The Recommendation Integrity Audit: What to Verify Before Referring a Reader

The reader experiences the whole handoff

A publisher may think its responsibility ends when a reader leaves the article. The reader experiences one continuous journey: search result, explanation, recommendation, merchant page, checkout, onboarding, support, renewal, and possible refund. A misleading handoff can therefore damage trust even when the publisher did not write the merchant's checkout copy.

A recommendation integrity audit asks whether the publisher has enough current, relevant evidence to help a specific reader investigate further. It does not certify a product as universally good, predict earnings, or guarantee the merchant's future behavior. It records what was checked, what remains unknown, and when the recommendation must be revisited.

This approach also creates original value. Google Search's spam policies distinguish thin affiliation, which repeats merchant material without added value, from affiliate pages that provide meaningful information, original reviews, testing, ratings, navigation, or comparisons. Due diligence is not only a compliance task; it is the editorial work that gives the page a reason to exist.

Evidence: Google Search Central; Google Search Central

Separate five evidence files

Keep five logical files even if one document stores them. Product facts cover features, delivery format, prerequisites, and limitations. Merchant operations cover identity, support routes, billing, renewal, refund process, and complaint patterns. Audience fit explains who may benefit and who should choose another path. Evidence quality records demonstrations, testing, independent sources, and unresolved claims. Commercial relationship records commission, free access, sponsorship, ranking influence, and disclosure language.

Separation prevents evidence leakage. A clear feature list does not prove responsive support. A generous refund policy on paper does not prove the process is easy to use. A high commission does not make a product unsuitable, but it creates a material relationship that readers need to understand. A positive customer story does not establish a typical result.

• Product: what is delivered, required, changed, or excluded?
• Merchant: who charges, supports, renews, and refunds?
• Audience: what problem, stage, budget, and constraints fit?
• Evidence: what was observed, tested, compared, or left unknown?
• Relationship: what benefit or influence connects publisher and seller?

Evidence: Google Search Central; Google Search Central; U.S. Federal Trade Commission

Build a vendor record before drafting praise

Record the legal or trading name shown to customers, official domain, contact route, product owner, payment processor where visible, current price and currency, tax presentation, subscription or renewal behavior, cancellation path, refund window and exclusions, delivery timing, support channel, privacy policy, and terms date. Capture direct URLs and a dated note rather than relying on a screenshot alone.

Test only what you are authorized and able to test. You can inspect public terms without buying. If you purchase or receive access, document the version, device, location, and scope of use. Do not claim to have completed a course, used support, received a refund, or achieved an outcome unless that experience genuinely occurred and can be described accurately.

Mark contradictions. If a sales page says “lifetime access” while the terms describe a fixed period, do not choose the more attractive phrase. Ask the merchant for clarification or withhold the recommendation. If support or refund details cannot be found, write “not verified” rather than assuming industry norms.

Evidence: Google Search Central; U.S. Federal Trade Commission

Use an evidence ladder and label your access

Google's guidance for high-quality reviews recommends evaluating from a user's perspective, showing knowledge, providing evidence of experience, using quantitative measurements where useful, explaining differentiators, discussing benefits and drawbacks, and identifying important decision factors. Follow only the practices you can honestly support. If there has been no hands-on use, publish a researched overview or terms analysis, not a first-person review.

Label evidence next to the conclusion: “Based on public documentation reviewed on this date,” “Tested in the trial version under these conditions,” or “We could not verify support response time.” This helps readers calibrate confidence and makes later updates efficient.

• Documented: supported by current public product or merchant material.
• Observed: directly visible in a demo or authorized account.
• Tested: a named task was performed under stated conditions.
• Compared: the same declared criteria were applied to alternatives.
• Unverified: important information could not be confirmed.

Evidence: Google Search Central

Write the non-fit section before the recommendation

A recommendation becomes more credible when it names boundaries. State the required budget, time, skill, equipment, language, region, tolerance for subscription, or willingness to perform manual work. Then name common situations in which a free resource, different category, or no purchase is more suitable.

Avoid invented personas such as “perfect for anyone serious about success.” Use decision conditions: “May fit a beginner who wants a structured sequence and can practice two hours per week; may not fit someone seeking done-for-you implementation or a guaranteed outcome.” Conditions are inspectable and can be revised when the product changes.

Evidence: Google Search Central; Google Search Central

Place disclosure where it changes interpretation

Disclosure should arrive before or with the recommendation, in language an ordinary reader can understand. “We may earn a commission if you buy through a qualifying link, at no additional cost to you” explains the financial relationship more clearly than an isolated label that assumes readers know what affiliate tracking means. Do not hide it in a footer, a general about page, or a collapsed section after the link.

The FTC's Disclosures 101 guidance says material connections should be made obvious and disclosures should be hard to miss, close to the endorsement, and expressed in simple language. The UK's CAP guidance likewise explains that affiliate content in blogs or news-style settings may need additional disclosure so its commercial nature is obviously identifiable. These are jurisdiction-specific authorities, so check the rules relevant to the publisher and audience.

Disclosure is necessary but not curative. It does not make an unsupported claim true, a fake review acceptable, or a pay-to-play ranking independent. Keep commercial influence out of evidence scoring and record any free product, bonus, sponsorship, or placement arrangement.

Evidence: U.S. Federal Trade Commission; UK Advertising Standards Authority and CAP

Audit the merchant-side experience

Open the handoff as a new reader would. Does the destination match the named product and current price? Are region, currency, tax, shipping, subscription, or upsell conditions introduced clearly? Does checkout request unexpected data? Can key terms be found before payment? Does the support contact work? Record observations without entering false information or completing a purchase unless authorized.

A merchant may change a landing page without notifying affiliates. Use scheduled link and term checks, but do not automate purchases or accept new terms. High-risk changes such as a new recurring charge, altered refund policy, aggressive income claim, broken domain, or different product should suspend the recommendation until a human re-verifies it.

• Destination and product identity match the article.
• Current price, recurrence, and key conditions are visible.
• No unexplained redirect or domain change occurs.
• Support and refund routes are discoverable.
• Material changes automatically pause or flag the recommendation.

Evidence: Google Search Central; Google Search Central

Score readiness without hiding red flags

A simple readiness table can score product facts, merchant transparency, audience fit, evidence depth, disclosure, and updateability from zero to two: missing, partial, or verified for the intended scope. The total helps prioritize work, but red flags override the sum. An unsubstantiated health or earnings claim, hidden recurrence, unverifiable merchant identity, fake reviews, or contradictory refund terms should stop publication even if other fields score well.

Record the decision as approved, approved with limitations, pending evidence, or rejected. Approval belongs to an exact article revision and evidence date. A product update, terms change, credible complaint pattern, or long review interval returns it to pending rather than inheriting permanent trust.

Do not turn the score into a consumer-facing star rating unless its method, weights, and limitations are suitable for that use. Internally, its purpose is completeness and accountability. Externally, explain the decision factors in plain language.

Evidence: Google Search Central; Google Search Central; U.S. Federal Trade Commission

Maintain trust after publication

Set review frequency by volatility. Static educational concepts may need an annual check. Price, subscription, platform features, and refund terms may need monthly or quarterly review. Event deadlines need explicit expiry. Use automated monitoring to flag changes, but require human review before rewriting conclusions.

The final recommendation should sound like a handoff: what was checked, why the option may fit, where it may not, what remains unknown, what the relationship is, and which current details the reader should confirm. That is more useful than certainty theater. It respects the reader as the decision-maker and gives the publisher a standard that can survive beyond one campaign.

Evidence: Google Search Central; U.S. Federal Trade Commission; UK Advertising Standards Authority and CAP; U.S. Federal Trade Commission