Theory Roadnewbusiness@theoryroad.com(866) 3-THEORYGet in touch

Half of Affiliate Link Clicks Are Bots: 349,893 Clicks Analyzed.

We logged every outbound affiliate click behind our own redirect on about 30 websites we own and operate for four months, then classified each one. Half announced themselves as bots. Most of the rest failed simple checks. Here is the funnel, who the bots are, and what it does to your conversion math.

By November 18, 202612 min read

Want this done for you?See what Theory Road runsget a free plan

Of 349,893 outbound affiliate link clicks we logged between June 14 and October 17, 2026, 49.7% came from declared bots, and only 13.9% passed three basic human checks. Just 3.3% passed those checks and came from the United States. The clicks came from about 30 websites we own and operate, through our own first-party redirect, from 86,885 distinct IP addresses. No client sites are in the data.

If you run affiliate links, the click count in your network or Associates dashboard is probably not a count of people. Ours was mostly software: SEO tool crawlers mapping links, AI company user agents fetching pages, social platform preview bots, and a large block of clicks with no referring page at all. This study shows how we sorted them, what was left, and how to apply the same checks to your own links. We call the method the three-check rule: a click counts as plausibly human only if it does not declare itself a bot, runs a current browser build, and arrives from one of your pages or an ad click.

Method.

  • Source. A first-party click log behind our /go redirect. Every affiliate link on our sites points to /go/<offer>, which records the click and forwards the visitor to the merchant. The log stores time, user agent, referring page, ad click ID if present, country and IP.
  • Window. June 14 to October 17, 2026.
  • Sites. About 30 websites we own and operate: content, review, ecommerce and service sites.
  • Volume. 349,893 logged clicks from 86,885 distinct IPs, an average of about 4 clicks per IP.

We classified each click with these rules, applied in order. A click lands in the first rule it matches, so no click is counted twice.

Declared bots and tools.
The user agent names a crawler, tool, script or headless browser (SEO tools, AI company agents, social platform fetchers, search engine crawlers), or is empty.
Out-of-date browser builds.
The user agent claims iPhone OS 10 to 13 or a Chrome version below 110. Those builds are years old. A small number of real people still run them, but they are a common disguise for scripted traffic.
No referring page and no ad click ID.
A person who clicks a link on one of our pages sends that page as the referer under the browser's default referrer policy. A click with no referer and no ad click ID did not come from a page view we can see. Some privacy browsers and in-app browsers strip the referer, so we call this group unattributable and very likely not a person, rather than certainly a bot.
Remaining.
Everything that passed all three checks. We then split it by country.

The click funnel.

Outbound affiliate clicks by classification step, about 30 owned sites, June 14 to October 17, 2026
StepClicks removedShare of all clicksClicks leftShare left
All logged clicks00%349,893100%
1. Declared bots and tools173,77349.7%176,12050.3%
2. Out-of-date browser builds18,6785.3%157,44245.0%
3. No referer and no ad click ID108,91031.1%48,53213.9%
4. Remaining clicks outside the US36,90810.5%11,6243.3%

Of the 48,532 clicks that passed all three checks, 24.0% came from the US. 41,547 (85.6%) were desktop and 6,985 (14.4%) were mobile. The desktop share stands out on sites whose readers we would expect to skew mobile. Our checks remove the obvious automation, not bots that copy a current browser and send a referer. Treat the 13.9% as a ceiling on human clicks, not a count of them.

3.3%.of 349,893 logged affiliate clicks passed all three checks and came from the US

Who the bots are.

Declared bots tell you who they are in the user agent string. We grouped them by family.

Declared bot and tool clicks by user agent family, June 14 to October 17, 2026
User agent familyClicksShare of all logged clicks
SEO tool crawlers (Ahrefs, Semrush, DataForSEO, Majestic and others)112,88032.3%
Anthropic (Claude) user agents28,3328.1%
Meta10,4353.0%
ByteDance8,5132.4%
Other declared bots7,0122.0%
Amazonbot2,0610.6%
Bing1,5480.4%
OpenAI (ChatGPT) user agents1,1260.3%
Google crawlers9180.3%
Scripts and headless browsers5010.1%
Perplexity2290.1%
Applebot1940.1%
Common Crawl22under 0.1%
Empty user agent2under 0.1%

Each declared click is counted once, in the first family it matches, so the rows add up to the 173,773 declared clicks in step 1.

  • SEO tool crawlers. The largest group by far. Link indexes crawl every outbound link they find to map who links to whom. Following a /go link is how they learn where it points. Each follow is logged as a click.
  • AI company user agents. Anthropic, OpenAI and Perplexity user agents together produced 29,687 logged clicks. Common Crawl, an open web archive, added 22. Some of these fetches are crawlers. Some happen when a person asks an assistant about a page and the assistant fetches the links on it. Either way, the fetch is not a shopper reaching the merchant.
  • Social platforms. Meta and ByteDance user agents fetch links to build previews and check destinations. That traffic follows links shared or seen on their platforms.
  • Search engines. Google, Bing, Amazon and Apple crawlers made up a small share. Together they were about 1.3% of logged clicks.

Clicks by month.

Affiliate clicks by month and classification, June 14 to October 17, 2026
MonthLogged clicksDeclared botsOld buildsNo refererRemainingRemaining USRemaining US share
June 14 to 3049,85241,525 (83.3%)8396,0981,390 (2.8%)390.1%
July106,82775,023 (70.2%)5,68313,98812,133 (11.4%)1,5981.5%
August58,71432,819 (55.9%)2,41714,0519,427 (16.1%)2,0573.5%
September89,73120,004 (22.3%)5,47752,11612,134 (13.5%)3,7634.2%
October 1 to 1744,7694,402 (9.8%)4,26222,65713,448 (30.0%)4,1679.3%
Total349,893173,773 (49.7%)18,678108,91048,532 (13.9%)11,6243.3%

Read the declared-bot column with care. Its share fell from 83.3% in late June to 9.8% in October mainly because we added user agent filtering to the redirect on more sites over the summer. Filtered clicks still redirect, but they are no longer logged, so they drop out of later months. The bots did not go away. We stopped recording them.

The no-referer group moved the other way. It grew sharply in September, to 52,116 clicks and 58.1% of the month, and stayed above half of October's log. As the declared bots were filtered, a large block of clicks with no referer and no ad click ID took their place. Since October 18, 2026, our redirect also keeps old-build and no-referer clicks out of the log and sends them to an untagged destination. The click still works for anyone who is real. It just does not carry our tracking tag.

The non-US share.

186,876 of all logged clicks, 53.4%, came from outside the United States. That matters most for Amazon. Our Amazon tags are US Associates tags, which work on Amazon.com. A visitor in another country who lands on Amazon.com is unlikely to buy there, so every non-US click on a US-only link inflates the click count with little realistic path to a commission.

Among clicks that passed all three checks, the US share was 24.0%. So even after removing the obvious bots, three in four remaining clicks came from places our US tags could not earn in. If your sites draw real international readers, geo-routing to local stores is worth setting up. If they do not, a high non-US share is one more sign of automated traffic.

What this does to EPC and conversion math.

Earnings per click (EPC) and conversion rate both divide by clicks. If the click count is mostly bots, both numbers come out too low, and they come out too low by a different amount on every page, network and month. That leads to bad decisions: dropping offers that convert well with humans, or ranking products by a metric that mostly measures crawler interest.

Worked example (illustration, not measured data): the same orders against raw and filtered clicks
MetricDashboard clicksPlausibly human clicks
Clicks1,000139
Orders1010
Conversion rate1.0%7.2%
Earnings (example: $40)$40$40
EPC$0.04$0.29

The example applies our 13.9% pass rate to a round 1,000 clicks. Ten orders on 1,000 clicks looks like a weak 1% conversion rate. Ten orders on the 139 clicks that were plausibly human is about 7.2%. The orders and earnings did not change. Only the denominator did. Your own pass rate will differ, which is the point: until you know it, you cannot compare EPC across pages, sites or programs.

This is also why commission rate alone is a poor guide to which program to use. A higher rate on a program whose clicks are mostly crawlers can earn less per real visitor than a lower rate elsewhere. Our guide to Amazon Associates commission rates covers the rate side. Out-of-stock destinations distort the same math from the other end, as covered in affiliate links to unavailable products. And if your Associates report shows a large unexplained row, read the Amazon Associates Other row explained before blaming traffic.

How to filter your own affiliate clicks.

You need a redirect you control. A raw merchant link in your page gives you no log at all, and network dashboards count clicks on their own terms. With a first-party redirect in place, these checks cover most of what we found.

Two cautions. Do not block filtered clicks outright. Some real people use privacy browsers, old phones or apps that strip the referer, and a working link costs you nothing. Untagged is safer than broken. And do not try to hide links from crawlers with tricks that show search engines something different from what people see. Filtering what you log and tag is enough.

If you run an affiliate program as the merchant, the same logic applies to the clicks partners send you. Server-side conversion reporting is far harder to pollute than click counts. Our guide to affiliate postback URL setup covers how to pass conversions back to a network or tracker so you can judge partners on orders rather than clicks.

Limits of this study.

  • One operator. These are about 30 sites we own and operate. Many are young content sites, which crawlers visit heavily. Established sites with loyal audiences may see a much higher human share.
  • Changing filters. We added user agent filtering on more sites during the window, so later months under-record declared bots. The full-window shares mix filtered and unfiltered sites.
  • Imperfect rules. The no-referer rule catches some real people, and the remaining group still contains bots that copy current browsers. The 13.9% is an upper bound on human clicks, not a count.
  • No order matching. We did not join clicks to orders at the click level, so this study does not show conversion rates by class. The worked example is an illustration.

How to cite this study.

Theory Road, Affiliate Link Bot Clicks Study, November 2026, theoryroad.com/insights/affiliate-link-bot-clicks-study. Please cite the source and the window (June 14 to October 17, 2026, 349,893 logged clicks) when you use the numbers.

Questions about affiliate bot clicks.

What share of affiliate link clicks are bots?

In our log of 349,893 outbound affiliate clicks on about 30 websites we own and operate (June 14 to October 17, 2026), 49.7% came from declared bots. Another 5.3% came from out-of-date browser builds and 31.1% had no referring page or ad click ID. Only 13.9% passed all three checks.

Which bots click affiliate links the most?

SEO tool crawlers such as Ahrefs, Semrush, DataForSEO and Majestic produced 112,880 logged clicks, 32.3% of the total. Anthropic (Claude) user agents produced 28,332, Meta 10,435 and ByteDance 8,513. Google, Bing, Amazon and Apple crawlers were each under 1%.

Do bot clicks hurt my affiliate account?

Bot clicks inflate your click count and lower your reported conversion rate and EPC. Programs set their own rules on invalid traffic, so read your program's agreement and policies. Filtering what you tag, while still redirecting every click, keeps your numbers clean without breaking links for real people.

Why do clicks with no referer look like bots?

A person who clicks a link on a web page sends that page as the referer under the browser's default referrer policy. A click with no referer and no ad click ID did not come from a page view you can see. Some privacy browsers and in-app browsers strip the referer, so treat this group as unattributable and very likely not a person.

How do bot clicks change EPC?

EPC divides earnings by clicks, so bot clicks push it down. As an illustration, 10 orders on 1,000 dashboard clicks is a 1% conversion rate, but if only 139 of those clicks were plausibly human, the rate is about 7.2%, and EPC rises by the same factor.

Should I block bots from my affiliate links?

Redirect them, but do not tag them. Send filtered clicks to an untagged destination so a real person caught by a rule still reaches the product, and log tagged clicks separately so your EPC and conversion reports use plausibly human clicks only.

Get in touch

Let’s talk.

Tell us what you sell and what you want more of. Email newbusiness@theoryroad.com or use the form, and a senior strategist will be in touch.

Tell us about your business.

A senior strategist reads your inquiry personally.

Rather email? newbusiness@theoryroad.com or call (866) 3-THEORY.

Thanks. We have it.

A senior strategist will be in touch by email or phone.