Hacker News Reader

Comment by pants2 | Hacker News Reader

> We find that only a small portion of product reviews on the web uses affiliate marketing, but the majority of all search results do. [...] We further observe an inverse relationship between affiliate marketing use and content complexity, and that all search engines fall victim to large-scale affiliate link spam campaigns.

I think this is an excellent methodology for testing the quality of search results. I would love to see a standard search engine test and scoring system based on this, maybe similar to some of the LLM scoring systems.

View on HN

This doesn't apply to the content complexity finding, but the finding that "product reviews which are in top search results are more likely to contain affiliate links than product reviews which are not" can also be explained by the fact that if and only I am getting a bunch of hits on my product reviews, I'm incentivized to monetize that with affiliate links.

We should also ask ourselves if affiliate links are really that bad. Someone could be making honest complete reviews and monetizing those with affiliate links, does that inherently mean that the search results are lower quality?

That approach also misses all the copied-a-github-issue low-effort content that seem to crop up on Google.

It's bad because most of the reviews will be for items you can only get on Amazon or another retailer with an affiliate link program and not for items available elsewhere.

I've seen so many "Best $WHATEVER of $YEAR" and it's really "Best $WHATEVER $PRODUCED_BY_WEIRDLY_NAME_FLY_BY_NIGHT_COMPANY_ON_AMAZON $YEAR"

How can an affiliate link not be in conflict with an unbiased review? They don't get paid if they don't recommend the thing.

In theory, if we had truly open marketplaces that offered equal monetization of all links to all similar products in the same category regardless of the price offered to the consumer, sure.

In practice, that is never the case. At best you're getting reviews that only compare products available in the same marketplace (say, for example, all the table saws you can buy on Amazon). At worst, you're getting reviews from vendors who offer the highest payout.

Also, a heck of a lot of consumer products are absolute garbage. (IMHO, most of them, but others may feel differently.) Who is going to write an honest, scathing review of a product and then monetize the link to it? Why even bother?

> that inherently mean that the search results are lower quality?

Almost? There doesn't need to be anything nefarious going on, but human beings align to incentives, basically always. So the trick with reviews is going to be to avoid any incentives coming from the product side, and embrace those coming from the consumer side.

Anecdotally, I can say that I’ve definitely noticed a negative correlation between affiliate marketing and content quality, though I would be very interested in seeing a formal study of it.

Affiliate links create misaligned incentives between content creators and consumers. This includes on the products listed themselves, where the affiliate is incentivized to select the products that give the most kickback rather than the ‘best’ for the consumer. But it also includes the content itself. Affiliate links create incentive to (a) write reviews, where a site wouldn’t have bothered before; (b) churn out lots of content with lots of links, that can be picked up by search engines; and (c) to not invest much in the actual reviews, and instead generate quick, low-quality content, since the content creator doesn’t actually care about finding the “best” product (which takes lots of time and money to do correctly), only on having readers click their links.

This is why sites with names like Celeb Rumor Central have dozens of articles like “top 10 coffee makers 2024”. They just hire a freelancer to do a few quick Google searches for coffee makers, then churn out 10,000 word articles with sections like “the history of coffee” and “why do people drink coffee?” Then the actual review is “this machine is purported to boil water and drip it through coffee grounds, and has high ratings on Amazon (we may earn a small commission when you click a link on our site)”. Increasingly, they’re just using AI to do it, cutting out the cost of a freelancer.

In my personal opinion, affiliate marketing is one of the worst things to happen to the modern web, and the source of a ton of content farm SEO spam pages.

Sure but I feel like you are taking game theory and applying it to real life. See for example https://www.rtings.com/monitor/reviews/best/by-usage/gaming they do a lot of actual reviews with tests and all their links are affiliated. If as a Google user I want a review of Gaming Monitor, having this as the first link would likely be an acceptable-to-good scenario.

There is a lot of money to be captured in paid search with these sites. There are some pretty large players in the space who just built these types of top 5 or top 10 pages for tons of verticals and then bid on keywords related to those verticals. They can go to brands and demand significantly higher fees than most affiliates would get and the brands will usually pay because that acquisition is usually more efficient than try to bid on the terms the affiliate is bidding on.

Back when I was still involved in acquisition marketing, I did a test where I killed all paid search and built relationships with the affiliates to negotiate our "ranking" higher up the page. It was a huge boom for business. We scaled paid acquisition at a profitable CAC, which was very difficult to maintain, let alone scale, in bidding directly and it was significantly less work to manage.

Consumers value this type of content quite a bit, even if many are skeptical of the quality. Sometimes it's nice to just see a pared down list of things with even a cursory rundown of features/differences.

Forgive my naivety, but wouldn't a simple way for a search engine (like Kagi) to avoid falling victim here to detect affiliate link programs? There's got to be a small handful of patterns for affiliate link tracking:

1. Domain Interception & HTTP redirects 2. Tracking codes embedded in the URL directly

It looks like Kagi does this[1]:

> Kagi surfaces shopping results featuring unbiased reviews and no affiliate links to help you identify the best product across categories. Top results include discussions focused on helping you find the best item to purchase - you are not bombarded with affiliate links and ads. Continue to scroll and you will see product comparisons across multiple vendors so you can pick what best suites you. Kagi's shopping search will always return a detailed discussion of which product to buy not a competition amongst advertisers promoting where you should buy. Kagi is focused on providing you the best results to make an informed decision not polluted by affiliate links and advertisements.

1. https://help.kagi.com/kagi/features/shopping.html

Kagi is better than Google, but just yesterday it returned the top link to Amazon-linking “check the price for this item on Amazon right now” affiliate spam. So it’s not perfect.

Oh, OK. I thought it was perfect.

I think it’s pretty good. I understand sometimes we root for the good guys like Kagi and look at their work with rose tinted glasses and even think they’re perfect, like you did, but it’s not really feasible.

A pattern I’ve noticed from people not working in tech, or more junior people in tech, is that they have a thinking that software can ever be perfect. If either of those categories describe you, I can tell you from experience no software is ever perfect!

In adversarial environments, this is true. In non-adversarial environments, software of a certain defined scope can at least get very close to perfection.

Ultimately a site can get around affiliate / referral detection with internal links that redirect though I suppose. So the crawler would need to follow and track all links to detect this. Also sites will surface different links / content based on user agents, so I don't know how effective this is.

Maybe there is room for one search engine that does this - for paying customers or as an investment to be better

Kagi is a paid search engine to avoid the manipulation that Google welcomes. To avoid every trick and targeted tricks it may be cost prohibitive. Fortunately Kagi isn't popular enough yet for targeted manipulations to have a noticeable impact.

I tried a search and found a site with amazon affiliate links in their featured "Shopping" snippet

https://i.imgur.com/vPRpBYJ.mp4 (yes the recording is very broken but you can still see)

Maybe the affiliate stuff is still there but ranked lower?

Sure but it sounds to me like it shouldn't be featured in the "Shopping" widget when it has 9 affiliate links?

When I hover over the "i" icon it says:

> Information provided by Looria, Amazon, Reddit and other sources. Shop confidently with Kagi - our results have no affiliate links.

The sites featured in the "Shopping" widget were:

- Reddit

- RTINGS: has 8 affiliate links (just from one article I picked, it was featured several times)

- NYTimes: has 47 affiliate links (just from one article I picked, it was featured several times)

- Techgearlab: has 9 affiliate links

Even if you ignored the widget, the first result was a site with 7 amazon affiliate links.

On second read, I think what the Kagi doc means is that they don't provide first-party affiliate links (i.e. if you click on an Amazon.com search result it's not using their affiliate). They will still show you search results that have affiliate links embedded (but presumably don't affect or negatively affect ranking).

I am working on a search engine that checks a site for affiliate links from known providers and demotes sites that have a large amount of them.

In addition:

* it demotes sites with popups (think newsletter sign ups)

* it demotes sites that block (or complain about) ad blockers

* it demotes sites with a high number of ads and favors sites with no ads

* it demotes sites using certain sketchy ad companies.

* It demotes sites that have paywalls

* It detects possible link networks and flags them for human review/removal.

* sites with RSS feeds get promoted.

* There is a toggle to hide all sites with ads or external trackers, but it is still WIP (The whole project is).

There are many other features. No idea if I am going to make it public, I created it to update my skillset. I actually thought about setting up a nonprofit and making it open source, but I haven’t decided.

Very interested, any way to follow you?

Me too.

What would be the difference between an affiliate link program and a web ring?

I don't know if discovery is actually a bottleneck to be automated away. It might be the fun part. I'm thinking back to the Napster approach where you could browse other people's libraries for music ideas.

Currently, Kagi has (if you hover/click the shield icon to the right of a result) an indication of the information it knows about a website (as well as a way for you to rank it higher or lower for yourself).

One of these is "ads/trackers". I imagine that it would be feasible for this to include some of the more common affiliate URL types, or third party lead/affiliate tracking bounce hops like awin.

Clearly there will always be some amount of ability to "defeat" this kind of measure by obfuscating links, but eventually the user needs forwarded to a link that has a referral parameter or a site that sets an affiliate cookie or similar.

The "tracker category" also can give a bit of extra information - things like "invasive fingerprinting, advertising"

I don't know if this would be a very long-term solution if the big ones (ok, google) did it. Advertisers would catch on very quickly, and some legitimate review sites which might get funding through affiliate links unrelated to the product being reviewed would lose out to straight-up paid for "reviews" that are funded wholly by the manufacturer and just don't use affiliate links.

this is one way to do it, but I wouldn't say its sufficient. If I search for 'things to do in Seattle', you get many 'blogs' and such that a writer gets paid by sources to insert their place into the things to do list. I didn't word that well, so for example: I own a coffee shop, I pay them moneys, and the '25 things to do in Seattle' writer puts my coffee shop in the list.

If I do an image search for the word 'strawberry', how many of those results are not stock images, images from a store, etc. of a strawberry? can you find an actual picture of a strawberry sitting in the wild? or just some picture of a strawberry a person uploaded without trying to sell you something?

Then the search engines will game the benchmarks. A different definition of "Search Engine Optimization" I suppose.

I know that tends to happen for LLMs but I don't know if Google/MS would care enough about some obscure benchmark system to try to game it.

Their entire business is gaming metrics. Most of those metrics are ones they invented for themselves, but if there's any advantage at all, they'll game it

Reply on news.ycombinator.com