounter it, the conflict tier is unassigned so nobody knows which rule applies, and no named person owns the claim text the model produces. Each needs an owner and a stored record before launch, not after a complaint.

Build the disclosure rule from the material connection test

FTC 16 CFR 255.5 requires clear and conspicuous disclosure when a connection between the endorser and the seller might materially affect the weight or credibility of the endorsement and the audience would not reasonably expect that connection. The rule lists business, family and personal relationships, monetary payment, and free or discounted products, including products unrelated to the endorsed item, as possible material connections. It also lists other benefits such as early access or the possibility of being paid. A disclosure does not need complete details, but it must communicate the nature of the connection well enough for readers to evaluate its significance.

The cited provision does not prescribe a single disclosure format. As an internal policy, consider plain language near the recommendation and before the relevant outbound link. Explain the nature of the connection, including who provides a benefit and what triggers it where relevant. This placement policy is not a legal safe harbor; review whether the disclosure is clear and conspicuous in each channel.

• Name the payer and the trigger, not just the existence of a fee.
• Place the disclosure before the link, not in a page footer.
• Repeat it wherever the recommendation repeats, including generated summaries and comparison tables.
• Keep the meaning consistent across surfaces, adapting wording and presentation to the channel and the actual connection.

Assign a conflict tier to every recommendation slot

A slot is a position in your interface or content where a supplier can appear. We recommend recording the connection for each supplier recommendation within that slot. These are internal review categories, not legal categories: Tier A covers paid placement and affiliate links; Tier B covers free products, samples, travel, event access or other potential benefits; Tier C covers business, family or personal relationships without a separate benefit; Tier D covers no known connection. Multiple categories may apply. Review materiality and audience expectations rather than assuming that a tier alone determines a legal disclosure obligation.

We recommend storing the applicable tiers and underlying connection details for each supplier recommendation, then using them to inform disclosure wording and review level. When a relationship changes, update the record and review the affected disclosures. Any proposed automated regeneration workflow should be evaluated and approved before use.

Decision table for common referral situations

Worked example: sizing a referral ledger (hypothetical)

This is an illustrative arithmetic exercise, not a quote, a market benchmark or a description of any live Teal program. Suppose a hypothetical operator runs 40 packaging referrals in a month at an average order value of 1,850 dollars and a referral fee of 7 percent. Multiply 1,850 by 0.07 to get 129.50 dollars per referral, then multiply by 40 to get 5,180 dollars for the month.

In the same hypothetical audit, 12 of those 40 referrals fail the operator's internal policy requiring visible disclosure before the first outbound link. Divide 12 by 40 to get 0.30, or 30 percent of the month's referral count. Assuming each referral has the same hypothetical order value and fee, multiply 12 by 129.50 dollars to get 1,554 dollars in fees associated with policy failures. These figures measure the internal placement policy, not whether disclosures were absent everywhere or legally inadequate.

If the operator moves to inline disclosure, rerun the same count monthly and compare the failure share over time rather than comparing revenue, which moves for reasons unrelated to disclosure.

Constrain what the assistant may assert

Separate recommendation language from factual claims. A model can rank suppliers against criteria you publish. It should not originate certification, compostability, recyclability, carrier acceptance or customs statements. Those come from a named source that a human has checked.

Washington Ecology's compostable-product labeling summary illustrates why claim review matters. It discusses third-party certification, industrial-composting standards, the FTC Green Guides, labeling and color signals, producer declarations, and restrictions on degradability terminology. Its product-specific sections distinguish film bags, other films and food service products. Do not use this summary alone to determine current applicability, an exemption or a product's compliance. Before publishing a compostability claim about a particular mailer, have a qualified reviewer check current authoritative requirements and supporting product documentation.

Build a claim blocklist, require a source link for any regulated term, and route carrier specifics to the carrier's current published guidance instead of a remembered rule. Carrier documents change, and a model's recollection is not a substitute for the live page.

Human review gates and the audit log

Not every recommendation needs a lawyer, but every new claim pattern needs a reader. Define three gates: a pre-launch gate for new slots and new disclosure templates, a change gate for edits to claim text or supplier lists, and a periodic gate that samples live pages.

We recommend logging the model version where available, the applicable tiers, supplier, displayed disclosure, reviewer and date. Retain rendered views as well as source text to evaluate placement across relevant devices and channels. Database presence alone does not establish effective disclosure, and position below the initial viewport alone does not determine legal adequacy. Check visibility and proximity in the actual user experience against your policy and applicable requirements.

Pre-launch checklist

• Every recommendation slot has a tier field and a named owner.
• Disclosure wording explains the actual connection, including the payer and payment trigger where relevant, and meets the internal placement policy.
• Relevant disclosures remain clear in generated summaries and comparison tables, with presentation adapted to each surface.
• A claim blocklist covers certification, compostability, recyclability, carrier and customs language.
• Each regulated claim links to a primary document a human has read.
• Carrier guidance is linked live rather than paraphrased from memory.
• The audit log stores the rendered page, model version and reviewer.
• A monthly sample rechecks placement, not just wording.
• Supplier relationship changes trigger a tier review.
• A named person can pause a slot without waiting for an engineering release.

Scope limits

This guide is an editorial planning framework for operators who run referral surfaces. It is not legal advice and it does not decide whether a particular connection is material, which depends on the audience and the facts. Have counsel review your disclosure program.

This guide does not cover tax treatment, supplier contract terms or specific marketplace rules. Check applicable agreements and published channel policies separately. Nothing here certifies a product, validates a compostability claim or confirms carrier acceptance. This is an educational evaluation guide, not an offer of a Teal referral program or evidence of built referral, ordering or payment capabilities.

Questions about AI referral disclosure and review controls