Know what 493 affiliate
programs actually pay
The headline percentage is the least useful number in the deal. Whether recurring beats a one-time commission depends on how long the program pays it, and how long the customers you refer stay subscribed. 20% recurring overtakes 50% once at its third commission payment. A 7-day cookie means the reader who thinks it over earns you nothing. A $250 payout minimum means you may never get paid at all.
Commission, recurring or one-time, cookie window, minimum payout, payout method and which network runs it — for every program listed, and not stated wherever the vendor doesn't publish it.
Read off the vendor's page
Not a roundup, not another directory. Every record carries the URL the figure was printed on, so any number is one click from being checked.
Dated, because terms change
A rate without a date is a guess with a number on it. Every record shows when it was last read, and a re-check tool reports drift rather than quietly patching it.
Not stated is a finding
39% of material fields are not published by the vendors at all, and 38 state no rate anywhere. Knowing who is silent is worth as much as knowing who pays.
What vendors leave out
The commission rate is the one almost everyone publishes. The terms that decide whether the money ever reaches you are the ones they leave out. What each of them means.
Connect your AI agent
https://affiliateprogramterms.com/mcp
Look up terms
Ask which programs pay recurring, or what a vendor's cookie window is. Each answer comes with the vendor's page and the date it was read.
Submit a program
Your agent sends a vendor's page to the same queue as the form. It is checked against the page before anything is published.
Book a sponsored card
Your agent drafts the card and hands you a link to pay. No tool pays for anything.
A taste of the free 50
| Program | Commission | Type | Cookie | Min. payout | Runs on |
|---|---|---|---|---|---|
| ActiveCampaign2/5 Email & newsletters | 20–30% of subscription value, rising with new business and retention | Recurring — 12 months | not stated | not stated | PartnerStack |
| Anymail Finder5/5 Email & newsletters | 30% | Recurring — 12 months | 60 days | $100 | not stated |
| Apollo.io4/5 CRM & sales | 15% on monthly plans, 20% on annual | Recurring — 12 months | 90 days | not stated | PartnerStack |
| Backendless3/5 No-code | not stated | Recurring — 12 months | 60 days | $50 | not stated |
| beehiiv4/5 Email & newsletters | 50–60%, tiered: 50% Bronze, 55% Silver, 60% Gold | Recurring — 12 months | 60 days | not stated | in-house |
Search and filter the whole directory. 50 are free to read in full; the other 443 are in the downloadable file.
How this directory is built
Affiliate terms are scattered across vendor pages, disclosed inconsistently, and changed without notice. A rate with no date and no link cannot be checked by the person reading it. This one is built so that every number can be traced back to the page it came from.
Every figure is read off the vendor's own page
Not a roundup, not a blog post, not another directory. The source URL on every record is the page the figure was printed on, and checkedOn is the date it was read. Where a vendor publishes nothing, the field says not stated — which is itself worth knowing before you build content around a program.
Pages are found by sweeping, not guessing
A program page can sit at /affiliates, /affiliate-program, /partners or a dozen other paths. Each candidate domain is swept across all of them and the results scored, because guessing one URL and believing the 404 is how a directory ends up with gaps it cannot see.
/affiliate-marketing is the trap: it reads like the most obvious path of the lot and is where vendors put marketing copy rather than terms. Four more paths won nothing at all in at least 988 tries each and were dropped.
A page that scores well is still read by a person
Scoring finds candidate pages; it does not read them. A vendor's own blog post explaining what affiliate marketing is scores higher than most real terms pages and contains none — /affiliate-marketing was swept 1192 times and yielded 13 usable pages. So does a reseller hub, and a vendor's own affiliate-campaign product. Every record here was written by reading the page rather than pattern-matching it, and roughly one candidate domain in four ends up with a publishable record.
Submissions are checked against the page, not taken on trust
Anyone can submit a program, free. What they send is a URL — not figures. Any figure a submitter volunteers is checked against the vendor's page and discarded; what gets published is what the page says. A claim that is not on the page is rejected outright, which is what stops a free form being used to plant a number.
The build refuses to publish an unsourced fact
A validator runs before every build and aborts it on a missing source, a missing date, a date in the future, a category outside the list, or a record that contradicts itself. No figure reaches this page by being typed in.
The data is re-read, and drift is reported rather than patched
Terms change. A re-check tool re-reads every source page and reports which published figures are no longer on it — it never changes a figure, because a tool that repairs itself produces wrong numbers instead of no numbers. A record's date moves forward only when its page was read, every figure on the record is still on it, and a second reader — a language model asked a yes-or-no question about each figure — agrees; anything else waits for a person. The last full pass, on 24 Sept 2026, re-read 411 source pages (2 more would not load); 408 still carried their published figures. 22 records have been added since, each read once, on the date it carries.
What the data says about the vendors
Five terms decide what an affiliate actually earns: commission rate, recurring or one-time, cookie window, minimum payout and payout method. A vendor either states each one on its own program page or it doesn't.
Across the 493 programs here, 39% of those are not published at all, and 38 state no commission rate anywhere. That share has held roughly steady as the directory grew — 41% at 38 records, then 37% at 124 records, then 37% at 219 records, then 41% at 351 records, now 39% at 493 — so it is a property of how vendors publish, not an artefact of which ones got collected first. The guide breaks all of it down.
terms stated, out of 5
79 of the 493 publish all five. Most of the directory sits in the middle, telling you some of what you need and leaving the rest to be found out after you join.