The Receipt · Issue 01 · September 2026

Most sales from a recommendation never touch your link.

You recommended a tool on an episode. Two hundred people clicked. Four bought. You concluded the recommendation was weak, or the audience was not buying, and you moved on.

Sources: Podscribe Q1 2026, eMarketerread 2026-09-24

The measured evidence says the more likely story is that fifteen or twenty people bought, and you were paid for four. The rest bought the way people actually buy — later, from a search, on another device — and nothing you could see recorded it. This issue is about that gap: how big it is, why it exists, and the one thing that closes most of it.

The figure

4.7×

more sales found by matching purchases back to the audience than by counting promo codes and vanity links — across 97,000+ podcast campaigns in Podscribe's Q1 2026 benchmark.

Podscribe sells the matching, so it has every reason to make the gap look large, and its matched total includes modelled sales. Treat the level with suspicion. Treat the direction as settled: every independent look at podcast credit for the sale finds codes and links undercount, and none has found them overcounting.

What actually happens after you say "link in the description"

Someone hears you recommend a tool. A few of them click the link right then. Most do not. They finish the episode, and some days later — on a different device, at work, on a laptop instead of the phone they were listening on — they search the product's name and buy. Your link was never involved. If the program gave you a code and they remember it, you get paid. If not, the sale is invisible to you and credited to nobody, or to whoever's search ad they clicked.

Take the ratio at face value for a moment and put money on it. A publisher with ten thousand listeners sends 200 clicks and sees 4 sales on a program paying $58 each: $232. If codes and links are catching one sale in five, the episode produced something closer to 19 sales — about $1,100 — and $870 of it went unpaid, or was paid to whoever ran the search ad the listener clicked a week later. Halve the multiple to be safe and it is still $300 you earned and did not receive. Per episode.

This is not a tracking bug. It is what the tracking is built to see. A click records a single moment on a single device; a purchase decision by a real person spans days and devices, and only one of those moments is a click.

There is a newer version of the same problem, and it is growing. When someone asks an AI assistant what to use and it answers in the chat, eMarketer's phrasing is exact: "the tracked link that triggers an affiliate commission never occurs." Nobody clicked anything. But if the assistant knows your code — because it read it on your page, or the listener gives it — the code still goes on the order. The click is the fragile part. The code is not.

How the sales arrive

Clicked your link or typed your code — what you are paid for today~21

Bought later, from a search or a direct visit — found only by matching~79

Bought through an AI answer — not measured by anyone yet?

Of every 100 sales a recommendation produces, by how they were found. Derived from one source: if matching finds 4.7× what codes and links find, codes and links see about 21 in 100. Podscribe does not publish the split between clicks and codes, so they share a bar. The third row has no bar on purpose. And none of this was measured on creator traffic specifically — podcast ad campaigns are the closest anyone has come, which is why we are instrumenting it ourselves.

What to do differently

Ask every program for a code, and say it before the link
Most programs will issue one if asked, even when the sign-up page does not mention it. Then say the code first and the link second — “use CHRIS20 at checkout, or the link in the notes.” The code survives a week’s delay, a switch from phone to laptop, and a screenshot. The link survives none of those.
Judge a recommendation by sales over 30 days, not clicks over 3
Clicks are the sales you can see, not the sales you made. Before you drop a program for “not converting,” check its dashboard a month after the episode, not the week of. Late sales are the rule.
Read two lines of the terms: the window, and who wins a tie
How long a click counts for (7 days at some programs, 90 at others), and whether a typed code beats an earlier click by someone else. Those two lines decide more of your income than the commission rate does. They are on the program’s own page, and the offer pages here quote them where we have read them.

What this changed on Ancaster

We publish these because they change what we build, not just what we say. Three things moved this month because of this finding.

Sign-up asks for your code before your link →So the thing that survives is the thing we build around, from the first minute.The store plugin credits you by code, not by cookie →Shopify’s own /discount/CODE link puts your code on the order itself. A purchase a week later on another device still finds you.Every offer page shows what the program lets you do →Including whether paid traffic is allowed at all — which is next month’s issue.

Sources

Every figure in this issue, and where it was read.

4.7× more sales by pixel matching than by codes and vanity links: Podscribe, Q1 2026 podcast advertising benchmark, 97,000+ campaigns · read 2026-09-24; A measurement vendor’s own data; matched sales include modelled ones. The ratio is the stable finding.
“The tracked link that triggers an affiliate commission never occurs”: eMarketer, on in-chat AI shopping answers · read 2026-09-24; Quoted exactly. The commercial consequence — that codes survive where clicks do not — is our inference.
Codes and links capture roughly one sale in five: Our arithmetic on the Podscribe multiple — not a published figure · derived 2026-09-24; If matching finds 4.7× the sales that codes and links do, codes and links together see 1 ÷ 4.7, about 21%. Every dollar figure in this issue follows from that one division, so if the multiple is inflated, they are too.
The worked example: 200 clicks, 4 sales, $58 each: Illustrative. Commission from ElevenLabs’ published terms (22% of $22/mo × 12), read 2026-08-28 · constructed 2026-09-24; The click and sale counts are chosen to be typical of a ten-thousand-listener show, not measured. Change them and the proportion holds; the dollars move.

Next issue · December

One finding, the source behind it, and what to do about it. Sent when it’s ready — not on a schedule.

One email a month. Every issue is published here in full, for anyone, the day it is sent.