News · August 10, 2026
Claude still has no affiliate program, and that tells you something about AI tool reviews
Here is an industry fact worth sitting with: OpenAI, Anthropic, Google, and Perplexity, the companies making the models everything else is built on, run no cash affiliate programs for consumers. Zero commission for recommending them. The most important tools in the entire ecosystem pay reviewers nothing.
Now look at the average “27 best AI tools” listicle. The frontier models that a hundred million people use get a paragraph up top. Then page after page goes to subscription tools you have never heard of, each carrying a 30 to 50 percent recurring commission behind the link. Tool number 19 is not there because it is the 19th best tool in the world. It is there because the list is a payout table wearing an editorial costume.
The economics that write those lists
Run the numbers from the writer’s side and the shape of the content explains itself. A recommendation of Claude earns the site nothing. A recommendation of a mid-tier writing tool at 30 percent recurring earns roughly $5 a month per referred subscriber, forever. Multiply across a list ranking in search, and an honest ordering of tools costs the publisher thousands of dollars a month compared to a commission-ordered one. Most publishers, most of the time, do not pay that price. That is not a conspiracy; it is just incentives doing what incentives do when nobody discloses them.
The result is a review ecosystem where visibility correlates with payout rather than quality, and where the actual best tool in most categories, frequently a free tier of a frontier model, gets buried under fourteen subscriptions.
How to read AI tool reviews like an operator
Check whether the site recommends things it cannot earn from. This is the single fastest honesty test. A review site whose top picks are all monetizable has told you its ranking algorithm. One that puts unmonetizable tools at the top when they deserve it has at least demonstrated the capacity for an honest verdict. Our answer, for the record: Claude sits at the center of nearly every workflow we publish and pays us no commissions. At most, a referral invite earns account credits. The affiliate tools are the spokes, we mark every one, and the full disclosure is one click from every page.
Look for evidence of use, not descriptions of features. A feature list can be written from the marketing site. Screenshots of real projects, specific settings, costs actually paid, and complaints about specific failures can only be written by a user. If a review contains nothing the vendor’s homepage does not, it is the vendor’s homepage with a commission attached.
Distrust lists longer than about ten. Nobody has meaningfully tested 27 tools in a category. Long lists exist to catch long-tail search traffic and maximize the odds you click something monetized. Short lists with strong opinions are where testing actually happened.
Check the “skip it” advice. Honest reviewers tell specific people not to buy. If every tool is right for someone and no tool is wrong for anyone, you are reading a catalog, not a review.
The takeaway for operators building content businesses
There is a second lesson here if you are on the publishing side, and it is the one this site is built on. The picks-and-shovels layer around the big models is where the affiliate economics live, and that is fine, as long as the editorial layer stays honest about it. Recommend the frontier model on merit, monetize the toolchain around it, disclose everything, and let the occasional “this pays us nothing and it is still first” be the proof that the verdicts are real. Trust is the only durable moat a review site has, because the payout tables are the same for everyone.