Most affiliate advice treats trust as a nice extra that helps conversions along. I’d argue it’s the reverse: trust is the asset, and conversions are what it pays out. A reader who believes your recommendations will click the next one, and the one after that. A reader who feels sold to once discounts everything you publish afterwards.
This piece separates what is established fact from what is my opinion, because that distinction is the point. The filter I propose at the end is an opinion about how to run an affiliate site. It is built on one plain fact: you get paid when someone acts on your recommendation, and readers know it.
The recommendation that costs you more than it earns
Picture a reader who searches for a tool, lands on your comparison post and buys through your link. The tool turns out to be wrong for their situation. You earned a commission, and the reader now knows that your site’s “best for beginners” label can’t be relied on.
The commission shows up in your dashboard. The cost doesn’t show up anywhere. It arrives later, as a reader who doesn’t return, doesn’t subscribe and doesn’t click the next link, and no report will tie that behaviour to the bad recommendation.
That asymmetry is why affiliate marketing ethics and trust belong in the strategy itself rather than in a footnote. Your income report records the sale, and it can’t record the damage. If you only judge recommendations by what the dashboard shows, you will keep making ones that cost you more than they pay.
Why commission rate is the wrong first question
Affiliate marketing is a performance-based model: you earn when a reader takes a defined action, such as a purchase or a signup. That’s the established part. The rate attached to that action says something about the merchant’s margins and marketing budget. It says nothing about whether the product suits your reader.
When you start from the commission rate, the research bends toward the answer you already want. You look for reasons to like the high payer and reasons to doubt the low one. Readers can often sense this, even when they couldn’t explain why.
A better order of questions:
- What problem does my reader have?
- Which products solve it, whether or not I have an affiliate relationship with them?
- Of the ones that do, which have an affiliate program?
Commission enters at step three, and only as a tiebreaker between products that are already good enough. Your own publishing standards should say the same thing: affiliate recommendations should reflect genuine relevance, not commission rates alone.
Trust is a growth metric, so measure it like one
Most site owners track clicks, conversions and revenue per visit. Those measure the transaction. Trust shows up in behaviour over time, so track the signals that sit upstream of the sale.
Some that are worth watching:
- Returning visitors as a share of traffic. A rising share suggests readers found the site useful enough to come back.
- Replies and questions. If readers email you or comment asking what you’d recommend, they treat you as an adviser.
- Unsubscribes and complaints after affiliate-heavy emails. A spike after a promotional send tells you where the line was.
- Clicks per page alongside post-click outcomes. If your affiliate program reports returns, cancellations or refund rates, a high click count paired with poor outcomes is a warning.
- Branded search. People searching for your site by name are choosing you before they know what you’ll recommend.
I can’t give you benchmark numbers for these, and I’d distrust anyone who does without a source. Rates vary by niche, traffic source and audience size. What matters is your own trend line: is it moving the right way as you add more affiliate content, or the wrong way?
My opinion is that trust is the one metric that compounds. Each honest recommendation makes the next one more credible, and each regretted one makes the next one harder. Revenue per click resets every month, and reader confidence doesn’t.
The human filter: five questions every affiliate link has to pass
A filter only works if it’s short enough to use every time. These five questions are mine, and they’re opinion rather than an industry standard. Run every link through them before it goes live. If a link fails one, fix the recommendation or remove the link.
1. Would I still recommend this if the commission were zero?
If the answer is no, you’re writing an ad. That can be legitimate when it’s labelled as one. It doesn’t belong in a page that presents itself as independent advice.
2. Does it solve a problem this specific reader has?
“Popular” and “well reviewed” don’t answer this. Name the reader, the situation and the problem. If you can’t finish the sentence “this is for people who…”, the recommendation is too vague to be useful.
3. Can I say who it is not for?
Every real product has a wrong customer. A recommendation with no limits reads like marketing copy because it is. Stating who should skip it is the cheapest way to prove the rest is honest.
4. Do I know this from first-hand use or from a source I can name?
Be strict about the difference. If you’ve used it, say what you did with it. If you haven’t, say what your assessment rests on, such as the vendor’s documentation, published pricing or reader reports. Never imply a test that didn’t happen.
5. Would I be comfortable if the reader saw the commission next to the recommendation?
Imagine the disclosure and the payout printed beside the link. If that makes you want to reword the paragraph, the paragraph is doing something your reader wouldn’t accept.
None of the five asks you to avoid earning money. They ask whether the recommendation would hold up if the reader knew everything you know.
Where AI helps with affiliate content and where it shouldn’t decide
For people building AI-assisted content workflows, the risk is specific: a language model will write a confident product review whether or not anyone has examined the product. Left unchecked, it can describe features that don’t exist, repeat outdated pricing and write “in my experience” sentences that no experience stands behind.
Where AI is useful:
- Organising your research into a comparison structure
- Summarising long documentation so you can check it against the source
- Listing questions a reader might ask that your draft doesn’t answer
- Flagging sentences that make claims you haven’t supported
- Tightening drafts once the substance is settled
Where it shouldn’t decide:
- Whether a product earns a recommendation. That’s question one on the filter, and it needs your judgement.
- Facts about features, pricing and terms. Check these against the vendor’s current page, because they change.
- Anything that implies experience. If a sentence says “I tried” or “we found,” a person has to be able to stand behind it.
- Who the product is not for. A model tends to flatter whatever it’s describing.
I’ve written separately about using AI without losing credibility, and the same principle applies here: the tool can draft, but a person is accountable for every claim that carries your name.
Disclosure that readers actually notice and respect
Disclosure is a legal and ethical requirement, and here is the established part. In the US, the FTC’s Endorsement Guides require you to disclose material connections, such as an affiliate relationship, clearly and conspicuously. Its guidance says the disclosure should be close to the recommendation and not buried in a footer or on a separate page. Other countries have their own advertising rules, and some affiliate programs set disclosure terms of their own, so check the rules where you and your readers are based.
Here’s my opinion on doing it well. A disclosure that exists only to protect you tends to be written in a way no one reads. One that serves the reader is short, specific and placed where the decision happens.
- Put it before the first affiliate link, not after the last.
- Use plain words: “If you buy through this link, I earn a commission at no extra cost to you.”
- Only say “at no extra cost” if that’s true for the program in question.
- Repeat it where links appear in a later section or a newsletter, since readers jump around.
- Say what the commission does and doesn’t influence, if you have a real editorial policy.
Readers who see a clear disclosure at the point of recommendation generally have no reason to suspect something is hidden. Hiding the relationship is what creates the suspicion in the first place.
What to do when the best product pays the worst, or nothing
This will happen. The product that fits your reader best may have a low commission, an awkward program or no program at all.
Recommend it anyway. That’s the filter working as intended. Say it’s the strongest fit, and say plainly that you don’t earn from it if that’s the case. That sentence does more for your credibility than any amount of enthusiastic copy, because it shows readers your list isn’t ranked by payout.
Some practical handling:
- Give it the prominent position. Don’t push the better-paying option above it.
- Keep the paid alternative in the comparison only if it fits. If it suits a different type of reader, say which type.
- Don’t pad the weaker option’s pros. Describe it accurately and let readers choose.
- Look for earnings elsewhere on the page. If another genuinely relevant product belongs in the article, include it on its merits.
The trade-off is real: in the short term you may earn less from some pages. I can’t promise it pays back in higher conversion, and nobody honestly can. My bet is that a site known for recommending the right product, even when it’s the unpaid one, holds its audience longer than one that never does.
Audit your own affiliate links this week
This takes an afternoon. Pull up your five most-clicked affiliate links, which are the ones with the most reader trust riding on them. Your affiliate dashboard or link-tracking tool should show which they are.
For each link:
- Run it through the five questions above.
- Check that the facts about the product, its pricing and its terms are current.
- Look at the disclosure near the link and ask whether a reader would notice it.
- Decide on one of four outcomes: keep as is, rewrite, add a limitation or disclosure, or remove.
Note which questions your links fail most often. If it’s question three, you probably aren’t saying who the product is wrong for. If it’s question four, your sourcing is the weak point. That pattern tells you what to fix in your process, not just in the five pages.
Pull up your five most-clicked affiliate links and run each one through the human filter this week. If you’d like a second pair of eyes, get in touch about a trust audit of your monetisation setup.