Use AI to refresh an affiliate guide without inventing product experience

Start with the reader's decision

Choose a page that already answers a concrete question: which desk fits a small room, which email tool suits a solo creator, or which travel bag fits a particular airline's published allowance. Avoid beginning with a request to make the page longer. Length is not the missing value if the reader cannot understand the tradeoff.

In a hypothetical home-office guide, a reader may care more about usable depth, monitor clearance, and return costs than the number of desks listed. Write that decision at the top of your brief. Then separate your original testing notes, official product details, and unanswered questions into three folders or document sections. Keep the dates attached.

Ask for a claim audit before a rewrite

Give the assistant your page and the source material you are allowed to use. Ask it to find claims that need checking, including prices, compatibility, availability, warranties, and superlatives. A model's confident answer is not evidence. Open the underlying source yourself and check that it describes the correct product version and market.

If browsing is unavailable, provide current source excerpts and tell the assistant that the research boundary is those excerpts. Mark gaps explicitly instead of asking it to fill them from memory. A comparison with two verified facts and a visible unknown is more defensible than a polished table of guesses.

Prompt to try

Audit this affiliate guide against the supplied evidence. Return a table with claim, evidence source, date, status, and recommended action. Use only verified, outdated, unsupported, or firsthand-note as statuses. Flag any sentence that implies I personally used a product without a matching testing note. Do not rewrite the guide yet. Reader decision: [decision]. Guide and sources: [paste].

Preserve the boundary between research and testing

Use direct language about your method. If you compared published specifications, say so. If you tested something, explain the conditions and limits using your actual notes. Do not let AI add statements such as we found, in our testing, or customers love unless you have the corresponding evidence and permission where needed.

The FTC's guidance says endorsers cannot describe experience with a product they have not tried. Its affiliate guidance also calls for a clear, conspicuous explanation of the financial relationship. Put a plain explanation near the recommendation, where readers can understand it before acting. A disclosure does not repair an invented review.

Improve the decision, not just the wording

Ask for a short fit assessment for each option: useful for this situation, less suitable for that situation, and information still needed. In the hypothetical desk example, a narrower desk might suit a laptop user but require a measurement check for a monitor arm. That is a useful distinction without pretending to have assembled the desk.

Make the recommendation conditional on the reader's needs. If the evidence cannot justify a winner, offer a comparison framework. Do not rank products by commission while presenting the ranking as an independent performance result. Keep affiliate terms separate from editorial judgment.

Prompt to try

Rewrite only the approved sections using the verified evidence below. For each product, explain who it may suit, the relevant tradeoff, and what a buyer should check. Preserve uncertainty. Do not invent testing, testimonials, performance results, or product benefits. Keep the affiliate disclosure clear. List any sentences you could not substantiate. Approved facts: [paste].

Review the page as a reader

Check every purchase link, comparison row, and mobile disclosure placement. Read the page without the product names: does it still explain the decision clearly? Keep a small change log describing what changed and why. Update a reviewed date only when you actually complete the review.

Google's people-first content guidance emphasizes original value and reliable sourcing. Apply that as an editorial test, not a ranking promise. Measure useful outcomes over a sensible period: broken-link fixes, clearer answers, relevant outbound clicks, and reader questions resolved. Revenue changes alone cannot prove the refresh caused the result.

Try it on one real task

1. Choose one existing guide and write the reader's decision in one sentence.
2. Audit ten factual claims and verify their sources manually.
3. Improve one comparison or fit explanation, then record the change.

Check before you use the result

• Firsthand statements match genuine testing notes.
• Prices and product details have current, relevant sources.
• The financial relationship is clear near the recommendation.
• Unknowns remain visible and every purchase link works.

Take the next step

Download the Affiliate Marketing Coach AI practice pack for these prompts, exercises, and review checklists.

New to AI? Take the free first-workflow course at AI Teachers. Already using it? Practice building a repeatable workflow.