Affiliate marketing in 2026: how it works for brands and publishers, networks, commissions and compliance

Affiliate marketing pays partners a commission only when a tracked click turns into a sale or a qualified lead. US advertisers will put about $13.81 billion into it in 2026 and get roughly $241.03 billion of ecommerce sales back, on eMarketer’s forecast. But the economics shifted this year, and Amazon is the reason.

How affiliate marketing actually works

Affiliate marketing is a revenue share agreement dressed up as an advertising channel. A brand publishes an offer, a partner promotes it through a tracked link, and if a reader clicks and buys within an agreed window, the partner earns a percentage of the order or a fixed bounty. Nobody pays for the impression. Nobody pays for the click. Payment lands after the transaction clears, which is why finance teams like it and why it is the last channel to get cut.

Four parties sit in the chain. The advertiser (or merchant) runs the programme. The publisher (or partner, or creator) sends the traffic. The network or software-as-a-service platform issues the tracking links, deduplicates orders, validates them and handles payment. And an agency or in-house programme manager recruits partners, negotiates rates and polices the whole thing. On smaller programmes the last two collapse into one person with a spreadsheet, which works until the first fraud incident.

Almost every argument in affiliate marketing comes down to one question: which touch gets credited. That is a measurement problem, not a partnership one, and our marketing analytics guide goes deeper on it.

The numbers that frame 2026

Affiliate is growing faster than the ecommerce it sells into. eMarketer has US affiliate spend at $13.81 billion in 2026, up 11.3% from $12.42 billion in 2025, while US retail ecommerce grows 6.7%. That gap explains the channel’s popularity with CFOs. When budgets tighten, the line item that only pays on results survives the meeting.

That last number is the uncomfortable one. A channel worth $241 billion in tracked sales is invisible to nearly half the people planning media. eMarketer also found 27.3% of marketers using mix modelling fold affiliate into a generic performance bucket, and 14.8% leave it out of the model entirely. So affiliate gets judged on last-click reporting, which flatters it, or ignored in the model, which starves it.

Commission models and what to pay

Cost per sale is the default: a percentage of order value, paid after a validation period. Cost per acquisition pays a fixed fee for a defined event, which suits purchases where order values swing wildly. Cost per lead pays for a verified signup and is standard in finance, insurance and education. Hybrid deals bolt a flat placement fee onto a percentage.

Rates vary by margin, not by sector prestige. Physical goods with warehousing and returns rarely support more than mid-teens. Software with near-zero marginal cost can pay 20% or 30% recurring and stay profitable. The mistake I see most often is a brand copying a competitor’s headline rate without checking whether its contribution margin can carry it.

On typical rates, the honest answer is that good public data is thin. Vendor analyses circulating in 2026 put median ecommerce commissions at roughly 8.4% of order value, software around 20% of first-year revenue, and travel near 4.2%, with digital products far higher because there is nothing to ship. Treat those as orientation, not benchmarks.

Networks and platforms after the 2026 consolidation

The infrastructure layer moved more in eighteen months than in the previous decade. Awin finished folding ShareASale into a single platform, with the legacy ShareASale system shutting down on 6 October 2025 and the combined network claiming more than 9,500 advertisers and 250,000 active publishers. Then on 28 April 2026, Rakuten and impact.com announced an alliance pairing Rakuten Advertising’s managed services and Rakuten Rewards’ cashback base with impact.com’s contracting, tracking and payments platform.

“Partnership marketing has become one of the most important growth channels for modern businesses, but the systems supporting it have remained fragmented.” David A. Yovanno, CEO, impact.com, 28 April 2026.

Two consolidations in a row narrows real choice for mid-market brands, and raises the switching cost: migrating a programme means re-onboarding every partner, re-issuing links and losing reporting continuity. Ask any prospective platform for its data export format before you sign.

What Amazon did to Associates, and why it matters

This is the single biggest affiliate story of 2026 and it was never formally announced. Adweek reported on 18 May 2026, after speaking to seven publishers and partners, that Amazon had cut Associates commission rates by up to 50%. eMarketer confirmed the reporting on 19 May 2026: categories that previously paid up to 10% dropped to 4% or 5%, the rollout began in Asia-Pacific in late 2025 and reached US publishers in March 2026, and the Associates operating agreement was quietly updated on 14 April 2026.

The rate cut is not even the worst part. Milestone bonuses and year-on-year performance bonuses were removed for most partners. Reporting was thinned out: the sales threshold for tracking-ID-level data was raised, SKU and ASIN level reporting disappeared, and access to some premium APIs was revoked. And commission narrowed to the linked product and its variants, so a reader who clicks your kettle recommendation and buys a toaster instead now earns the publisher nothing. One deal-site publisher told eMarketer they expect 2026 Amazon revenues to come in 50% below plan.

Google’s AI overviews collapsing organic traffic at the top of the funnel and Amazon paying less at the bottom makes the conversion “super challenging.” Andrew Perlman, CEO, Recurrent Ventures.

That is the squeeze in one sentence, and it explains why every commerce publisher spent 2026 diversifying. If Amazon is more than about a third of your affiliate revenue, that is now a concentration risk rather than a convenience. Direct brand programmes pay more, convert worse, and require actual relationship work. Most publishers who moved found the blended rate improved anyway.

Publisher types and the incrementality problem

Not every partner does the same job, and paying them the same rate is how programmes waste money. Content publishers and creators write and film the reviews that create demand. Coupon and cashback sites intercept demand at the last moment. Technology partners sit inside the checkout or the browser and take a cut of what passes through.

eMarketer’s data on where the money actually lands is blunt: cashback, loyalty and rewards platforms took 35% of affiliate ad spend in 2024, discount and coupon publishers accounted for 42.4% of revenues in the first half of 2025, and content publishers took just 16% of spend. Technology partners have grown from around 5% of budgets in 2021 to about 7%. So the majority of a typical programme’s payout goes to partners who appear after the customer has decided to buy.

None of that makes coupon or cashback partners worthless. They keep voucher-code searches from sending a customer off site, and cashback platforms genuinely drive new customers in insurance and utilities. But the default should be a lower rate for interception and a higher one for creation, and only a holdout test proves which is which. Our conversion rate optimisation guide applies the same discipline on site.

Creators are the group to fund next. On Awin’s network the creator share of revenue rose from 15.9% to 19.5% year on year. On TikTok Shop, affiliate creator content drives about 42% of US GMV, according to platform analyses collected in 2026. The overlap between influencer deals and affiliate deals is now almost total, and the practical differences (rights, briefs, exclusivity, whitelisting) are set out in our influencer and creator marketing guide.

Tracking and attribution without cookies

Affiliate tracking was built on third-party cookies and has spent six years migrating off them. Safari has capped cookies set through JavaScript at seven days since February 2019, and since 24 March 2020 it has blocked cookies for cross-site resources by default, along with deleting script-writable storage after seven days without interaction. Firefox does something similar. That is roughly a third of web traffic where a browser-cookie programme is guessing.

The fix is not clever cookie handling. It is moving the identifier server side. The network appends a click ID to the destination URL, your server reads it on landing, stores it against the session in your own database, and fires a server-to-server postback when the order confirms. No browser storage in the critical path. Most major platforms support this pattern, and most mid-market programmes still have not implemented it because it needs developer time rather than a tag manager change.

Three practical consequences. Attribution windows longer than seven days only work with server-side tracking, so if you are running a 30-day window on client-side cookies you are quietly under-crediting publishers. Mobile app conversions need their own deep link and postback path. And once you have click IDs in your own database, you can finally join affiliate orders to customer lifetime value instead of arguing about last click.

The rule here is the rule everywhere in 2026: attribution tells you the order of touches, incrementality tells you whether the spend caused anything. Geo holdouts and partner-level pause tests are cheap. Media mix models rarely handle affiliate well, which is why 27.3% of modellers dump it in a generic performance bucket.

Google’s site reputation abuse policy and affiliate content

Google’s spam policies define site reputation abuse as publishing third-party content on a host site “mainly because of that host’s already-established ranking signals”. That is what hit the big publisher commerce hubs from 2024 onwards. The policy has stayed manual action only, which means a human reviewer decides, and recovery requires the pattern to actually change rather than a slight rewrite.

The detail worth reading properly is Google’s own coupon example. Coupon pages run with a specialist provider are unlikely to be actioned when the section sits in a fully integrated subfolder, is organised into publisher-specific categories, discloses editorial responsibility, is cross-referenced from other editorial content, and is navigable from the main pages. Miss those conditions and you are running a separate business on someone else’s domain authority, which is what the policy targets.

Separately, Google’s thin affiliate policy has not changed and still catches most low-effort affiliate sites: publishing product content where descriptions and reviews are copied from the merchant with no original content or added value. The stated contrast is original reviews, genuine comparisons, real testing and price navigation. That is a workable specification, not a vague warning, and it maps closely to the E-E-A-T work described in our SEO guide for 2026.

Ranking volatility across affiliate and commerce content has been heavy through 2026, with core and spam updates landing through the year. The month-by-month detail sits in our SEO news for September 2026.

AI search, shopping agents and where the click goes next

Here is the structural threat. Affiliate publishers built businesses on comparison, deal hunting and recommendation. A chatbot does all three well, and when the answer arrives inside the chat there is no tracked click and no commission. eMarketer noted shopping queries on ChatGPT grew faster than any other query type between December 2024 and June 2025, and that genAI users reach for it to compare prices (54%), find deals (41%) and check reviews (41%).

But there is a counter-signal that most doom coverage skips. AI assistants have to cite something, and what they cite is often affiliate content. In one brand analysis eMarketer highlighted, nearly 70% of the sites cited when ChatGPT discussed the eyewear brand Zenni were affiliate pages. More than a quarter of OpenAI’s publisher content partnerships since 2021 have been with companies running scaled affiliate commerce operations. Comparison content is not obsolete. Its distribution changed.

What that means in practice: the page still has to exist, it has to contain the specifics an assistant can lift (prices, specs, tested results, dates), and it has to be structured so a machine can extract a claim without reading marketing prose. Getting cited is a different discipline from ranking, and it is covered in our AI in digital marketing guide and, on the commerce side, in the ecommerce and retail media news for September 2026.

Monetisation inside the assistants is not settled. OpenAI has no public affiliate programme with a published commission schedule or cookie window, though Sam Altman has discussed roughly 2% on purchases completed through ChatGPT. Analyst Yoram Wurmser’s read for eMarketer was that OpenAI is “clearly setting up a structure that’s ideal for affiliate fees”. Plan for it. Do not budget for it yet.

Fraud and where programme money leaks

Affiliate fraud data is genuinely bad, in the sense that almost all of it comes from vendors selling fraud detection. Figures circulating in 2026 put losses at 8% to 15% of gross programme spend, with cookie stuffing misattributing 5% to 15% of ecommerce commissions. Treat those as directional. The behaviours behind them are real, and you can check for each one yourself.

Cookie stuffing drops tracking cookies for users who never clicked, usually through hidden iframes or extensions, so the partner harvests sales it had nothing to do with. Brand bidding means partners buying your own brand name in paid search, often as “brand coupon” or “brand discount”, so you pay commission on a customer who was typing your name anyway. Toolbar partners fire at checkout and take last click from every other channel.

The defences are dull and effective. Ban brand bidding in the terms and monitor it weekly with a competitor tool. Check click-to-conversion time by partner: conversions clustering within seconds of the click mean stuffing. Require manual approval for new partners, and actually look at the site. And audit what fires at checkout, because the technology partner sitting in your basket page is invisible in most reporting. Brand protection in paid search is covered in our PPC guide.

Disclosure and compliance in 2026

Disclosure law did not change much in 2026. The FTC’s Endorsement Guides were revised in 2023, and the agency’s own guidance is specific about affiliate links: tell readers “I get commissions for purchases made through links in this post”, and remember that “the closer the disclosure is to your recommendation, the better”. The FTC treats “paid link” next to the link as adequate and considers a bare “affiliate link” tag or a “buy now” button insufficient on its own.

Two points brands get wrong. The disclosure has to be seen before or as the link is seen, not after the click. And liability is shared: an advertiser cannot outsource disclosure to partners and then claim ignorance. If your terms do not mandate wording and nobody spot checks partner pages, you own the risk.

Outside the US, the UK’s Advertising Standards Authority applies the CAP Code to affiliate content, and India’s Advertising Standards Council of India has run influencer advertising guidelines covering material connections since May 2021. New this year: the EU AI Act’s transparency obligations under Article 50 began applying on 2 August 2026, requiring that synthetic audio, image, video and text output be marked in a machine-readable format and that deepfake content be disclosed. If your partners are producing AI-generated product imagery or voiceovers for the EU market, that is now a programme compliance question, not a creative preference.

If you are the publisher: what still works

The publisher side of affiliate got squeezed from both ends in 2026: less search traffic at the top, lower rates at the bottom. What survives has three characteristics.

First, original testing. Google’s own thin affiliate definition rewards rigorous testing and real comparison, and an assistant citing sources prefers pages with specifics it can quote. If your review could have been written without touching the product, it is now worthless twice over.

Second, direct programmes rather than a single marketplace. Amazon’s cuts hurt most where Amazon was the whole business. Brands running their own programmes on Awin, impact.com or CJ frequently pay double the rate, and will negotiate a new-customer bonus if you ask. Smaller traffic, higher revenue per session.

Third, an owned audience. Newsletters, subscribers, a community. Search referrals fell hard across publishing through 2025 and 2026, and the sites that held up had a channel Google does not control, an argument made at length in our content marketing guide.

And a fourth thing, less comfortable: cost base. If your model needed 100 articles a month to work, it does not work now. Fewer pages, more depth, more updating. The affiliate sites still growing in 2026 publish less than they did in 2023.

A 90-day plan for a programme that needs fixing

This assumes an existing programme, a network in place, and someone with three or four hours a week. Do it in order. The first month is diagnosis, and that is the part everyone skips.

Two things to resist. Do not announce a relaunch before the tracking fix, or you will onboard partners onto broken measurement. And do not cut the headline rate across the board: the partners who leave first are the ones with alternatives, which are usually the good ones. Where affiliate sits next to every other channel is mapped in our complete digital marketing guide, and the store-side work it depends on is in the ecommerce marketing guide.

Last researched and updated: 7 September 2026.